Do you rank every horse in making an oddsline?

[bettheoverlay]

In another thread there was a discussion of value and making an oddsline for all horses that added up to 100%. I have read books that recommend this and I tried to do it, but soon gave up the idea as cumbersome, inexact and time consuming.

Now I give a betting odds line to my top 2 contenders. In large or highly contentious fields, I may go to 3 or 4. But my records show that after my top 2 selections, the winning % for the rest is so low its not worth the effort to decide if a horse should be 8-1 or 10-1. The losing streaks on such horses would be humongous.

I will also revise my odds line based on the tote board. In todays 9th at Belmont there was a Clement first time US horse who was 3-1 in the ML. I didn't know what to make of him, but when I saw he was getting bet in accordance to the ML, I revised upward my betting line on my 2 contenders.

I am curious to know if anyone actually assigns odds to every horse in a race and what success one has with that approach.

[Dave Schwartz]

I make a "probability" for every horse... although I only consider wagering on the "contenders."

Dave

[Lefty]

In his books, Dick Mitchell always stressed that he divided 80% of the line among his contenders and the non-contenders got the other 20%

[SLAM DUNK(MW)]

Been making an oddsline off and on for 21 years, so I have done so for quite a while. While I go through periods where I may just A B C my contenders and ballpark it, I always go back to lining races because it is(at least for me) the only way to know if I am getting value on my plays(yes I use 100%-I am not overly concerned with who is overbet or underbet, I want to know who offers value). Sure you can devise charts and methods such as contenders get 80% and non-contenders get 20% or first 3 contenders get 60-72% depending on how competitive the field is... But ultimately what works best for me is capping a race, getting a feel for how much chance each horse has and assigning a fair price. Remember a line is simply a numeric representation of your opinion. If it is inaccurate, it is because either your opinion is inaccurate or your ability to read your own opinion is inaccurate. In either case you will have a hard time making money no matter what strategy you use. I still feel you will do better when you take the time, effort and preparation to make a line not to mention the fact that it is MUCH easier to stay disciplined, controlled and within the confines of a bankroll.

It is interesting how some races i line the percentages add up to 98 to 104% and I just need subtle adjustments and other races may add up two 120 or even 130, thus I have overestimated everyones chance and I have a major overhaul to do on my line. Would hate to be betting the latter type races without a line.

Do I line every horse, ABSOLUTELY, If it is a race I have a handle on every horse. Some races have too many unknowns and making an accurate line is a near impossibility. In those situations I just denote my main contenders and look for something attractive. A lot of times lining the non contenders takes a minute or two, 20-1, 30-1, 50-1, 100-1,..it is usually a fairly easy guess and to be honest these horses do not effect your line on the contenders much anyhow(thus if I make a horse 30 who should be 50 or vice versa, not a big deal, but you want to try to be as accuate as possible). Thus if you have a half a dozen horse in a 12 horse field, rather than use some generic label such as 20 % for them why not just go through the field and 1 is 20 and 2 is 30 and 3 is 20 and 4 is 30 and 5 is 20 and 6 is 30. In this case the non contenders are approximately 24% and I will not be cheating by 4% when lining the contenders. More importantly, by lining EVERY horse I force myself to really look at every horse. Many times I have not liked a horse originally, but when looking further, I realize that he has a shot, and maybe should be 12 or 15-1. Also just because a horse is not a contender does not mean that he is not playable. When I make my line I label a hors playable or a lineout(meaning I will not play that horse even if he is value). The reason being I may not like a horse, but I do not want to UNDERESTIMATE his chance of winning. If I have a horse 15-1 and I do not line him out if I get 30 or 40 to one I take it(I do require double and no line out on horses I make 10-1 or higher a couple of ticks higher on horses below). For those of you who like to concentrate on 2 or 3 contenders, I ask, why on a menu of 80 horses on a typical card do you limit yourself to playing only 20 or 25. One further not if you play on the betting exchange(ehorsex) a line is a TREMENDOUS asset(almost mandatory) since you not only can lock in prices often much better than track odds you can bet against horses and it certainly helps to have your line in front of you when determining who to bet against.

Regarding the procedure, it takes practice especially if you are from the bet the most obvious horse school. It will likely be uncomfortable too. Like anything else new, take it slow and wait to you get comfortable with it.

Regarding if I make a horse 4-1 and he goes off at 12-1 he might only have a 15% chance of winning whereas is he goes off at 6-1 my 20% assesment is more likely accurate, I question what purpose that fact serves. I have not charted it so I cannot tell your one way or the other(I play horses recreationally) though if they say that is true I have no reason to argue, but so what? I am going to bet the horse either way, so does that mean I just bet less on the 12-1? than the 6-1, I would rather bet more(personal taste). Do I make adjustments to my line. YES. If I make a firster 15-1 and he looks live at 2-1, I bring him down to a 3-1 line out and adjust the line on the other horses, same with shippers and layoff horses or any horse that I feel has a right to have a better chance than I projected for him. Also there is a such thing as smart money(horses that are getting money that are not obvious to the public) and there is a such thing as a dead board horse so you have to sort of factor these things into your decisions, or in adjusting your line.

One final note, is regarding exactas. I do not believe that in thoroughbred racing that using your win odds line is an accurate model for betting exactas. It worked very well for me back when I used to play the Harness races. In t-breds it has failed me for too long and I have completely abandoned it. Now my focus is more on straight betting and place betting if the horse is not what I consider a win type(meaning that if he ran off the board I would hardly be surprised-some reasons might be bounce, suspicious drop, dangerous pace situation, off long layoff, 1 st time turf or 2 turns etc). I use other strategies for the exotics such as identifying races with just a few contenders or keying my value with the key contenders in all 3 slots fo the tri, tryng to anticipate pace casualties and using the tri as a means to capitalize if correct etc).

[SLAM DUNK(MW)]

By the way I just read further into the Steve Fierro thread, so now I get at what Game Theory was implying. He actually will eliminate a bet on a horse at certain points and adjust his line on other contenders based on his projected reduced probability on the value horse. The thought of betting a horse because he is 5-1 and passing on him because he is 7-1 is insane to me. Also what happens when I pass on him at 7-1 only to watch him drop ot 5-1 after the race goes off. Ouch. I am not going to insist that a 4-1 is going to win 20% of the time whether 6-1, 10-1 or 12-1, but in theory they should and to me it seems the extra value should more than make up for any loss in win %. If GT has a huge database proving I am wrong, good for him, but I have to do what feels right and playing at 5-1, passing at 7-1 is something I WILL NEVER DO.

[alysheba88]

I usually just make it on 3-4 horses. In the races I bet trifectas I make on 5-6 horses.

[formula_2002]

I make an odds line for every horse in the race, but am only truly interested in the top pick after adjusting the entire odds line for scratches.
My top pick, who's min odds line exceeds the final odds, perform
SIGNIFICANTLY better then those that do not exceed the min odds line.

Just a note:
It would be almost impossible for me to handicap a 10 card race day without my computer program. The odds line for every horse can be generated in less then 1 minute.

Joe M

[Valuist]

I say you can't make a realistic line on the main contenders if you don't know the expected odds of the longshots, some of whom may be no-chancers and some who could be sleepers. The easist way to make a line, IMO, is to line the garbage first. Should the longshots be 15-1 or 50-1??

[Lefty]

Mitchell also said Ron Ambrose would get a race down to 3 contenders and just make a line on ea. one at 3-1. Seemed to work for him. Guess many ways to skin this cat.;here kitty, kitty.

[alysheba88]

Valuist, I see your point. The one thing I would add is if the non-contenders are not much behind the contenders then i pass the race. Specifically if I have a hard time coming up with my 4 contenders and can make a case for three of the non-contenders, then I usually just conclude the race is too wide open for me and pass.

[Dave Schwartz]

Slam Dunk,

The resaon this works is that we are not really making a line. Instead, we are determining the probabilities for the horses.

If you know the probabilities and the current odds, you can determine the $net (or ROI).

Thus, instead of saying, "I am going to bet horse A if he is over 3/1" we say "I am going to bet the horse that has the highest $Net among my contenders."

The entire process is very dynamic. As the tote board changes, the probabilites change (somewhat). Generally the most profiable play will remain the most profitable play because, as the probability goes down, the odds are going up to compensate for it.

When I was at Saratoga, I had a 50/1 horse come up as my "best pick." After smoothing with the toteboard, his probability changed and he became the horse that was 3rd in terms of win probability. He was still the best play in the race, of couse. He just no longer looked like he should be 2/1.

Dave

[Fastracehorse]

About the odds of my selections.

After a while you know when the public will like or ignore your style of selections.

fffastt

[SLAM DUNK(MW)]

Dave, I understand what you and GT are saying and I am sure you base this off of your own data. My 4-1 and your 4-1 and GT 4-1 are three different things. I made mention in my post that you do have to read the board and make adjustments accordingly, bringing down your line on a "live" horse or possibly passing a stone cold dead horse(or cutting the bet way down). Of course you have to know what a "live" or "dead" horse is and that takes experience and is an ever evolving process. These are adjustments I make. I contend that whatever sample you have is TOO SMALL and just because GT may show 100 or whatever # 4 to 1's at 7-1 winning at only 12 %, the next sample of 100 may easily win at 18% or 20%. I would also argue that if you have a very large sample of 4-1's at 7-1 winning at 12% there is something very wrong in your capping. You brought up using the odds to determine the best roi, why limit yourself to the best. If I have a 4-1 at 8-1 and a 2-1 at 3-1 I will play them both. My biggest frustation over the years has been that too often that 2-1 is 2-1 and that 4-1 is 6-1 by the time the race finishes. Of course the betting exchange has put an end to that frustration for me. Moreover, even if I was to buy that your odds adjustment is accurate, what does it matter? You do not know the final odds until the far turn(or is it after the race now) and anything prior is just a guess.

Where it might have the most promise is in the exacta pools, only I feel that in too many situations in t-bred racing there is a major disparity between the chances of a horse winning and coming second and even when I used to make a place line as well, it did not really seem to help me out and never really got a feel for doing so..

Just to conclude you are taking the inexact toteboard(not sold that the Wednesday crowd and Saturday crowd will come up with the same line on the same exact race), using the current odds which is just a partial reflection of what the true final odds will be and this is what will make your incorrect line, correct?

Hey I am glad it works for you guys and seriously hope it makes you rich, but this is just not a theory that excites me in any way, shape or form.

[Dave Schwartz]

Slam,

Yes, that is about right.

Understand that I do not expect to make the "right" line; only a better one than I had before.



Dave

[cj]

What good does adjusting your line for the public's line do? I just don't get it. I understand the principle, but here is my problem with it.

You adjust a contender you have at 4-1 who is 20-1 on the board. No matter how you adjust your line, it will be between 4 and 20, and he will always be an overlay.

Now you have a horse on your line at 6-1, and he is 2-1 on the board. Same thing, except he will always be an underlay.

Granted, an adjusted line will be more accurate, but once the race is over, it doesn't really matter.

One other problem, the public odds are unknown to a great extent until after the bell goes off. You can guess, but you will probably be wrong.

[GameTheory]

If you adjust one horse, you're really adjusting all horses because you've got to renormalize. You don't really see how it all shakes out until you renormalize (get the probabilities so they sum to 100% again after adjusting them up & down). Some overlays will get bigger, some will disappear, some underlays will get bigger, some will disappear, and some underlays & overlays appear where before they seemed to be at a fair price. And of course a lot of the time, nothing much changes. But the effect is more complex than it would seem at first thought....

[cj]

I understand all the other horses would have to be adjusted, but in the real world, I just can't see pulling this off. There simply isn't enough time as the odds are constantly changing, sometimes big time, especially in the last minute. So all the "normalization" would still just be a guess when its time to place your bet.

Also, who do you adjust first in this normalization? If you do them all at the same time, the effect will be the same. Overlays will still be overlays, underlays will still be underlays. Do you start with the favorites and work your way up? Or reverse it?

I guess what I'm saying is it sounds great in theory, but how can this possibly help out in the real world where time is a factor?

[Dave Schwartz]

CJ,

When we were at Saratoga, it occurred every time I updated the tote odds.

The interesting thing is the relationships become very stable with 4-5 minutes to post. Oh, the odds might shift downward or upward, but the relationships don't change much. Remember that as the odds come down the prices go up to compensate.

At least this is how it works in our system.

I would not want to do this by hand... without a computer, though.


Dave

[Larry Hamilton]

On the other hand......If you do your horse work by computer, thru some kind of db that can be fed by Excel, Game theory has a program that will go on the internet, download all the scratches in real time. It then creates a data sheet in Excel that is the scratch list. IIIIIIIIIF you are working by computer, then you could weave the scratch list into your daily work and recalc everything on the fly.

In addition, it "could" monitor which horses are acceptable bets. So if you can catch him with nothing to do, and odds-monitoring is your angle, this is a program you really need.

[Dave Schwartz]

CJ,

Missed your previous question...

WHere it does the most good is that it gives you a clearer picture of the race.

For example, that $103 horse (again). I started with that horse as my top horse, looking like a world beater. But when adjusted I see that there are actually 2 horses that are more likely to win than him. I still play him but I now have a clearer picture of how he fits into the race.

Another thing that happens... I had a race where I had tossed out the 3/5 favorite. Since he was a non-contender, my system assigned him about a 5% probability. After adjusting for the tote, he was boosted to 30%, something much more reasonable. (Just take my word: 3/5 shots that you throw out just aren't THAT bad in the long run. That is why the beat you sometimes.)

And, finally, one more thing that I am working on now... Linking my playable contenders to some of the longshot non-contenders in exactas. Here is a race that I handicapped for our next video...

(Please note: This would be considered the ultimate redboard. No, I did not handicap/play this race on September 1st.)


PHA02 Monday, September 1, 2003
=====================================
Distance: 8.32 Furlongs
Age 3
(CL) Claiming: $7,500
Purse=$11,000 Field=11 (expected=11)
Level=98 Track Level=101 EP=71 FT=68
ES: (25) 8-6-6-5
RS: 1-4-4-0-2


PHA02 (o:PSR) Val=0) Bet Int=11
Prg Horse Cont Odds Prob $Net Opt% Fair Odds
10 Karakorumblackjack C 3/2 34.1 $1.71 -9.7 2
6 Typo C 5 14.6 $1.75 -2.5 6
5 Johnny Loves Jazz C 25 14.4 $7.51 11.0 6
7 Fabulous Dasher C 5 10.2 $1.22 -7.8 9

Cont Prob= 73.3 Odds=0.36:1 Book Pct= 77 Dutch= 0.30:1

Exactas (Min DV=1.80)
5- 6 5.48
5-10 5.34
5- 7 3.83


Payoffs
_______
5 Johnny Loves Jazz $54.80 $28.60 $13.20
11 Dancer Sonata $60.20 $16.20
4 First Dixie $4.20

$2 Exacta: $4,615.60
$2 Trifecta: $34,523.00




Here are my original contenders. The #5 is the eventual winner. The exactas are the 3 polayable combinations among the contenders. (These win probabilities have already been adjusted.)

Now, this race would not naturally be defined as a "crush race" in my approach because the "fair odds" for these contenders are 0.36:1 and the "Dutch odds" are only about 0.30:1. (It probably would have been a "crush play" at post time, if the top horse had been 9/5 or more, but I handicapped with final odds. That is why the final odds shifts often work in my favor.)

So, in my "standard approach" I would play the #5, win/place and bet the "profitable" exactas. (Then I would add in the non-contenders and look for other profitable exatcas to play but only the ones with my "playable horse," the #5. Basically, this race will scream for me to wheel and back-wheel the #5.)

But notice what happens when I also consider the non-contenders in the picture:




PHA02 (o:PSR) Val=0) Bet Int=11
Prg Horse Cont Odds Prob $Net Opt% Fair Odds
10 Karakorumblackjack C 3/2 34.1 $1.71 -9.7 2
6 Typo C 5 14.6 $1.75 -2.5 6
5 Johnny Loves Jazz C 25 14.4 $7.51 11.0 6
4 First Dixie 9/2 10.2 $1.13 -9.7 9
7 Fabulous Dasher C 5 10.2 $1.22 -7.8 9
3 Joyeux Occasion 10 3.4 $0.75 -6.3 30
2 Danny Boy O'Leary 12 2.9 $0.75 -5.2 35
8 Holy Man 19 2.5 $1.01 -2.6 40
1 Seaside Tony 25 2.5 $1.32 -1.4 40
9 Fireman Rahill 35 2.5 $1.82 -0.3 40
11 Dancer Sonata 80 2.5 $4.10 1.3 40


Exactas (Min DV=1.80)
5-11 12.85
5- 9 5.67
5- 6 5.48
5-10 5.34
1- 5 4.12
5- 7 3.83
4- 5 3.54
5- 8 3.15
9-11 3.10*
6-11 2.99
10-11 2.92
2- 5 2.33*
3- 5 2.33*
1-11 2.25*
7-11 2.09
4-11 1.93*




It does 2 interesting things here:

First, notice how the #4, a non-contender, slips into the top 4? This tells me that he deserves consideration even though I don't like him. Wouldn't want to leave him out of trifectas with my key horses.

But, second, notice the horse on the bottom? He is the only non-contender with a positive Opt%. He should be considered in the exactas with my profitable contenders.

Of course, looking at the updated exacta combination list shows me the extra exactas that SHOULD be profitable. (note: I never actually need to look at the exacta payoffs to determine this.)

The "*" exactas are those without a contender and will not be played. The others are all viable bets.

The point is not that one should run out and purchase my software... so, please, naysayers, resist the urge.

The point is that this is the effect of "smoothing" the odds.


Dave Schwartz

[Larry Hamilton]

I did it this way for a long time, until I realized there is a flaw in this over and underlaw stuff. To do this you need one bed rock correct system of assigning probability of winning to a horse, to this you compare the crowds odds (probability of winning). The only thing you can do to the crowds odds is bet more driving the odds on your bet down--not a smart tactic, senseless really.

The other variable is a measure of how well I handicap. Can I predict the acceptable odds of a horse with little error? Of course not, who can? Herein lies the rub. If you can't accurately calcualte one side of the equasion, what is the value of your comparison?

So, it would appear to me that the conversation about over, under and other relations between real odds and my odds is a discussion about how to divide. The real questions are: How do you create an odds list and how accurate is it. Hell, how do you even measure how accurate it is?

[Larry Hamilton]

Nothing makes my skin bunch up around my neck faster than hearing a guy say, "I make this horse 8 to 5." My thought is: Oh really? How? And How often do you get it right.

Oh shit, I should have read daves post before I put my shoe in my mouth

[Dave Schwartz]

Larry,

You are dead right. When I make a horse 40% (i.e. 3/2) that does not mean that he is REALLY 40%. How would I know?

What it does is give me a better picture. (My favorite phrase these days.)

It stops me from looking at the race above and saying, "Wow! That #5 can't be beat." It reminds me that his real chance of winning is around one-in-seven. It falls into the category of "could happen" instead of "likely to happen."


Dave

[SLAM DUNK(MW)]

Regarding some of the latter questions and points. Yes the objective of an oddsline is to make an accurate line. Yes making a truly accurate line is very difficult. Besides having to handicap flawlessly(which means you have to be able to, you have to make no mistakes and you have to convert your opinion in to that oddsline sans error), you have to be privy to the actual physical condition of the horses, for which the toteboard/form/appearance(personally I have a hard time differentiating from a 3 claimer and a stakes horse so there is liitle I can do on the appearance end) may or may not provide a clue. The fact that there are guys privy to this information influencing the odds, it makes sense that using the toteboard to iron out your own line will be effective. That being said there is a mass of people betting horses using a wide array of techniques. They also influence the odds to a much greater degree and personally I do not want them influencing (my odds line). I also still maintain that everyones odds line will be influenced by this ironing out process differently(not everyone will benefit or benefit to the same degree) and the sample sizes to actually determine whether this will work for you would have to be quite large.

My second point on the objective of an oddsline is more personal in nature. The objective of my oddsline is to PUT MY OPINION OF A RACE EXPRESSED AS AN ODDSLINE, so I personally can MAXIMIZE my performance. So while I aspire to make the "perfect" oddsline and we all know that I will not even with the adjusments I make, at least at the end of the day I am betting the horses that offerred value in my book. Thus if I make 3 horses 4-1 and two are 7-1 and the other is 9/2, I am not really interested in saying gee over my 77 4/7 I only won 12% of the time so they should actually be 7-1 shots and my 4-1 at 9/2 should now be 3-1 after all my other "ironed out adjustments" so I am going to bet the 9/2 shot. Personally it would drive me crazy after one of the 7-1 shots won(which as far as I am concerned is twice is likely as the 9/2 shot winning).

[Fastracehorse]

Sorry CJ, not trying to offend you by associating yourself with me, but I have no interest in comparing the probability of my horse winning; a hypothetical odds line; and, the tote.

What difference does it make??

Sometimes I'm going to get a bit less than I hoped for - sometimes a bit more. Does it matter?? Does it even out in the end?? I'm not sure ( well I am, it doesn't even out ).

My point being: I like reasonably priced contenders anyways and I like to play what I deem: High probability horses.

So to the hypothetical odds-line makers: Why does it matter?? Do you guys play short-priced horses?? Schwartz was talking about a bomb, but is that a real rarity in your system Dave?? I doubt it is.

Fastracehorse.

[Dave Schwartz]

Fast,

Well, one like this is certainly rare, but, no, there are lots of price horses.

I think the real issue here is the picture of the race that it presents. (My favorite word again.)

Think of it this way... Most people, myself included, want to quantify as much as we can with numbers. We want to measure things.

Now I can look at a race (as in the above example) and see that I have overlooked the 4 horse. I would not bet him to win, but if I were betting Trifectas, I would not want to leave him out.

And, without the exacta play list I would never consider linking that $54 horse to the eventual place horse.

To me, the "Picture" is a look at the race that brings it all togerther.

But it has to work FOR THE INDIVIDUAL. I just know that this approach appears to be taking me where I have wanted to go for 25 years and nothing else I have ever tried has done that.


Dave

[cj]

I guess for me it would all come down to this. The purpose of making a personal betting line is to compare it to the real odds, and make a betting decision based on the comparison. Why in the world would you want to include the very thing you are supposedly comparing on both sides of the equation? Who cares in the long run how accurate your line is, the question is does it guide you to good wagering decisions? I know when I assign a horse 2-1 odds, that probably doesn't really represent his "true" chances in the race. It doesn't matter. What matters is that the horses I bet who are "overlaid" at 2-1 show a profit, nothing else.

[Fastracehorse]

I like quantifying too, but 2 things:

1) I look at quantification like a dart board. That is, if several horses hit the dart board then I consider them contenders. I don't have a need to hit a bulls-eye.

That is not saying I don't identify prime contenders, I do. But they would arise when they hit the dart board and none others do.

2) I like speed figures - big time quantification. I believe strongly in my figures accuracy but I have learned that for me; relying on them too much, is a dangerous thing to do.

Basically my handicapping is like this: Identifying horses that will run close to their tops through trainer intent, with a dash of race-shape.

fffastt

[Dave Schwartz]

Fast,

As they say, different strokes for different folks.

Personally, without being able to quantify things with numbers, I am lost. But that is just my style.

I am always impressed by guys that can decide whena race should be bet with both hands just by non-numeric handicapping.


Dave

[Fastracehorse]

We all create a niche and then adapt to it the best we can.

And, I do admit that I'm a better bettor, sice I have been using some quantification methods.

fffastt

[VetScratch]

Dave,
The resaon this works is that we are not really making a line. Instead, we are determining the probabilities for the horses.LOL - for every probability there is an equivalent odds value and vice versa. Whether you choose to show both representations, probability and odds, is a matter of preference.

When AmTote finds that 33.3% of the money was bet on Chalky at 15% takeout, they elect to publish odds of 1.50/1 instead of the 33.3% probability. When you estimate that probability is 33.3%, you also arrive at the same odds value whether or not you elect to present it.

[Dave Schwartz]

VS,

No kidding.

Dave

[Tom]

In betting pools, the probability/odds number represents a population and is absolute - the 2-1 shot is a 2-1 shot in terms of total money bet on it - there are no other factors to conside,
When you make a betting line, your number is never absolute becaseu you never have all the available information at hand-you can only factor in what you have to work with-you don't know that the 1a stepped on a safety pin in the paddock, or you don't know that the jock's batteries are low and will not deliver the stretch punch needed to win.

[maxwell]

I toss out all horses that will probably go off at less than 5/1. Most of the time, it will be the morning line picks or horses that are selected three or four times by the DRF cappers.

I use a factor system that doubles the odds for each factor that is lacking.

9 = 5/1
8 = 10/1
7 = 20/1

One example of a 9 would be: three races in 45 days, today's rider has a previous good effort on the horse, a relevent Beyer, a good finish vs. today's class or better, and several races within a furlong of today's race. I would not consider any runner without contending speed.

I don't know how people can bet on horses without some sense of a value line.

[VetScratch]

Dave,

Terminology creates perceptions, and perception differences cause most disagreements on this board.

In another thread (Steve Fierro A Quote) we haggled over the merit of blending public and handicapped probabilities/odds. I couldn't understand why you and Game Theory seemed to put more emphasis on blending than fundamental handicapping.

Now, from the examples you posted here, I see that I misunderstood why and what you are blending. You previously said you were building a "new science of value." Instead, your example implies that you are grafting value concepts onto your old winner-selection system. That certainly explains why blending works for you. However, I thought "building a new science" meant you were scrapping winner-selection methods and adopting pure handicapping methods after Fierro's book influenced you to become a value player.

It seems to me that you are blending public probabilities/odds with winner-selection probabilities/odds instead of handicapped probabilities/odds. As a consequence of winner-selection methodology, you are starting out with a rather artificial set of probabilities/odds. Wouldn't blending work better if you started out with a properly handicapped set of probabilities/odds?

Perhaps you are suggesting that your initial probabilities/odds are less artificial than they appear. However, to me they have the same template-like characteristics that are usually found in morning lines and winner-selection lines.

None of this means that blending is bad. To the contrary, it explains why you emphasize the merit of blending. It also explains why we haggled. Before you start blending, you are simply further removed from reality than where I assumed you would be after building a "new science of value."

By definition, the goal of handicapping is to estimate probabilities/odds rather than pick winners in individual races. Success is measured by the accuracy of the probabilities/odds, not by the outcome of any single race. When many races are considered, all that matters is whether there is a high-degree of correlation between handicapped probabilities/odds and actual results.

If your 33% (3/2) horses are winning extremely close to 33% of the time (and your tests pass muster for statistical confidence), you have attained a small measure of success. If most other probability/odds plateaus are also winning extremely close to their expected percentages, you have attained a larger measure of success.

"Value handicapping" is a misnomer that has emerged because the word "handicapping" is seldom used properly. Many horseplayers use "handicapping" to mean selecting winners, or even worse, to mean selecting wagers (i.e., playing the races).

Most advertised "handicapping" methods and programs are designed to select a winner and/or explicitly provide guidance for playing a race. At the same time, one school of horseplayers that we call "value players" is interested in methods and tools that simply handicap the races. The term "value handicapping" emerged to appeal to "value players" and differentiate pure handicapping systems from selection-oriented systems. This is admittedly misleading because handicapping and value play are discrete processes.

When you think about it, a perfect handicapping system could never legitimately claim to "beat the races" because perfection implies that you accurately predict results, which dooms profits on account of takeout and breakage. This is why the vast majority of horseplayers want a winner selection system rather than a handicapping system.

Properly understood, handicapping is similar to formulating par speed ratings that merely predict how fast today's race "should" run. Thus, good handicapping is merely a tool like par figures. How you handicap will never beat the races. The only path to profit is how you play the races. "Value play" is simply a theoretical approach to playing the races profitably. While good handicapping is a tool for value players, they must ultimately rely on their personal judgment just like winner-selection players. Actual Pure Default Optional
Track Handicapped Selection Selection Morning
P# Odds Odds Odds Odds Line
1 4.3 2.7 5/2 5/2 7/2
11 105.5 31.2 99/1 30/1
4 4.2 4.6 24/1 5/1
2 4.0 4.8 5/1 5/1 5/1
9 43.3 15.7 49/1 12/1
10 103.8 99.0 99/1 30/1
12 21.5 24.0 99/1 12/1
7 99.5 82.3 99/1 30/1
6 9.0 9.2 10/1 10/1 5/1
8 67.0 31.1 99/1 20/1
5 109.5 13.9 99/1 20/1
3 1.6 3.9 9/5 9/5 3/1
(--1--) (--2--) (--3--)

1) Odds-line from so-called value handicapping software.
2) Default contender odds-line from winner-selection software.
3) Optional full odds-line from the same winner-selection software.Probably ninety percent (90%) of horseplayers would be happiest using winner-selection software because it shares their objectives and suits their style of play. Most players want to study races, identify contenders, then select a key horse to bet via win and/or exotic wagers. In this regard, good winner-selection software is a productivity tool and also decision-support tool that helps players choose among their contenders. Few horseplayers, probably less than ten percent (10%) of the total population, would be happy with pure handicapping software because many races will appear more difficult to play. Winner-selection usually makes a race appear more playable by accentuating the differences between horses. Pure handicapping software will often predict that there is little discernible difference between contenders. This is not what most players want because it suggests that too many races should be passed, and we have all observed that the vast majority of horseplayers hate to pass a race.

Because winner-selection software accentuates differences between contenders, I think it best supports the style of play preferred by most horseplayers. In my example, the winner-selection odds accentuated the differences between horses to help players easily make selections. This a natural consequence of the way winner-selection software tends to evaluate data.

If anyone is interested, we can discuss why winner-selection systems work this way, but many of the reasons have already been revealed in Dave's excellent thread about evaluating current form.

In my example, I used winner-selection software that tends to exaggerate the differences between horses. This is the true objective of winner-selection software, and the software I used is very good at picking winners. It is widely acclaimed because it picks a high percentage of winners, has a huge loyal following of users who report success, and has been reviewed by many respected authorities (who gave it high or even highest-ever ratings).

In comparison, a small minority of horse players might prefer a pure handicapping system. I say this because strict "value play" does not appeal to most players. Many talk about "value play" but few actually adopt it as a formal strategy. If you disagree, it is because all players intuitively "consider" value, but few have defined formal value tactics. Most players prefer to devise formal winner-selection tactics and intuitively consider value. I know many successful "natural born" horseplayers who can intuitively assess value better than so-called value players who assess value mechanically.

Right now, what Dave and Game Theory have achieved through blending probabilities/odds has me intrigued. The success of Benter's Hong Kong team has been well publicized. However, I submit that anyone considering this approach would be better off starting out with good handicapped probabilities/odds rather than probabilities/odds derived by winner-selection methods.

[Valuist]

There's another reason to make an odds line on all horses besides just making simple comparisons to the board. Recogizing steam is important, especially when its a barn that is very good when they bet. These horses are more than likely to be better value in the exotic pools. These horses are even more interesting when they are a first timer or a horse coming off a layoff and they look very sharp physically. Many of the best steam plays are secondary contenders; horses who should be 8-1 or 10-1 but are holding at 9-2/5-1 on the board. Since these aren't prohibitive favorites, many casual players don't realize these horses are alive. A good barn, early money and a sharp paddock/post parade appearance is a very solid angle.

[cj]

Noone has really ever answered my main question. I stated that, and its obvious, blending any single horse with its real odds won't change its status as an overlay or underlay. What will change is the relationship among the other horses.

For example, you have a horse as 4-1 on your personal line. The horse is 7-5 on the board. For arguments sake, lets say your formula changes your personal line to 5-2 on the horse. How do you decide which of your other contenders will go up in price? For that matter, which horse do you start with? Do you begin with your personal favorite, the actual favorite, the longest of you contenders, etc? All these decisions could be programmed, but what is the order and what is the logic behind it?

By the way, I think Fierro accomplishes the same thing with his race filters.

[Valuist]

If your 4-1 turns into a 5-2 w/a redone line, were talking about going from 20 pari-mutuel pts to 28. We would have to subtract 8 other pts from the other horses. A 2-1 shot would get approx. 2 pts deducted; a 5-1 would get approx 1 pt deducted. As long as it adds up to the 120-124 range, you'll be ok.

[cj]

I know how it would work, what I'm asking is what horse do you start with?

Example:
Your line Real Odds
2-1 5-1
9-2 7-2
5-1 6-5
6-1 10-1


Pretty common occurence. Where do you start? Do you adjust the overlays or underlays first? Once you decide that, which one of the category do you adjust first? Also, how would this change anything regarding who is overlaid and who is underlaid?

[Dave Schwartz]

VS,

You said: "It seems to me that you are blending public probabilities/odds with winner-selection probabilities/odds instead of handicapped probabilities/odds. As a consequence of winner-selection methodology, you are starting out with a rather artificial set of probabilities/odds. Wouldn't blending work better if you started out with a properly handicapped set of probabilities/odds?"

First, I do not know what you mean by "a properly handicapped set of probabilities/odds." What does that mean?

I did not mention what our slection process was as I did not feel it was germain to the discussion.


Later you said: "However, I submit that anyone considering this approach would be better off starting out with good handicapped probabilities/odds rather than probabilities/odds derived by winner-selection methods."

What does that mean? - "good handicapped probabilities/odds"


Dave

[Valuist]

CJ-

If you need to subtract points, I'd say look at the 3 main other contenders to split it up. Maybe 3 pts, 3 and 2. If there's only 2 other horses that realistically can win, then subtract 4 and 4.

In the example you cited above, why adjust at all? Let's say the 6-5 was a severe seconditis horse, you probably should just accept your line as being correct. OTOH, let's say the 6-5 shot was coming off a layoff or one really bad race, the public probably isn't the one pounding him. In that case, the barn probably has a lot better idea of how the horse would run. Then, I would scrap the line altogether. I'm a big believer in making odds on all the horses, but I do believe there's plenty of races in which you really can't.

[Dave Schwartz]

CJ,

I don't know that you were looking for a reply from me, but I'll try.

>>How do you decide which of your other contenders will go up in price?

For that matter, which horse do you start with? Do you begin with your personal favorite, the actual favorite, the longest of you contenders, etc? All these decisions could be programmed, but what is the order and what is the logic behind it? <<

I adjust the entire field. Some go up, some down.

>>By the way, I think Fierro accomplishes the same thing with his race filters.<<

I entirely disagree. In my opinion, Steve's "race filters" was about deciding what races (for him) were not playable/profitable. That section of the book is enough to turn the almost-break-even player into a winner.

For me, 4-Quarters became a defining moment in my horse racing development. Not so much because it was a great book (sorry, Steve) but because FOR ME it caused a light to go on. I really think that "turning on the right light" is what it takes to move into consistent profits. For some people it might be Steve Crist's book that adds what was missing. For others, Steve Davidowitz. For me, it was Fierro's. (These Steve's are everywhere, aren't they?)


Dave

[cj]

Dave,

I'm looking for an answer from anyone. I'm not trying to pick on anyone or cause a ruckus, I just don't understand and I would like to see it more clearly.

I understand you adjust the entire field, but unless I'm missing something, there must be a pecking order, especially if you are using a computer. In almost any race, you will have some overlays and some underlays when contrasting a race with a personal odds line.

I say you cannot adjust the entire field at one time, or there would be no benefit. As I've stated before, underlays would remain underlays and overlays would remain overlays, just to a different degree than before adjustments. You would have to do one horse, then reallocate the change in percentages. It may appear to be all done at the same time by a computer program, but it couldn't be coded that way. I do a lot of programming and just can't even fathom a way to "correct" odds lines without some sort of order to the process. Maybe, if someone would give me an example, one race, of how you could do this, I'd understand.

As far as the race filters, I was referring to things like when a contender is 3-2 or less and all others are 5-1 or more, he passes. It has nothing to do with race type. It has to do with him greatly underestimating the chances of a horse and deferring to the public, the same as saying I made him too high, so the others are adjusted upward to not be overlays. Same when he says 3 overlay pass or 4 overlay pass, he is using the publics odds to realize he missed something in the race.

[Foolish Pleasure]

Making a line for all horses is an interesting proposition until you try to differentiate between 12/1 and 20/1 shots,

best of luck correctly assessing these miniscule differences accurately and without the accuracy, what exactly would be the point?

[Dave Schwartz]

CJ,

When I referred to Fierro filtering on "race type" I meant "a race with 3/2 or below" was a type of race. I did not mean claiming sprints versus maidens.

WHen I adjust each horse, I do all of them at once. Program number order... Adjust the #1, then the #2, etc. When all have been adjusted, then the probabilities are re-normalized to 100%.

Recall at the start of this that I said I used "probabilities" rather than "odds lines?" (While the two terms to me are synonomous, apparently they are not to other people.) Using a probability is much easier to adjust.

While our actual method is a bit proprietery, the basic concept is this:

If a horse has a 30% probability and only 10% of the pool then the 30% needs to come down "some." If the horse has a 10% chance of winning and 30% of the pool then the 10% needs to go up "some."

So, some horses go up and others come down. After all is said and done, THEN we re-normalize to 100%.

The major changes are for non-contenders that are well-bet. Imagine that I have a 10-entry field with 4 contenders which I have assigned an 80% total probability to. Within the other 20% are the favorite and 2nd choice, at 2/1 and 5/2, respectively. These two horses comprise over 50% of the pool and I have given them perhaps a combined total of 12% or so.

I have probably underestimated these horses. How much? Who knows? (And how would one find out?) All I know is that these horses need to be increased SOME.

Now, I am not saying that these guys will become bets. They will NEVER be considered profitable, but it will give me a BETTER assessment of the TRUE possibilities in the race.

Please forgive me if I seem to be trying to win a point here... I am not. Frankly, it does not matter to me whether anyone considers this or not. I can only say that this has lifted me from slightly better than a break-even player to... well, almost an unbelievable ROI. It has made THAT much of a difference.


Dave Schwartz

PS: As I look back on this thread, I realize that it appears I am really trying to soap box here... that is not my intention. So, to that end I think I'd like to end my part in this discussion, unless someone really feels they'd like to hear more from me.

[GameTheory]

CJ --

Your intuition that all the overlays would remain overlays & all the underlays would remain underlays is a natural one, but it isn't true. Much of the time that will be the case, and not much changes, but because of renormalization and the degree by which each horse is adjusted is different, the gaps between the adjusted line & the public can end up increasing as well as decreasing (although it does decrease mostly, but maybe one horse will go the other way). Order doesn't matter. You start out with probabilities adding up to 100%. You adjust each horse up or down or not at all. Now you have a new sum above or below 100%, and you simply divide each horse's adjusted number by that new sum to get its new proportion/probability.

Note that this isn't actually the way I do it, it just the easiest way to explain the concept. I just throw the odds in as a variable along with a bunch of other normal handicapping variables, and my probabilities are re-calculated as the tote board changes. So in my case, I don't have any line at all until I plug in some odds.


VS --

Like Dave, I don't know the difference between "handicapped probabilities" and probabilities arrived at by "winner-selection methods". I have probabilities that estimate the chance of each horse winning. What other kind are there? (My guess is that you are getting at something to do with looking at races individually versus looking at a series over time, but I can't think what the actual practical difference would be between the probabilities.) Please explain.


Also, you said, "Success is measured by the accuracy of the probabilities/odds, not by the outcome of any single race. When many races are considered, all that matters is whether there is a high-degree of correlation between handicapped probabilities/odds and actual results."

While that is part of the measure of success, it is not all of it. Like I have said to formula_2002 (that's who Hurrikane is talking about when he mentions "Joe"), you can get the highest possible correspondence (100%) by simply diving 1.0 by the size of the field in every race. If you do that, EXACTLY 10% of your "10% horses" will win. But you wouldn't claim that those oddslines were any good, would you? (Each horse given an equal probability in every race.) The real measure of success is how high the probability you gave to the eventual winner, while still generating lines that add up to 100%. In other words, the ideal line for a race is where you give 1.0 to the eventual winner, and 0.0 to everyone else. Please refer to this old post:

http://www.paceadvantage.com/forum/showthread.php?s=&postid=36443#post36443

Read that whole thread to see a related debate on what we've been talking about...

[SLAM DUNK(MW)]

Dave, do not think what we do is all that different(I am however more concerned about "smart" money than public money). I think the two main differences is that you use your "adjusted oddsline" to isolate fair value other contenders to use in the exotics(a strategy that makes sense-though I prefer using the exotics in a different matter-keying my value horse only when there are a very limited number of contenders). The second is that you use your adjusted lines to maybe find value on a horse who would not have been value otherwise. The fact that you have had a tremendous increase in roi is obviously a good sign that it works for you. I would actually be really interested in whether you have looked at using the opening odds or (closing odds relative to morning line) as a possible adjustment tool. In other words you make a horse 20-1 and he is morning line 8 opens at 18 before coming down to 13-1(I do not think you need to make any adjustment). However a horse you make 20 is morning line 15 opens 6 before closing at 10 probably should be used to to adjust your line on other contenders upward. Likewise if your horse is 3-1 on your line, morning line 8 opens at 4 and closes at 6 that look a lot liver to me(probably no need for upward adjustment) than if he was morning line 3 opening at 9 and closing at 6(which looks on the dead side and probaby should be adjusted upwardly). Also I am a little bit curious in the difference between say your ironed out lines using the odds at say 5 minutes to post and the ironed out lines using the unavailable closing odds, just to get a difference in roi.

FP-I totally disagree that the distinction between a 12-1 and 20-1 is difficult. I find it no more difficult then the distinction between a 2-1 and 4-1. It is all fairly simple. Analyze everything and decide whether the horse seems like he should be 2-1 or 4-1 or 10-1 or 30-1, whatever feels right. Go through the entire field, add up the %'s and see what adjustments need to be made(upward or downward). Once you get to 100 it is simply looking at each horse and verifying if the odds you assessed seems right. If so great if not adjust. When you have 100 and everything feels right, next race.

Regarding another point. I do not use the analogy if this race was run 100 times how many times would this horse win. That does not apply in many instances(layoff horse, suspicious droppers.....) I mean if a horse wins for 40, 50, 50 earning 98 Beyers each time and now drops for 20, what good is this frame of thought. Well is he in one piece he will win 95 out of 100 times. The important question is WHAT IS THE CHANCE HE WILL WIN TODAY given all the information you have. That is the question you need to answer. I think after a while it is more feel than anything.

[hdcper]

Hi Dave,

Couldn't resist trying to figure out how you calculate some of your output figures since you didn't go into detail.

Your output:


PHA02 (o:PSR) Val=0) Bet Int=11
Prg Horse Cont Odds Prob $Net Opt% Fair Odds
10 Karakorumblackjack C 3/2 34.1 $1.71 -9.7 2
6 Typo C 5 14.6 $1.75 -2.5 6
5 Johnny Loves Jazz C 25 14.4 $7.51 11.0 6
7 Fabulous Dasher C 5 10.2 $1.22 -7.8 9

Cont Prob= 73.3 Odds=0.36:1 Book Pct= 77 Dutch= 0.30:1

Exactas (Min DV=1.80)
5- 6 5.48
5-10 5.34
5- 7 3.83

For horse #10 Karakorumblackjack

$Net = Odds(expressed as payoff) times Prob.

$Net = 5 times .341 = $1.705(rounded to 1.71)

Opt% = .5 times (($Net - 2.00) / Odds)

Opt% = .5 times ((1.71 - 2.00) / 1.50)

OPt% = .5 times (-.29 / 1.50) = .5(-.193) = -.096 = -9.7

Exacta number = (($Net) times ($Net)) / 2.40

For example: 5-6 combo

((7.51)(1.75)) / 2.40 = (13.1425) / 2.40 = 5.476 = 5.48


Again these calculations are only as good as the success of the handicappper's horse probability figures.

Thanks for sharing,

Hdcper

[Dave Schwartz]

Hdcper,

Yes, you've got it. Sorry I didn't explain them. I sometimes forget that not everyone uses the same abbreviations.

Permit me to help with the rest:

First, the "fair odds" for the contenders:

If you have 73.3% then the n/c= 26.7. That make fair odds 0.36:1.

The Dutch: This is how we measure "profitability" for the contenders as a group.

We add up the "booking pct" for each contender, in this case totalling 77%. Since there is a total of 100% to return, then we would win 23%. 23% / 77 % =0.30:1.

Since the 0.30 is less than 0.36 this race is not playable as a box between the contenders.



The Exacta stuff you have almost right but 2.40 is not part of it, at least not directly. Let's say that the average $net (per $2) is $1.60. If a horse has a projected $net of $1.90 he has a "DV" (i.e. "Dollar Value") of 1.19. If a second horse has a $net of $2.20 then his DV is 1.38. If they are involved in an exacta, then their combined DV is 1.19 x 1.38 = 1.64.

I have set the minimum DV at 1.80 for an exacta wager.


Hope this helps.

Dave

PS: The exact "base $net" is around $1.55.

[MV McKee]

This is one of the most interesting and informative threads I have come across in some time, very interesting reading.
I’ve been doing what most of you do (making my own line) for almost 3 years now. I started doing this after a brief hiatus (and retooling) I took from the game. Like others have stated in this thread, truly understanding the concept of value, and embracing it, has been a breakthrough for me. It has also taught me that perhaps the most important part of my game is “handicapping my handicapping”.

My process for making a line is relatively simplistic, especially when compared to the level of refinement some people have taken the art of line making to. If anyone has any interest, I thought I would share my methodology, and the reason(s) I utilize it.

Step 1) I identify non-contenders. These are horses that I feel have no chance of winning whatsoever. I am very liberal in my inclusion of horses into the “contender” category. I have handicapped 23,680 races in the time period since I began tracking myself in the manner I do now. From that 23,000 + races I gleaned the following information:

A) ~32% of the starters were categorized as non-contenders. This does not mean that I eliminated 3 horses in every 9-horse race or 4 in every 12-horse field. There were a number of races where I had every horse in a group of 12 as a contender, and others where I had as few as 2.

A) ~5% of the above group won.

Based on this, I assign the group of horses identified as non-contenders a 5% chance of winning.

Step 2) I assign all the other starters (including the individual starters in an entry) points based on a 40 point scale. In other words, if this race were run 40 times, how many times this horse would win. I initially started with a 100-point scale, but found that I spent far too much time making ridiculous fine-line distinctions. After I was satisfied that my 5% win-rate on my non-contenders was valid, I briefly adopted a 20 point scale, but encountered (obviously) the inverse of the problem I had with the 100 point boondoggle.
The 40-point scale has been a good fit for me, and I have used it for upwards of 2 years now.

Step 3) Every 3 months I do a diagnostic on my line and the public (tote) line. Basically, I group all the horses I handicap by the points I gave them. Every horse I assigned 1 point (a 1 in 40 chance of winning) is grouped together, every 2 point horse (20-1) etc. up to @17 points, at which point I lump everything upwards from there together into 1 grouping. I then adjust the odds at varying percentages by the public (tote) odds. Each increment from 2.5% Public + 97.5% Mine to 2.5% Mine + 97.5% Public. I then re-group by the new, hybrid odds calculation and compare the expected win percentage of each group to the actual. I then graph the average of the absolute value of the differences between the expected/actual. I then find the real peak of the graph (sometimes an individual calculation set may differ from the real shape of the trend). The low spot is my line, generally improving as the public odds are factored in to a greater degree, and then sloping downwards again. Once I find the optimal ratio, I use that as the determinant of real odds to base my betting on. Currently, I am using @ 70% public, 30% my odds line.
While I realize that this is not necessarily accurate for each race, it is (as it relates to my own handicapping) the optimal percentage to use in a blanket calculation. Obviously, I do incorporate some subjective judgments into individual races.

That is about it. I do basically the exact same things that have been discussed in prior postings in this thread. The basics are 1) Calculate an odds-line. 2) Adjust it accordingly by the public’s odds.

There are a few things I do a bit differently than some, that I feel give me an edge however.

In regards to Step 3 above:

1) I look at all non-contenders (from Step 1) as a single betting interest, and apply the total percentage of the public odds to the group as a whole. In this case, I have found that a 22.5% Public 77.5% my line is the optimal ration to apply to the group.
2) For 2 years I based the public percentages strictly on the final win odds in the results charts. For the past year + (since I have been retrieving tote data from the internet), I have been using a different formula to calculate public odds. This is based on the win AND exacta pools at the first tote cycle after the 1 MTP mark. I do this for two reasons A) This is the point in time where I am making my wagering decisions and I need an accurate assessment of how accurate (as a whole) the odds are at that point in the wagering. B) Just as the hybrid personal/public line is more accurate, so too is a combined win/exacta public line.


For Exacta wagering:

I also make a line for the 2nd place finisher. Not a PLACE line, but a 2nd place finisher line.
I cannot fathom that anyone could make an accurate 2nd place expectancy model based on a win odds line, no matter how accurate. My 2nd place lines often bear little if any resemblance to my win line. Non-Contenders in my win line are routinely not part of that category for my 2nd place line, and Win Contenders are on occasion, 2nd place non-contenders. Numerous scenarios come into play here, I can detail a couple of the obvious ones:

1) One dimensional speed horse-- Lead=Win, Hooked=Cooked
2) Long layoff class horse – In One Piece = Wins For Fun, Damaged Goods = Kal-Kan

I am overstating the obvious in saying that a place line, if done properly, should be a completely separate calculation and impossible to derive from a win odds line. IMO the single greatest detectable (in the absence of trifecta tote data) deficiency in the mutual pools is that 2nd place money distribution in the exacta pools is very (to an extreme) closely correlated to the win distribution in both the exacta and win pools.

I also recalculate the 2nd placing odds using a percentage of the public 2nd place exacta odds (derived for the same procedure I detailed in Step 3 above). However, I utilize 65% my line 35% public as the optimal weighting in this calculation.

Well, that’s my $.02. If anyone is still awake, I’d be interested in hearing any feedback. This is a great thread. Really addresses what this game is ultimately all about.

[MV McKee]

I forgot to mention something else that I do in relation to exacta wagering. I did it originally as an experiment, because I felt it held promise, but it has turned out to be perhaps my biggest edge in the exacta pool.
I do what I call "ticking" (no idea why I call it that).
I make 2 adjustments to my exacta odds line (based on my win and 2nd place lines).
The first adjustment is automatic. It accommodates for a factor that is more prevalent at some tracks than others that has to do with exacta combinations.
Exacta combinations in which the place horse exits from a gate within X stalls of the winner decrease in actual/expected frequency as X increases. In other words, if horse 3 wins, the 2 or 4 will complete the exacta more frequently than the expectation. The 1 and 5 slightly less so, etc., etc. I did not originally want to factor this in to my expectation calculations, but doing so has proven quite valuable.
The second adjustment is manual. I notate horses whose fortunes are tied to each other. This sounds very strange, but again, it has proven (for me) to be very valuable. It is relatively simple.
Each horse in a race has a scenario(s) which will have either a positive or negative effect on his performance in a race. There are a million scenarios, but I only look at pace and condition (when in question). We all factor these into our handicapping each individual horse, but I correlate them to others in the field a particular scenario (in essence an unanswered question) may affect.

I code them this way: 3PNN1
3 is the 1st place horse
P is a Pace Scenario (C would be a Condition Scenario)
NN is Negative-Negative, (PP Positive-Positive, NP, PN)
1 is Strong, 2 is Weak

A simple, but representative example of what I do:

In handicapping a race I determine that 1 and 2 are both speed
horses who appear to be need the lead types. The 2 horse is a steep class dropper coming off a win, so he has some CONDITION questions.

For the 1 I would give him 2PPP1 and 2CNP2
For the 2 I would give him 1PPP1

2PPP1 - In a scenario where the 2 wins, it was likely due to a non-speed duel (P) scenario which also benefits the 1 (P). The likelihood is weak since both are need the leads.

2CNP2 - In a scenario where the 2 runs poorly (the N) due to a valid condition concern (the C) the 1 benefits (the P). Lets say this is a strong possibility.

1PPP1 -- In a scenario where the 2 wins, it is due to the same scenario as the 2PPP1 above.

After doing this for a little over a year, I again have seen a fair amount of benefit. I adjust the individual exacta expectancies based on the codes (the others adjust automatically in sympathy).
I apply the pace scenarios only to on the pace types, and the condition scenarios only to obvious form cloud horses if they are on the pace types.

OK, am I certifiably out there, or does this make sense to others?

[GameTheory]

Originally posted by MV McKee

OK, am I certifiably out there, or does this make sense to others?

Not at all, I have had similar ideas. Excellent stuff...

[formula_2002]

THE COMPLETE HANDICAPPER

If you were to develop an odds lines based upon, track, surface, distance, race type and track condition, the combinations could look something like this for each track;

13 RACE TYPES, 10 TRACK CONDITIONS, 4 TRACK SURFACES, 4 DISTANCES (SHORT, SPRINT, MIDDLE, LONG) =1040 COMBINATIONS FOR EACH TRACK.
SAY YOU STUDY THE rounded odds ranges from 0 to 19
(20 ranges), then that’s 20,800 combinations per track.

Let’s say we look at 10 tracks, that’s 208,000 combinations.

Finding 1000 races for each combination =208,000,000 races.
(to find those 1000 races for each combination, you would have to REVIEW many more)

I think that would almost do it!!!

Joe M

PS, say you are trying to determine which of 70 factors would make a SIGNIFICANT DIFFERENCE BETWEEN your odds line and the public's, then you may have to look at an additional $%&#@
races..

[VetScratch]

Game Theory,
Like I have said to formula_2002 (that's who Hurrikane is talking about when he mentions "Joe"), you can get the highest possible correspondence (100%) by simply diving 1.0 by the size of the field in every race. If you do that, EXACTLY 10% of your "10% horses" will win. But you wouldn't claim that those oddslines were any good, would you? (Each horse given an equal probability in every race.)Let me propose that these debates would make sense to the broadest possible audience if we all tried harder to explain "why" our assertions/opinions are true.

What you said is true: "you can get the highest possible correspondence (100%) by simply dividing 1.0 by the size of the field in every race."

Now, for the sake of readers at all levels of knowledge, wouldn't it have been fair to add that this is like stating 2 + 2 = 4?

The reason your assertion is true is that program numbers tend to be random. This is what makes your clever assertion generally true (but still susceptible to small bias on account of coupled entries). You can randomly bet program numbers with the results that your assertion predicts only because program numbers tend to be randomly assigned to the horses in every race.

[VetScratch]

Game,
And you needn't point out that my "practical" example overlooked post position biases, which do have a correlation to program numbers. I should have said "randomly select plays" instead of "randomly play programs numbers." Thus, I was the first poster guilty of failing the test of helpful exposition, which was the very point I was trying to make! :)

[JustMissed]

MV, You are very literate and have a way of explaining concepts that is easy to understand.

You said earlier "Like others have stated in this thread, truly understanding the concept of value, and embracing it, has been a breakthrough for me".

Would you please explain:

A. What is value?

B. How is value used to select bets?

C. How did your embracing value cause a breakthrough for you?

Thanks,

[GameTheory]

VS --

Are you just trying to be difficult? Do you actually read these posts?

It is a bit like saying 2 + 2 = 4, but I find some people around here (like Joe) often will not admit that simple facts like 2 + 2 = 4 are true, while others (like you) will attempt to obfuscate simple assertions with all sorts of crazy nonsense.

It has nothing to do with program numbers or randomness. I have no idea where that came from. It is not generally true, it is precisely true in every case. You have clearly misunderstood what I've said, as usual.

If you make your probabilities like this:

every horse in race gets => 1.0 / field size

when we go to check the accuracy of our oddslines based on probabilities assigned in such a way (1.0 / field size), what do we find?

Well, let's look at the .10 horses. EXACTLY 10% of them won. Amazing. That correspondence is even better than the public! I guess 1.0 / field size is the best way to make probabilities.

Here is a 1-race sample with 5 runners just to show you:

HorseA .2
HorseB .2
HorseC .2
HorseD .2
HorseE .2

Now, even though we've only got one race, let's go ahead and look at observed/expected anyway:

Since we only have "20% horses", I guess we'll start with those. We have 5 horses to which we've assigned the probability .2. .2 x 5 = 1. That means we expect to find 1 winner out of this group. How many winners did we actually have? 1! Wow, we sure did a good job of assigning probabilities! And even though there is only one race in this sample, I am VERY CONFIDENT the correspondence will be just as high even over a large sample as long as we stick to the 1.0/field size method.


The point being, judging the utility of a set of oddslines by looking at groups of (assigned) odds ranges and checking observed/expected correspondence doesn't work. That's not the way to do it. It can be helpful, but using the above method of making probabilities (1.0/field size) -- which I hope we can all agree will make totally useless odds lines -- has a higher correspondence (in fact, a PERFECT correspondence) than can be achieved any other way. So *CLEARLY* that can't be the right way to judge oddslines or we would have to conclude that my 1.0/field size based lines are the best possible.

Please re-read this post repeatedly until it is crystal clear. You are not allowed to respond with anything other than 100% agreement, because it really is as simple as 2 + 2 = 4.

(Hint: if you still think program numbers has ANYTHING to do with it, you haven't got it.)

[GameTheory]

Originally posted by VetScratch
Game,
And you needn't point out that my "practical" example overlooked post position biases, which do have a correlation to program numbers. I should have said "randomly select plays" instead of "randomly play programs numbers." Thus, I was the first poster guilty of failing the test of helpful exposition, which was the very point I was trying to make! :)

I will not point that out, I assure you, because that would imply that this talk about program numbers & post positions was somehow relevant...

[GameTheory]

VS --

I am also still hoping you'll explain what you were talking about with that "handicapped odds" vs. odds assigned by "winner-selection" methods. What's that about?

[Dave Schwartz]

MV,

Great post.

What I took from it is very similar to my own approach in that your line is mostly "your factors talking" with "a little public" mixed in.

And a question for you... Did you find that using this approach improved your play? That is, could you see a measurable difference?


Regards,
Dave Schwartz

[VetScratch]

Dave:
First, I do not know what you mean by "a properly handicapped set of probabilities/odds." What does that mean?
I did not mention what our selection process was as I did not feel it was germane to the discussion.
Game Theory:
Like Dave, I don't know the difference between "handicapped probabilities" and probabilities arrived at by "winner-selection methods".
I have observed that the differences between a "selection process" and a "handicapping process" is usually rooted at the most elementary level where data factors are evaluated. I addressed this once before in another PA thread.

PA Thread: The Last Piece--FORM http://www.paceadvantage.com/forum/showthread.php?s=&threadid=7313

In that thread I posted the following opinion about Dave's contribution: "Good work, Dave... most "informative" series of posts since I've been a member! Other handicappers may agree or disagree with you on specific issues, but no one can say you failed to address FORM in a manner that flatters this board when any visitor opens this thread."

Towards the end, I also asked for further discussion about data evaluation at the most atomic level.

What I was getting at was that most "selection methods" evaluate data in a way that helps select winners by accentuating the degree of difference between members of the data sets being evaluated. Dozens or hundreds of comparative data factors may be examined, sophisticated methods of matrix evaluation may be used, and thoroughly researched coefficients of significance may be equated to various data factors (e.g., using weights or Impact Values), but most "selection methods" begin to distort pure non-linear relativity at elementary levels of evaluation. Here is an example.
Data Ordinal Linear Relative
Factor Ranking Scoring Indexing
94 1 100 100
92 2 87 96
87 3 75 85
86 4 62 83
76 5 50 61
72 6 37 52
68 7 25 43
50 8 12 04
48 9 0 0

Scoring Formula: 112.5 - 12.5 x rank
Most comparative data sets (such as measurements for speed, pace, recency, earnings box ratings, etc.) plot as a non-linear function such as the data values in the example. As Dave posted, selection methods often translate raw data into a different representation by ranking and scoring (as shown by Dave's examples in the thread on FORM). The scoring formula shows how this is often done. Note that this transforms the raw data set into scores that plot as a linear function, which does not preserve the non-linear relationship between the raw data values.

By comparison, a pure handicapping method would strive to retain the non-linear relationship between data values. This might be accomplished without altering values if the spectrum of raw data has known bounds (upper and lower), or more commonly by translating raw data into a relative index that is easy to work with, like scores, but preserves the non-linear relationships between data values.

This example of low-level differences in the way data is evaluated can eventually make it easier to select the most probable winner, but unless probability determination is totally divorced from such ranking and scoring methods, you inevitably end up with probabilities (and odds lines) that tend to reflect linear data evaluation techniques: 34.1%, 14.6%, 14.4%, 10.2%, 10.2%, 3.4%, 2.9%, 2.5%, 2.5%, 2.5%, 2.5% (as shown by Dave earlier in this thread). Few "real" sets of probabilities will ever have duplicated values.

None of this detracts from the efficacy of blending public and projected probabilities. To the contrary, it would seem to support the merit of blending. Dave and Game do great work; it is just my suggestion that what works good may work better if you started out with probabilities derived less synthetically.

[formula_2002]

I think the following represents just how I feel about creating and odds line...the more data the better, but there is a limit.

http://www.apmaths.uwo.ca/~bfraser/version1/vanderpol.html

change the epsilon setting to .09

Joe M

[VetScratch]

Formula2002,
change the epsilon setting to .09At .09, I started to get dizzy after watching it after a while. Changing epsilon to -.09 is easier on the eyes. :)

[Fastracehorse]

I believe that a handicapper cannot precisely assign weightings of factors or prescribe exact values to data in horse racing.

I say the above; and I formulate my own speed figures.

Horse racing isn't like that for me.

Horse racing isn't a physics lab where the environment electrons flow is controlled.

This is horse racing: To weight factors numerically seems impossible let alone unimportant.

For me, a type of abstraction of factors to know what they can lead to is important. Ex., if this factor works well I'll consider the horse.

Scientists know that it is human nature to try and 'pigeon-hole' data. That is, we try and measure something numerically ( retrieve data ) for something that is immeasureable - so scientists try not to do it.

Again, horse racing isn't a physics lab. Sure some data arrive at a postive conclusion ( winning horse ). But, I'm sure an abstraction process could arrive at the same conclusion without the reliance on data.

For ex., Belmont dirt races last Sun. - how much data does one need to bet speed when so many races went w-to-w. Statistically, speed may be a good bet, but there are more times when it is not - it is just stats say it is a good bet in general.

An all encompassing process would have to be abstract - less work, more intellectual IMO, and, more effective.

fffastt

[bettheoverlay]

I tried applying some of the concepts discussed on this thread over the weekend. I was continually frustrated by the dramatic last minute odds changes. Over half of my value bets actually had final odds less than my odds line, most considerably so. One of my winners was 7-2 going in the gate and ended up winning at 8-5. With so many people pounding perceived value in the final minute, the practicality of such an approach seems questionable. How often to you actually get value? Especially on lower odds horses.

I was also frustrated by my buddy who has never read a handicapping book in his life, doesn't own a computer, and is always amused by my mathematical tinkering. He looks for favorite angles, spends little time on each race, trusts his instincts and designs very small exotic bets. He hit 2 PK 4s, a trifecta and a significant exacta Saturday! He believes horse racing is far more about betting than it is about handicapping.

Sometimes I think all this figuring is just a roundabout way to get to somewhere you might have been anyways.

[VetScratch]

Fastracehorse@DRF,
Scientists know that it is human nature to try and 'pigeon-hole' data. That is, we try and measure something numerically ( retrieve data ) for something that is immeasureable - so scientists try not to do it.You are fearless! Most folks feel that such opinions should be discreetly whispered. Many handicappers "tremble" at the thought of antagonizing the ghosts of Sengbush and other scientists who died trying to beat the horses. :)

[hdcper]

MV and Dave,

I sort of read it differently, that MV used 70% public and 30% his line to derive his value or odds line.

Maybe I am just reading it wrong or miss understand what you are saying Dave.

Just trying to clear up the mud,

Hdcper

P.S. MV, I really enjoying your post, great ideas!!!

[Dave Schwartz]

Hdcper,

You are right! I misread. Thanks for correcting me.

Well, I am doing more the other way around.

Whatever works. <G>


Dave

[GameTheory]

Originally posted by VetScratch

By comparison, a pure handicapping method would strive to retain the non-linear relationship between data values. This might be accomplished without altering values if the spectrum of raw data has known bounds (upper and lower), or more commonly by translating raw data into a relative index that is easy to work with, like scores, but preserves the non-linear relationships between data values.

This example of low-level differences in the way data is evaluated can eventually make it easier to select the most probable winner, but unless probability determination is totally divorced from such ranking and scoring methods, you inevitably end up with probabilities (and odds lines) that tend to reflect linear data evaluation techniques: 34.1%, 14.6%, 14.4%, 10.2%, 10.2%, 3.4%, 2.9%, 2.5%, 2.5%, 2.5%, 2.5% (as shown by Dave earlier in this thread). Few "real" sets of probabilities will ever have duplicated values.


Hey, I think you said something that actually made sense!

But the relationship between maintaining non-linear relationships and making good probabilities is only indirect, and as you say doesn't really have anything to do with the blending process (so why'd you keep bringing it up?). After all, your final synthesis of your numerous (linear) handicapping factors is liable to be non-linear. It just comes down to what works best and what is practical. I make may lots of data transformations or none at all depending on the factor I'm working with. Much of the time discretizing data will improve your probabilities rather then degrade them, and it is often neccessary to make certain problems computationally tractable.

But since the above paragraph has probably the smallest target audience of any I've ever written, I will leave it at that, since the best way to deal with such questions is heavily dependent on the method of data analysis you are using and the actual data itself...

[VetScratch]

One morning on the backside at CD, an owner who is prominent in Kentucky business circles came into my uncle's tack room with the Daily Racing Form.

After reviewing the afternoon race for his horse, the owner asked my uncle for his opinion. My uncle replied, "It looks like we're in a little tough. Maybe, with your connections, it would be a good idea to boost our chances by enlisting as much public sentiment as possible. If I were you, I would get back downtown and tout as many of your friends and associates as you can reach before they head for the races. I will do my part to generate some steam here at the racetrack."

[VetScratch]

Game Theory,
But the relationship between maintaining non-linear relationships and making good probabilities is only indirect, and as you say doesn't really have anything to do with the blending process (so why'd you keep bringing it up?). I brought it up because in the Fierro quote thread (by Karl) you and Dave put forth the blending theory as if it would cure all inadequacies in fundamental handicapping techniques.

[GameTheory]

The assumption was that you've made the best lines you are capable of making, and the blending process will make any line better or allow you to exploit your lines in the best way possible.

It is a bit magical though, in that blending can take a line that you cannot make money with, and turn into a profitable tool. The reason is that if you make a line that contains infomation not accounted for by the public, it often won't be profitable on its own. However, once you've added to the public it becomes so (you've got "value-added" information you're adding to the public line). And sometimes you might find your original line isn't as good as you thought because it only contains information that the public does account for and post-blending you might find your line to more or less vanish -- it is not adding anything to the public line. For this reason, you can't really tell the profit potential of your own handicapping until after blending so you can see how much you're adding to the public...

[MV McKee]

I didn’t do a very thorough job of explaining exactly how I determine how I weight the public odds against my own odds line. I wasn’t sure there was anyone interested. I will explain my process more thoroughly below.
I made a statement in my first posting that the single most important thing I do is “handicap my handicapping”. I really don’t put much effort into attempting to determine any precise ratios relating to the weighting of various individual data elements, or determining “impact values” etc.

I can concisely explain why I have chosen to utilize the process I currently do through a series of very simple logic statements. In doing so, I will hopefully answer some of the questions posed earlier:

1) In a horse race every entrant has a chance (X as a decimal) of placing in any specific finish position. X is always >0 and <1.
2) X is a single quantitative value (not an estimate)
3) X is unknown.
4) I create an estimate of X by translating a subjective judgment into a quantitative value (M as a decimal)
5) The public creates an estimate of X by translating a subjective judgment into a quantitative value (P as a decimal)
6) M cannot be compared to X as X is unknown.
7) P cannot be compared to X as X is unknown.
(I believe Game Theory pointed out 6 and 7 above in a prior posting)
8) M can be compared to P. This comparison is valuable because P determines the return side of an investment/return equation.

The method I use to compare M to P

1) I group each entry into a “bucket” defined by the odds line (M) I assigned it to. Since I utilize a 40-point scale this is very tiered. 1 point = 40-1, 2 points = 20-1, 3 points = ~13-1 etc. With M equating to .025 (M1), .05 (M2) and .075 (M3) respectively. 20 points and upwards = 1-1 and below (3-5 and below tote odds) are grouped together (M20). All non-contenders (if any) from a single race are grouped with M equating to .05 (MN).
2) I split each of the above (M) buckets into groupings based on whether the (P) line was higher or lower than the (M) line. I create this (P) line by averaging the decimal values derived from the percentage of money wagered on the entry in the win pool, and the same decimal value derived from the exacta (1st place finish) pool. Because the exacta pool is larger than the win pool, combining the pools and deriving the decimal value from the total distribution would weight the exacta pool more heavily. I have chosen average the odds of the 2 individual calculations. This gives more weight to each dollar wagered in the win pool. I generally derive this value from the last cycle prior to the scheduled post time (<1 MTP and >0 MTP). At this point, generally only @50% final pool money has been wagered. However, in practice, this will have to be the data I base decisions on.
3) I calculate an estimate of the unknown X for each of the groupings. With a large amount of data, this estimate is very accurate.
4) The expected X for M is already known (since every value of M in a given grouping is equal. The expected X for P is easily calculated for a particular group (Sum of Values/Count).
5) IF the estimated value of X lies between the expected M and the expected P, then a line calculation based on a percentage of both M and P is indicated.
6) An optimal percentage of M and P to use can be derived using a number of methods. However, I have found it useful to look at some generalized subsets of the data. I have the luxury of a fair amount of history to work with though. Be cautious and use only very broad groupings. In my case, I have found that at Emerald Downs .7M+.3P is indicated, whereas in So Cal, I use .4M + .6P.
7) Turf races .6M + .4P, Maiden races .25M + .75P. If one breaks it down into too many categories, you are left with difficult questions. For instance, based on the above, what do I do in a Maiden Turf race? Or a Maiden race at Emerald Downs? I have learned to only group by track.

Hope someone found this useful. It has proven invaluable for me. Now, I just need to learn to explain things a little more clearly.

[Dave Schwartz]

MV,

Thanks for sharing. I would never have thought to do it your way. Very creative.

I find it amazing that three of us do something similar in three different ways.

If ever there was a place where "the end justifies the means," it would be betting on horse races.


Dave

[MV McKee]

Dave,

Thanks.

In answer to your previous question regarding whether or not I have seen my bottom-line improve since doing this (not your exact words).
Yes.
MY ROI has decreased , but my total revenue has increased substantially.
I was (in retrospect) foolishly enslaved in the practice of looking for value only in my top (ordinal) selection. Now I routinely find value in my 1st as well as my 2nd, 3rd, 4th and sometimes 5th win selections, or in a combination of them.
To be specific, my ROI has decreased by 2/3rds. But I am wagering 5 times what I was when I had a sky-high ROI.
On the surface, a 17% interest rate sounds moderately attractive, but when you consider the fact that I cycle through my bankroll roughly 50 times a year, it's a decent yearly return.

[formula_2002]

There are about 20 members that have posted to this note.

Suppose we have a little project. Not a contest but a project.
This weekend, Saturday and Sunday, everyone post as many picks as they can. One pick per race to win.

On Monday, we each analyze the results as we may see them.
There is nothing ment to be proven here.

The analysis could be more interesting than the results.

I'll offer to tabulate the results, although anyone can tabulate the results in any manner they wish.

I can tabulate the results for some of the more common factors I follow such as my speed, class, pace, form, trainer, jockey, etc.
I'll show win%, expected win % 1/ (odds+1), # of picks, win net, place net and show net. I'll also indicate the % profit/loss for flat and dutch betting.


Post your picks here at least 30 minuets prior to 1st race post time.


Just indicate track, date, horse name, program # if possible, your odds line if possible, a brief statement for conditional betting such as "fast track only" , "turf only" any odds line condition, entry stipulation if any, etc.
Win bets only.

Again, this is not nothing ment to prove anything.
Hopefully, it may offer a futher basis for conversational interchange.. I think this group can do well with that.

Joe M

[alysheba88]

I am going to Belmont Saturday. Would be more than willing to post my contenders and odds line. Sounds like fun

[VetScratch]

Originally posted by GameTheory
The assumption was that you've made the best lines you are capable of making, and the blending process will make any line better or allow you to exploit your lines in the best way possible.

It is a bit magical though, in that blending can take a line that you cannot make money with, and turn into a profitable tool. The reason is that if you make a line that contains information not accounted for by the public, it often won't be profitable on its own. However, once you've added to the public it becomes so (you've got "value-added" information you're adding to the public line). And sometimes you might find your original line isn't as good as you thought because it only contains information that the public does account for and post-blending you might find your line to more or less vanish -- it is not adding anything to the public line. For this reason, you can't really tell the profit potential of your own handicapping until after blending so you can see how much you're adding to the public... That was not the assumption in the Fierro-Quote thread. Dave said Fierro's book had helped convince him about the merits of value play. Dave said he working on his "new science of value." Blending was proposed before anyone discussed how to make "the best lines you are capable of making."

Meanwhile, very talented handicapper/programmers have been quietly making progress for decades... value is not a "new" science. In fact, the way you and Dave discussed parimutuel odds and probabilities earmarked you as rookie oddsmakers.

Many "selection" experts have exuberantly charged into oddsmaking for value players. What they soon discover is that the best odds line they can make with their "selection" methods are not very good decision-support tools for value play.

If you have found that blending is required to turn your odds lines into a profitable tool, then that just proves my point. I suggest that you need to raise your standards with respect to what a good odds line is. Having done this, I think you will find blending is "magic" with diminishing returns. You seem to refute that public opinion is influenced by a lot of bad information and also reflects purely emotional waves of buy/sell decisions similar to all public markets. I think you overstate the value of blending and understate the value of handicapping.

I also think you and Dave have been "circling the wagons" on several issues. Why can't you agree to an acceptable method of testing odds-line accuracy? I proposed the widely accepted Quirin methods from his 1979 book. You never responded.

I have followed your posts with great interest since I joined this board. I do not doubt your credibility when you describe how your homegrown A.I. software is capable of finding non-intuitive, non-obvious, "seemingly anomalous" circumstances that represent statistically profitable plays. You have said that building a personal arsenal of these plays has been the key to your success. Since you profit, no one can assert that you ignore value, but your approach sounds very much like the way the business intelligence community hits paydirt via sophisticated A.I. data mining technology. Maybe, I missed something, but I cannot recall you showing much interest in software oddsmaking.

Why don't you show us what you mean by the "best lines you are capable of making?" A few examples would be enlightening.

[JustMissed]

Hey Vet,

They probably can't show you a value line prior to the race because they adjust their lines based on the toteboard at 4 MTP(as I recall).

This would not allow enough time to transmit the blended value line before the gate opens. At least this is the excuse I would use.

The only way to prove the validity of one's line, selections and bet is to post one's unblended line and selections prior to the race(similiar to what Jim Lehane does).

Didn't mean to jump in your query, but I have also gotten a little sceptical about claims of "life changing" discoveries without any sort of result validation.

See ya,

JustMissed
:confused:

[JustMissed]

After my last post to you I went over to Jim Lehane's private web page and sure enough he had handicapped the 9th at Belmont for today. He ranked his top 3 contenders and 1 periphery play, all with morning line and value odds, with a description of why he picked 'em.

I can't post the blanket numbers because they are on a private web page but the odds line goes like this, ML/VO:

5-1 5-2

15-1 8-1

10-1 9-2

2-1 2-1

Pretty neat huh?

Anyway, just thought you might be interested in this. I may play the 9th at Belmont, if I win I will post the results, if not, you won't hear a peep from me.Ha!.

JustMissed
:)

[VetScratch]

JustMissed,

I really don't reject blending "out of hand." I mainly take issue with how it was presented as a "magical" boon.

I assert that the value of blending is inversely proportional to the accuracy of handicapped odds, and that the accuracy of handicapped odds can be measured in the same way that the accuracy of public odds have been measured for decades.

I particularly thought it was misleading to use Benter's Hong Kong results as proof of this "magic" breakthrough.

Let's put the history of blending in proper perspective before we hail it as a miracle.

Benter's tremendous financial success with blending has more to do with Hong Kong racing than was implied by either Game or Dave. Blending works best when circumstances constrain accurate computer-based oddsmaking in an environment of great opportunity. That was Hong Kong when Benter originally developed his approach.

In North America, we are obsessed with quantitative analysis (especially for sports like racing and baseball). This is why downloads from the big data providers often supply 8,000 to 12,000 data values for a field of ten horses. Outside of North America, quantitative measurements are much harder to come by. This is reflected in PPs for foreign ship-ins. The details concerning interim times, positions and lengths are often missing because they may not have been officially charted. Brohamer would have probably been carted off to Bedlam if he originally tried to sell his pace analysis methods to village horseplayers in Ireland, England, Australia, and many other countries.

Although the gap has narrowed in recent years, professional horseplaying syndicates in foreign countries either handicapped with far less quantitative data (than we take for granted) or they employed their own clockers, chart callers, and statisticians to build quantitative databases. In Beyer's original account of visiting Hong Kong, he pointedly observed that surprisingly little quantitative data was published in English, and that Chinese-language data did not appear to cover the broad scope of data that we expect to find in the DRF or downloads. I have no reason to think that "Mr. B's team" failed to collect and build better quantitative information than was available to the general public. However, when blending was "discovered," much of it's value had to do with constraints associated with deriving handicapped odds from a scarcity of cheap quantitative data. The second dimension of merit for blending was the huge parimutuel handle in Hong Kong, where Beyer said as many as 50,000 online betting machines were scattered about the city. This is what provided an environment of great opportunity without prohibitive logistical difficulties.

Less publicized than Benter were other advocates of blending where opportunity was unfortunately shunted. In England, a group of ICL programmers (called the Reading group) adopted blending as way to improve their "rough" computer-generated odds lines. In theory, they should have gotten rich. To make a long story short, they were undone by canny British bookmakers who simply barred them just like Las Vegas casinos bar professional card counters in single/double deck Blackjack games.

In addition, while a minor segment of U.K. handle was channeled through parimutuel systems, the constant change in racemeet venues presented logistical problems that they could not overcome at the time (1980s and early 1990s). Not only couldn't they effectively acquire real-time parimutuel odds, but their real opportunity to make a killing was denied when bookmakers shut them off (by recognizing their trusted agents and value-based playing patterns). Short of denial, all a bookmaker has to do to discourage exploitation is abruptly offer much lower odds to anyone they are leery of; which, when combined with "possible" layoff hedging actually shifts the edge back to the bookmaker. If you doubt that this is done, walk up as a stranger to any on-track bookmaker at Ascot and try to bet 2,000 Quid on a horse he has listed at 8/1 on his chalkboard.

[VetScratch]

JustMissed,

What may be overlooked as "value" in the odds comparisons you posted and will make judgments about prior to each post, is that you may well end up passing all races! That sure beats forcing a bet out of every race by resigning yourself to playing a "Pick" when nothing better emerges.

[GameTheory]

Originally posted by VetScratch
That was not the assumption in the Fierro-Quote thread. Dave said Fierro's book had helped convince him about the merits of value play. Dave said he working on his "new science of value." Blending was proposed before anyone discussed how to make "the best lines you are capable of making."I think it is a pretty obvious assumption -- that the person who made the line in question made it as well as they could. The alternative would mean they purposely made it worse then they knew how.


Meanwhile, very talented handicapper/programmers have been quietly making progress for decades... value is not a "new" science. In fact, the way you and Dave discussed parimutuel odds and probabilities earmarked you as rookie oddsmakers.Really? What gave us away? Our preference for speaking plainly & clearly on the subject instead of making obscure references to things like "pure handicapped odds" and "winner-selection methods" that any non-rookie would surely know means the difference between maintaining non-linear relationships and not?


Many "selection" experts have exuberantly charged into oddsmaking for value players. What they soon discover is that the best odds line they can make with their "selection" methods are not very good decision-support tools for value play.Agreed. So what?


If you have found that blending is required to turn your odds lines into a profitable tool, then that just proves my point.No it doesn't.


I suggest that you need to raise your standards with respect to what a good odds line is. Having done this, I think you will find blending is "magic" with diminishing returns.It doesn't matter how good the line is or whether it is already profitable or not. If it is ignoring the real-time odds, it can be improved. Obviously if line A is better than line B (both ignoring odds) then line A has less room for improvement than B because there is a finite upper limit that is acheivable by any method. All I'm saying is you cannot reach that upper limit while ignoring odds.


You seem to refute that public opinion is influenced by a lot of bad information and also reflects purely emotional waves of buy/sell decisions similar to all public markets.No, I don't refute that. If the public line didn't have bad information in it, you could never profit. Another way to look at the blending process is that you are improving the *public* line with your adjustments. The idea is to end up with a line that is better than both your original one & the public's.


I think you overstate the value of blending and understate the value of handicapping.Nonsense. It is good handicapping that allows you to make a successful blended line. One thing you are leaving out of the discussion is that your handicapping needs to be different from the public as well as good. If I handicap using factors X, Y, & Z which are mildly predictive but the public doesn't consider much, that is liable to generate more profit than if I use speed figures (say) or something the public considers very heavily. From a pure handicapping perspective, speed figures may much better assess the relative chances of the horses than do factors X, Y, & Z, but chances of making money are much better using the weaker but less considered factors. So you need to have quality handicapping, but also handicapping that does not mirror the public's.


I also think you and Dave have been "circling the wagons" on several issues. Why can't you agree to an acceptable method of testing odds-line accuracy? I proposed the widely accepted Quirin methods from his 1979 book. You never responded.I didn't know my job was to present a united front with Dave on every issue. I have no idea what his opinion on judging odds lines is. And I actually don't even know what Quirin's method is either since I haven't looked at his stuff in about 8 years, and I don't have his book around here anywhere. But you obviously weren't paying attention to what my opinion was, because I made it very clear. I even gave you a link to a previous post (from the thread where we debated blending last time -- before your time) where I offered an exact mathematical formula (entropy-based) for scoring odds lines and explained (one of the many times I have explained) why just looking at observed/expected groups doesn't work.


I have followed your posts with great interest since I joined this board. I do not doubt your credibility when you describe how your homegrown A.I. software is capable of finding non-intuitive, non-obvious, "seemingly anomalous" circumstances that represent statistically profitable plays. You have said that building a personal arsenal of these plays has been the key to your success.Well, not quite. I maybe was unclear or was mixing past success with future (hopeful) success. I should explain that although I've been working on A.I. methods applied to horse racing for 10 years now, it is only in the last couple of years or so that I've become a "computer handicapper". It took me many years to figure out how to get these methods to work successfully, and improvements in computer-power and availability of data made a big difference. So although I was dabbling all those years, that isn't how I was betting. I was a DRF, pen-and-paper spot play guy, focusing mainly on trainers, claims, layoffs, pedigree, turf, & maidens. (Those are probably the plays I was referring to.) I was never very good at comprehensive handicapping just on my own. When we were talking about some of those obscure computer discoveries, those were new discoveries. All part of the project I've been working on to fully automate the whole process. My goal is for me personally to do no handicapping whatsoever.


Since you profit, no one can assert that you ignore value, but your approach sounds very much like the way the business intelligence community hits paydirt via sophisticated A.I. data mining technology. Maybe, I missed something, but I cannot recall you showing much interest in software oddsmaking.I'm using sophisticated A.I. data mining technology, so it's no wonder. Why this would restrict my interest in making good odds lines I can't fathom. After making some good discoveries, I have to be able to exploit them for profit, after all.


Why don't you show us what you mean by the "best lines you are capable of making?" A few examples would be enlightening.That is not possible presently because the system is not that far along. I could offer some "proof-of-concept" odds lines and then the blended versions, but they would be neccessarily contrived and wouldn't prove anything to the skeptical. Which is one reason why I offered to use YOUR odds lines (on past races) since you think yours are already so great that blending won't be helpful.

[GameTheory]

Originally posted by VetScratch
I really don't reject blending "out of hand." I mainly take issue with how it was presented as a "magical" boon.

I assert that the value of blending is inversely proportional to the accuracy of handicapped odds, and that the accuracy of handicapped odds can be measured in the same way that the accuracy of public odds have been measured for decades.

I particularly thought it was misleading to use Benter's Hong Kong results as proof of this "magic" breakthrough.

Let's put the history of blending in proper perspective before we hail it as a miracle.


I assert that maybe instead of all this tangential BS, you look at some actual evidence. i.e. try it and see.

Everything you assert is based on how clever you think you are. My assertions are based on experience. It sounds fishy to you, so you spend days and days arguing about whether it *might* be a good idea or not. The thing is, there is no need to rely on intuition or theory here, because it is quite easy to see results on real data. Just look at your own data and see! What are you so afraid of?

[Foolish Pleasure]

You're right Slam it's easy.

Hey good luck with the ponies.


BTW while your laying vs bad chalk,

I'll be signing for big gimmicks.

Afterall juice is the end all right?

LOL, can't say I didn't try buddy, maybe next time you consider some of us have been swimming in the pool for decades.

Fastracehorse's post should immediately go into the racing posting HOF.

[Foolish Pleasure]

BTW SLam,
the big winners in racing post at our home.
believe or not. If you don't you can always ask Dana.;)

[VetScratch]

Game,

When you are further along with your system and Dave has completed his new science, I'm sure I will see the light. In the meantime, I will ponder how I overlooked entropy, which must explain why you guys pegged 33% horses at 3/1 odds when this whole debated started. Does entropy explain what gets lost on the way through the parimutuel system? Shame on me, I always thought takeout and breakage were the culprits!

Seriously, I will go look for that link to the older thread. :) :)

BTW, If your odds before blending cannot be used profitably, I submit that my point stands... you are not making the best odds that you can make... are understating the importance of fundamental handicapping... are overstating the value of blending.

[VetScratch]

Or even 2/1 after adjustment for editing!

[JustMissed]

Sorry Vetscratch to be so late getting back to you. I went down the road to Tampa Bay Downs to cash my ticket and got to thinking about "blended lines" and stopped in for a little "blended whiskey" and time kinda got away with me. I haven't had any
7&7 since grad school and they tasted mighty good.

Any way, I can now reveal the Jim Lehane picks for the 9th at Belmont today. In order by number, Name, Morning Line Odds & Value Odds, his picks were:

1. Lady Bi Bi, 5-1, 5-2
5. Run Alexis Run, 15-1, 8-1
11. Courting Visions, 10-1, 9-2
4. Little Buttercup, 2-1, 2-1, Jim's Periphery play

The order of finish with final odds was:

1. Lady Bi Bi, 2.85
4. Little Buttercup, 2.30

Lady Bi Bi paid $7.70 for the win.

Anyway, I am not an expert on "value plays" and cannot comment as to whether or not a value line of 5-2 that goes off at 2.85 ,and wins, is a value play or not but I can only comment about a few items:

1. Jim Lehane has enough balls to publically anounce his picks and "value line" and gets my respect.

2. If you support a horse that you absolutely believe will win, you will walk out with more money than you walked in with with.

3. Blended whiskey taste pretty good.

Anyway, I know you want get any of your boys to post their picks or blended value lines but hang in there, you're doing a great job.

JustMissed
:)

[VetScratch]

Originally posted by GameTheory
I thought the purpose was to compare your probabilities (adding up to 100%) with the *prices* given by the toteboard (adds up to more than 100%).

If I make my line so it adds up to 118% or so, then my 10% horses will be equal with the toteboard's 10% horses, and it would then appear I have a fair (break-even) bet, when of course that is not true. The public has to underestimate the horse's chances merely to create a break-even bet.

Actually, I think we all agree, but VS misunderstood Dave in her zeal to disagree with him... Just to clarify things, this is where things started in the Fierro-Quote thread. The parimutuel probabilities always add up to 100% (or something is crooked) and if my line adds up to 118%, then I have not made a parimutuel line. Your 10% horses should be about 7.30/1, just like they are at the racetrack. If your 10% horses are 9/1, you haven't made a parimutuel line. Zeal has nothing to do with it.

(1.18/.10)-1 = 10.8/1 by your 118% method which ignores takeout, and this would be even worse than 9/1, and further from 7.30/1.

[formula_2002]

Vet..
Let me add this simple report to the discussion.

It's the results for all my "basic model plays". Includes back fitted and "untested data".
Public returned a loss of 14% in 42688 plays.
Basic model returned a $2 loss in 6279 plays.

The untest data (over 1600 model plays, returned a loss of 2%)
all without using, as Benter would say "combining the probability estimate" with the public's.






ODDSRN DBT WNET sys dbt SYS WNET
1 577 -43 223 -20
2 2104 -249 1108 -129
3 3018 -279 1425 -91
4 3150 -416 995 31
5 3150 -452 702 -79
6 2930 -667 467 -78
7 2985 -256 354 115
8 2856 -126 258 110
9 2771 -528 176 2
10 2595 -312 131 -11
11 2398 -352 95 24
12 2162 -400 80 37
13 2092 -394 50 32
14 1799 -306 49 26
15 1767 -117 36 -4
16 1477 -173 40 11
17 1408 -206 27 28
18 1391 -80 29 -10
19 1360 -268 26 14
20 694 -363 8 -8
42688 -5988 6279 -2
-14% 0

[SLAM DUNK(MW)]

FP


You're right Slam it's easy.

--------
it isn't? Determining that horse 1 should be 2-1 and horse 4 should be 6-1 is no problem, but that horse 3 should be 12-1 and horse 7 should be 20-1 is some form of rocket science. Moreover, it is a little hard to determine whether horse 1 should be 2-1 or 3-1 or 8/5 unless you have gone through the entire field. Not challenging your expertise in horses(you obviously have plenty), just trying to figure out why you think quantifying what the chances of a non-contender is such a difficult process.

-----------------

Hey good luck with the ponies.

-----------------

Hey thanks, but isn't it skill that ultimately makes you a winner over the long haul?

---------------------


BTW while your laying vs bad chalk,

I'll be signing for big gimmicks.

--------------------


Hey if dialing up big $ superfectas and pick 4's is your game and makes you rich, go on with your bad self, I have to resort to the lowly grinding game myself.

-------------

Afterall juice is the end all right?

---------------


I think that is your point. As I recall you are always preach efficient market, the take is too high(which of course I agree with). Making a line is just simply a tool to avoid making negative expectation wagers. Whether you win or lose of the long run just depends how good a line you make day in and day out(or blend according to some of the guys in this thread). Also I just find that I make SMARTER wagers when I have my line in front of me. It is when I take the races on without a line I make a lot of mistakes(finding value that isn't there, overbetting, missing key capping points.....)

--------------------------------

LOL, can't say I didn't try buddy, maybe next time you consider some of us have been swimming in the pool for decades.

Fastracehorse's post should immediately go into the racing posting HOF.

------------------------------

Okay, I'll bite. What is a better alternative than making a line? I have played this game a long time and no matter what I try, I keep going back to making a line. Not saying that it is the best or only way to play horses, but it sure works best for ME and ultimately that is all that matters.

[GameTheory]

Originally posted by VetScratch
When you are further along with your system and Dave has completed his new science, I'm sure I will see the light. In the meantime, I will ponder how I overlooked entropy, which must explain why you guys pegged 33% horses at 3/1 odds when this whole debated started. Does entropy explain what gets lost on the way through the parimutuel system? Shame on me, I always thought takeout and breakage were the culprits!

Seriously, I will go look for that link to the older thread. :) :)
Another one I can't make sense of. But it would be nice if you actually read my posts before responding how I'm wrong...


BTW, If your odds before blending cannot be used profitably, I submit that my point stands... you are not making the best odds that you can make... are understating the importance of fundamental handicapping... are overstating the value of blending. When I say the best odds you (or anyone) is capable of making, I mean it literally, and once again you are taking a very simple statement and twisting it every which way. Notice the word "capable" -- I am assuming that when someone makes a line they are doing it to the best of their current ability. Maybe their ability is great, maybe it is little -- it is a relative capability based on the knowledge and skill of the person.

If I can take an oddsline that cannot be used profitably, and then blend it with the public and the new line CAN be used profitably, how does that prove your point? It would seem to me that it would prove mine.

[GameTheory]

Originally posted by VetScratch
Just to clarify things, this is where things started in the Fierro-Quote thread. The parimutuel probabilities always add up to 100% (or something is crooked) and if my line adds up to 118%, then I have not made a parimutuel line. Your 10% horses should be about 7.30/1, just like they are at the racetrack. If your 10% horses are 9/1, you haven't made a parimutuel line. Zeal has nothing to do with it.

(1.18/.10)-1 = 10.8/1 by your 118% method which ignores takeout, and this would be even worse than 9/1, and further from 7.30/1. I still don't understand this either, but I guess that is because I'm a rookie.

Since you have no interest in actually dealing in evidence and only want to argue nonsense and take things out of context ad infinitum, I'm through with you. Are you just another troll like Joe? (Don't bother answering that.)

[VetScratch]

MV McKee,

About your last detailed post...

P = Public line
M = My Line (handicapper's line).

Given the furious posting pace in this thread, I hope everyone took a close look at what you are doing. Disregarding the issue about what is a good handicapped odds line (M), your approach to blending P with M presents a very meaningful consideration for anyone who blends.

The basic concept of blending is that P reflects awareness of unpublished performance factors that the handicapper cannot know and cannot incorporate into M.

You point out that ratio of known to unknown factors should vary (much like par expectations) from track to track and among race types. Two simple ways of looking at this are that:
(1) What should be unknown to handicappers should be greatest where the least data is generally available, such as Maiden races.
(2) What should be unknown to handicappers should also be greatest where unknown/unpublished considerations have the most opportunity to precede a race, such as SO-CAL where starts-per-season are below average, gaps between starts are above average, and trainers are more apt to quietly condition for races rather than race for conditioning.

Your research and blending ratio examples sure seem to support this important consideration, and your 17% profit is the proof of the pudding!

[VetScratch]

Game,

Lighten up! I know you are not a rookie just as you should have known that ZEAL was not my motive for correcting the assertion that 33% horses should have 2/1 odds and 10% horses should have 9/1 odds. We see odds in result charts, at racetracks, at OTBs, and on Internet toteboards... in each case they are ALL parimutuel odds.

What goes around, comes around!:)

[formula_2002]

FOR THE ANALYSIS PROJECT

BASIC MODEL (WITH ODDS CONDITIONS & MINIUM ODDS SHOWN)
TOP PICK (A) MUST BE >=2.5 AND EXCEED THE MINIMUM ODDS.
IF THE TOP PICK IS A PLAY THEN;
OTHER PICKS (B) IN THE RACE ARE PLAYABLE IF ODDS>=2.5 AND EXCEED THE MIN. ODDS
UNACCOUNTED SCRATCHES MAY AFFECT PICK SELECTION.

TRACK DATE RACE # HORSE NAM PG# ML MIN.ODDS PICK
BEL20031004 1 SAFARI 2 3.00 1.50 A
BEL20031004 2 MANHATTAN 6 3.00 4.71 B
BEL20031004 2 MY LEGAL 1 4.00 2.33 A
BEL20031004 3 BWANA CHA 7 2.50 3.63 B
BEL20031004 3 CARTAGE A 4 4.00 1.18 A
BEL20031004 5 DREXEL MO 10 2.00 3.33 A
BEL20031004 7 CHAPEL RO 8 1.40 3.73 B
BEL20031004 7 PADDINGTO 2 10.00 3.33 A
BEL20031004 9 SIGHTSEEK 2 0.60 2.31 A
BEL20031004 10 PRETTY WI 1 1.60 5.00 B
BEL20031004 10 ROARING F 3 5.00 3.67 A
BEL20031004 14 SIGHTSEEK 2 0.60 2.31 A
CRC20031004 1 REAL HOT 6 3.00 1.44 A
CRC20031004 2 MISTI LIG 5 6.00 3.44 B
CRC20031004 2 RARE HONO 4 3.00 2.33 A
CRC20031004 3 INNOCENT 3 2.50 3.20 B
CRC20031004 3 PRINCESS 7 3.00 2.82 A
CRC20031004 8 AEROPETER 2 6.00 4.57 B
CRC20031004 8 WESTERN M 1 3.00 4.57 A
CRC20031004 9 VISION IN 6 2.50 3.09 A
CRC20031004 10 SWEEPING 1 2.50 4.29 B
CRC20031004 10 WHEATON'S 8 4.00 4.29 A
CRC20031004 11 GOLDEN O' 6 3.00 3.20 A
CRC20031004 11 WINNING F 2 8.00 5.00 B
CRC20031004 12 FOXY GUY 7 5.00 3.22 B
CRC20031004 12 PAR RULES 5 3.00 2.46 A
CRC20031004 13 PLANTATIO 4 1.60 1.00 A
KEE20031004 2 RUPERT HE 4 4.00 1.56 A
KEE20031004 3 SAVANNAH 1 3.50 3.40 A
KEE20031004 4 A LITTLE 6 4.00 3.50 A
KEE20031004 6 VALUE PLU 3 2.50 3.08 A
KEE20031004 7 REVVED UP 7 3.50 7.75 B
KEE20031004 7 ROCHESTER 9 5.00 7.75 A
KEE20031004 8 POLLARD'S 6 3.00 3.17 A
KEE20031004 9 TRADEMARK 1 6.00 2.07 A
KEE20031004 10 ART VARIE 2 1.60 3.30 A
KEE20031004 10 DOC D 8 12.00 5.14 B
KEE20031004 11 VALUE PLU 3 2.50 3.08 A
KEE20031004 15 TRADEMARK 1 6.00 1.81 A
PIM20031004 2 INDIXIE 1 2.50 1.85 A
PIM20031004 3 REVERENCE 14 2.00 3.64 A
PIM20031004 4 BERANJIZ 11 8.00 5.17 A
PIM20031004 5 PRIVATE C 14 2.00 7.00 B
PIM20031004 5 WESTERN S 3 2.50 5.22 A
PIM20031004 6 FUNNY FAR 2 5.00 1.54 A
PIM20031004 8 HOLY BUBB 2 2.50 2.86 A
PIM20031004 8 RANDAROO 4 4.50 3.91 B
PIM20031004 9 BURT'S CO 1 8.00 5.22 A
PIM20031004 9 ETCETERA 4 10.00 7.00 B
PIM20031004 10 AMERI PRI 10 6.00 4.38 B
PIM20031004 10 WILD CHAR 8 2.50 3.78 A
SAX20031004 1 T.J. DIAM 4 3.50 2.90 A
SAX20031004 2 FURTHER 3 15.00 1.92 A
SAX20031004 3 DIXIE THR 4 1.60 3.00 B
SAX20031004 3 FORLAAN 5 2.00 2.27 A
SAX20031004 4 DOMESTIC 2 10.00 3.44 A
SAX20031004 4 SWEET RET 6 2.50 4.00 B
SAX20031004 5 PLEASANTL 2 2.50 5.29 B
SAX20031004 5 TASTE OF 4 15.00 3.40 A
SAX20031004 6 BOBBY DAZ 2 15.00 3.27 B
SAX20031004 6 BUT OF CO 3 2.00 2.62 A
SAX20031004 7 ROUSING A 1 3.00 1.06 A
SAX20031004 8 ARDUM REL 6 4.50 5.83 B
SAX20031004 8 SUSIE JOE 3 6.00 2.42 A
SAX20031004 10 ROI CHARM 6 3.50 1.77 B
SAX20031004 10 ROSBERG 8 6.00 1.77 A
SAX20031004 13 DOMESTIC 2 10.00 3.44 A
SAX20031004 13 SWEET RET 6 2.50 4.00 B
SAX20031004 16 PLEASANTL 2 2.50 5.14 B
SAX20031004 16 TASTE OF 4 15.00 4.38 A
SUF20031004 2 GRAY MORN 4 3.50 4.00 B
SUF20031004 2 SISTER RO 2 2.50 2.57 A
SUF20031004 4 CHANCES A 6 1.80 4.11 A
SUF20031004 4 HESA DEVI 5 15.00 4.75 B
SUF20031004 5 MINER'S T 1 1.40 1.69 A
SUF20031004 5 YUKON LOO 3 6.00 3.30 B
SUF20031004 6 BALLADO M 2 2.00 3.38 B
SUF20031004 6 IMMORTAL 5 4.00 3.38 A
SUF20031004 8 BINGO LIZ 1 8.00 4.67 A
SUF20031004 8 SARAH'S C 2 2.50 5.38 B

[Dave Schwartz]

Formula,

The real question then, would be "How did the C-D-F picks do during the same period?"


Dave Schwartz

[Rick]

This is amazing! The subject of incorporating public odds, or "blending" if you will, has only been rarely mentioned until very recently and now we've got a debate about it's merits. Naturally though, I'll have to throw my two cents worth in on the subject.

The idea of incorporating public odds is to improve your probability estimates, so what you really want to do is incorporate the "smart" money and bet against the "dumb" money. By "smart" money I mean the proportion of the money that predicts probabilities better than average. In US parimutuel pools the odds fluctuate quite a lot in the last few minutes and even after post time. So, the question is whether there exists a period of time that will give you good information about the "smart" money and still give you enough time to make a bet.

Some have suggested that there is a period from about 3-6 minutes to post where the additional money is more predictive than average so it may be possible to use that. Other studies suggest that, overall, early money may be less predictive than average for example at 10 minutes to post and earlier. Anyway, my point is that if the quality of money as a predictor varies over time, then one ought to specify exactly at what time he incorporates it as an additional variable in his odds model in order to determine it's true effect.

In my own betting, I totally ignore the toteboard and just bet according to the morning line. While that procedure is primarily motivated by laziness I find that it makes very little difference in the overall results. Since about 80% of the variance in public odds is explained by the morning line, and one could harldly consider it "smart" money, I bet against it as if it was a good measure of "dumb" money. If the "smart" money reduces the odds on my selection, I assume that the probability will go up enough to compensate for it.

If I were to follow the toteboard I'd probably be betting against 10 minute odds, and if I really wanted to create maximum stress for myself I'd probably try to incorporate late money from some time period into my odds in order to make them more accurate. I might even add the morning line here in order to further filter out the "dumb" money. This procedure could lead to the seemingly unnatural result of calculating overlays against odds from an earlier time period.

I don't know how much of this is true, or significant enough to use in a practical sense, but I just thought I'd throw out some more possibilities just to complicate the issue further. Have fun with the debate.

[VetScratch]

Game,

Quotes:
--------------------------------------------------------------------------------
VetScratch:
I also think you and Dave have been "circling the wagons" on several issues. Why can't you agree to an acceptable method of testing odds-line accuracy? I proposed the widely accepted Quirin methods from his 1979 book. You never responded.
--------------------------------------------------------------------------------
Game Theory:
I even gave you a link to a previous post (from the thread where we debated blending last time -- before your time) where I offered an exact mathematical formula (entropy-based) for scoring odds lines and explained (one of the many times I have explained) why just looking at observed/expected groups doesn't work.
--------------------------------------------------------------------------------
If you don't like Quirin's methods, I'm open to considering whatever you meant, but I searched the Fierro-quote thread in vain for the link you referenced. Where should I look?

[GameTheory]

Originally posted by VetScratch
Game,

--------------------------------------------------------------------------------
If you don't like Quirin's methods, I'm open to considering whatever you meant, but I searched the Fierro-quote thread in vain for the link you referenced. Where should I look? I didn't say I didn't like his methods, I said I didn't know what his methods were. The link you are looking for is in this thread, not the other one. First post by me in page 2. Once you follow that link, you will be in the middle of an old thread where we were talking about blending also (in the dark days of ignorance before you arrived on the scene).

[formula_2002]

Just a brief recap.

Full analysis asap

Per defined odds structure
'A' picks: 14 plays, 29% wins, 80% profit
'B' picks: 12 plays, 17% wins, 27% profit.



All plays at all odds;

60 picks, 23% wins, 11% profit

[Dave Schwartz]

Formula,

So, if I understand what you are saying here, you have the following:



A 14 +80%
B 12 +27%
C-F 34 -23%




Does that seem about right? Although this is a small sample, this is pretty close to what I am seeing selecting and adjusting "my way" ...


A +70%
B +12%
C-F -21%


Dave

[Fastracehorse]

<In my own betting, I totally ignore the toteboard and just bet according to the morning line.

+++++++++++++++++++++++++++++++++++

Do you mind expanding somewhat on the above??

:confused:

fffastt

[VetScratch]

Game Theory,

Thanks for the thread link: http://www.paceadvantage.com/forum/showthread.php?s=&postid=36443#post36443

I read the thread and now see why you fuss over using the word "accuracy" to describe the characteristic that most odds studies measure when they compare expected-win-percentages to actual-win-percentages.

Let P represent the probability represented by a winner's odds. Then your vaunted "entropy-based" method is take the average of Ln(1/P)/Ln(2) as a measure of surpise? The smaller the result, the less the degree of surprise? Therefore, accuracy is greatest when surprise is least.

Your formula is just an expression of variance from 100% probability. As your probability approaches 1 (100%), your result approaches 0 (zero), or perfect accuracy (no suprise).

Now I see where your concept of "entropy" comes into play... between your ears!

You are basically asserting that probabilities don't apply to race results like they govern games of chance, such as craps.

This is not a radical misconception. Many folks who undertand that probabilities apply to rolling dice do not accept that the winner of a horserace is a probability proposition.

When we roll dice, the geometry of the dice doesn't change, but even small differences in the way we roll them will start a chain reaction of chaotic events such that the central tendency is to obey the laws of probability.

In the case of a set of races repeated between the same field of horses, an almost infinite number of performance-related characteristics for both the racetrack and betting interests may change in the intervals between repeating races. The same horse will not win all repeated races. Thus, horseracing is characterized by plenty of entropy, and this invites probability theory into the picture.

[GameTheory]

That's all fine and good, but you're using the definition of entropy from the wrong field of science. In information theory, it is something different. In fact, John von Neumann suggested that Claude Shannon (father of information theory) call it "entropy" because no one really knew what it meant. (The spirit of the meaning is somewhat the same, but entropy in information theory is something that can be quantified precisely.)

And yes, it measures the degree difference from 100% accuracy (from the "perfect" line of making the winner 1.0 and everyone else 0.0). But the rate of change is important as well. The formula is a short-hand version (suitable because we really only need to measure the probability of the winner instead of all the probabilities) of the formula called "relative entropy", also known as the Kullback-Leibler distance. It is a distance metric (but not a true one because it is not symmetric) that measures how good a substitute one probability distribution is for another. The lower the value, the shorter the distance.

So in a horse race -- the winner having 1.0 and everyone else having 0.0 in our ideal probability distribution -- the relative entropy formula tells us how good a substitute our actual value line was for the perfect line. This places a premium on giving the highest probability that you can to the winners while still making lines that add up to 100%. While looking at observed/expected only tells a part of the story because the bigger the "spread" in your lines (avg difference between the high & low probabilities) the harder it is to maintain strong correspondence. However, the bigger the spread you can muster, the more profitable your line will be if you can maintain accuracy (more & bigger overlays will emerge). Attempting to maximize observed/expected correspondence will tend to take you in the opposite direction -- to giving all horses probabilities not very far apart (because 100% correspondence is achieved by giving all horses in a race the same probability).

So the minimizing of relative entropy will take you in the right direction, but maximizing observed/expected correspondence will not.

[formula_2002]

Dave, I don’t know if there are C,D,E,F..... picks in any race.

Every race has an "A" pick. Some races also have a "B" pick. Those picks are only playable when they satisfy the odds conditions I have outline.
If "A" qualifies for a play, "B" MAYBE qualified for a play proving it also meets the odds criteria.
The "B" pick is never a play without the "A" pick being a play.

The main thrust of my analysis is based on odds ranges.

I calculate the public's win % in the various odds ranges and my win % in the various odds ranges. Then I calculate the differences between the two.
Then I calculate the STD dev for the differences and then determine the "Z scores" for each difference.

I'll post that table some time after Sunday’s races.

I have not attempted to calculate the formula for the relation ships between the public's final odds any my odds line..yet.

I guess that's what Benter calls "The combined probability estimate"

I don't understand his formula. I'll try regression analysis.
Do you have any ideas on this?

Thanks

Joe M

[formula_2002]

FOR THE ANALYSIS PROJECT

TOP PICK (A) MUST BE >=2.5 AND EXCEED THE MINIMUM ODDS.
IF THE TOP PICK IS A PLAY THEN;
OTHER PICKS (B) IN THE RACE ARE PLAYABLE IF ODDS>=2.5 AND EXCEED THE MIN. ODDS
UNACCOUNTED SCRATCHES MAY AFFECT PICK SELECTION.


TRACK DATE RACE # HORSE NAM PG# ML MIN.ODDS PICK
BEL20031005 1 COUSHATTA 8 5.00 3.09 B
BEL20031005 1 Y TWO J 1 4.00 2.46 A
BEL20031005 3 AFTER THE 3 6.00 2.70 B
BEL20031005 3 ENDER'S S 4 2.00 2.08 A
BEL20031005 4 CHARM APP 7 10.00 3.22 A
BEL20031005 4 CIMARRON 9 3.00 3.75 B
BEL20031005 6 MOLOKO 11 2.00 2.91 A
BEL20031005 7 NORTHERN 4 3.00 2.64 A
BEL20031005 7 SAARLAND 1 1.60 2.92 B
BEL20031005 10 ON THE BU 15 4.00 5.71 A
CRC20031005 1 DUCE'S IM 3 4.00 3.33 B
CRC20031005 1 TAIWAN CH 6 2.50 2.55 A
CRC20031005 2 KING'S FA 8 3.00 3.73 A
CRC20031005 2 SENDA 2 5.00 4.78 B
CRC20031005 4 BYARS OF 5 3.00 4.14 B
CRC20031005 4 SPEEDY RU 4 2.50 3.50 A
CRC20031005 5 TERMINATI 3 4.00 3.30 A
CRC20031005 6 MAIDEN TO 1 2.00 0.90 A
CRC20031005 6 SAVANNAH 6 4.00 3.75 B
CRC20031005 7 CHIEF EXC 3 3.00 3.60 B
CRC20031005 7 GOOD EXPE 1 2.50 3.18 A
CRC20031005 8 EXCLUSIVE 2 3.00 3.78 A
CRC20031005 9 HEART OF 7 6.00 2.18 A
CRC20031005 10 CONSENT 5 3.00 3.00 A
CRC20031005 10 YAKIMA CA 4 8.00 4.14 B
DEL20031005 1 ELABORATI 2 2.00 3.00 B
DEL20031005 1 THUNDER B 6 3.00 2.08 A
DEL20031005 2 MORETHANA 6 8.00 3.10 A
DEL20031005 2 SEATTLE G 8 3.50 4.13 B
DEL20031005 6 NIMMER 6 2.00 3.50 A
DEL20031005 8 HEAVENS B 8 8.00 8.75 B
DEL20031005 8 WIRED N R 3 3.50 1.17 A
DEL20031005 9 LILITH'S 8 3.00 4.71 B
DEL20031005 9 PTICHKA 7 2.00 4.00 A
KEE20031005 2 WORLD TOU 1 6.00 5.13 A
KEE20031005 3 KENAI RIV 9 5.00 4.25 A
KEE20031005 5 EMILY RIN 7 5.00 2.23 A
KEE20031005 5 MISS LODI 6 2.00 3.67 B
KEE20031005 6 BIEN NICO 1 1.20 1.65 A
KEE20031005 8 YOU 2 2.50 1.41 A
KEE20031005 9 CRIMSON H 7 5.00 4.10 B
KEE20031005 9 SPANISH E 10 4.00 3.25 A
PIM20031005 1 MARY BOYD 1 2.50 4.71 B
PIM20031005 1 PRINCESS 4 3.00 1.67 A
PIM20031005 2 POKER DIC 7 1.40 1.67 A
PIM20031005 3 BANKER BE 11 4.00 5.57 A
PIM20031005 3 HE'S A BR 12 5.00 6.67 B
PIM20031005 5 JEREMIAH' 13 10.00 4.50 A
PIM20031005 5 PAS DE ME 5 3.00 6.33 B
PIM20031005 6 DEVIL MAN 1 0.80 3.00 A
PIM20031005 6 MALIBU MA 1 0.80 4.71 B
PIM20031005 7 MADAME GL 8 8.00 3.27 A
PIM20031005 7 STARRY DE 6 2.50 4.88 B
PIM20031005 8 MISS CASH 4 3.50 3.63 B
PIM20031005 8 PARSIPPAN 1 1.00 2.08 A
SAX20031005 1 RESPLENDE 1 2.00 2.18 A
SAX20031005 6 THIRD HAL 2 5.00 3.20 B
SAX20031005 6 YOUNG PIO 3 2.50 3.20 A
SAX20031005 7 CAPTAIN S 6 2.00 2.80 A
SAX20031005 10 LUCKY PUL 2 20.00 4.22 A

[VetScratch]

Game,

Handicapped probabilities are kin to actuarial probabilities.

Actuarial science properly asserts that there is "some" probability that neither of us will live to make another post. Conversely, there is no horse that has 100% probability of winning until the after race is done.

[VetScratch]

Reading this thread, it occurred to me that odds blenders may have conducted extensive studies of late-odds fluctuations.

Suppose I have horseracing data from before and after 100,000 races.
1) A handicapping database consisting of 100,000 past performance downloads.
2) A race results database for the same 100,000 races, where takeout averages about 17%.

Suppose I have a handicapping program that:
(1) Has absolutely NO access to the results.
(1) Does have access to late program change scratches.
(2) Identifies two (2) to five (5) contenders per race.
(3) Predicts fair-value parimutuel odds for each contender.

Suppose I propose Wagering-For-Dummies rules:
(1) Bet within two (2) minutes of post-time.
(2) Pass any race that includes a horse with parimutuel odds under 3/2.
(3) Flat-bet $20 on ANY and ALL contenders where track odds exceed projected odds.

Suppose such a simple system, without dutching, tests profitably against 100,000 race results.

Is there credible evidence that late odds fluctuations will have a "tendency" to defeat such a system?

Has anyone studied odds fluctuation patterns that occur after let's say two (2) minutes to post?

[formula_2002]

Vet, it may be worth mentioning here.
I have studied 1000's of races with two minuets to post time.


The study was looking for overlays in the exacta pool based upon the win pool odds.

What I found was that the final win pool odds trended to the exacta pool
"Win odds"

That is to say, if a horse was 3-1 in the win pool with 2 min to post, and at the same time was 8/5 in the exacta pool, then final win pool odds tended to be much closer to 8/5 then 3-1.

To make the comparison, I normalize the win pool and exacta pool odds, each adjusted for their respected take-outs.

I wrote a program that scans in the on line exacta pool matrix, and then makes all the calculations.


Joe M

Ps. I still have a day job.

[sjk]

Vet,

I bet around 2 minutes to post using overlays on my line more or less as you described. The difference is that I require that the track odds be greater than my odds by a factor (I use 175% of my line as the cut-off). This is the percent that maximizes my profit as judged by experience and significant back testing. I believe it is cutting things to way too fine to use a factor of 100%.

As I pointed out in an earlier post, using my line and 175% is mathematically equivalent to using a blended line (25% mine and 75% public) and a factor of around 120%. It still seems to me that you can effectively be blending by changing your factor.

Late fluctuations have not defeated me; perhaps due to the choice of factor. Experience tells me that horses which are 6-1 or less on the morning line and which are higher than the morning line at 3 minutes to post, will generally come down as post time approaches. It is also my experience (as pointed out above) that exactas, especially on short priced horses, which are paying significantly more that the parley generally come down as post time approaches.

I often delay my bets when experience tells me the price will come down.

[Dave Schwartz]

Formula,

Yes, I have lots of thoughts on this subject.

I have posted the simplest ones here. The deeper ones need to remain proprietery at this point.


Dave

[Rick]

fast,

I just throw out any top selections that are also the morning line favorite. Those only lose about 10% so it's also hard to bet against them and exactas with the 2nd choice on the bottom seem to be the best way to go. I'm also testing a filter that throws out about half of the rest that only break even, but that's another story. I could also throw out 2nd favorites or demand minimum ML odds of 4-1 or better and boost the ROI but doing that also lowers the win % more than I'd like. Eventually I'd like to come up with a better method of identifying "dumb" money to bet against, probably involving speed ratings and odds of recent races.

Anyway, based on many years of experience with different methods, I still think that the easiest way to get genuine overlays with a good method is the throw out the ML favorite. Try it with whatever you're doing now. It's almost sure to improve the bottom line.

[JustMissed]

[QUOTE]Originally posted by Rick
[B]fast,

I just throw out any top selections that are also the morning line favorite. Those only lose about 10% so it's also hard to bet against them and exactas with the 2nd choice on the bottom seem to be the best way to go.[END QUOTE}.

You are kidding aren't you? What you said would mean that the inverse of only losing 10% would be only winning 90%. What the heck kind of system would cause a player not to bet on horses that win 90% of the time.

JustMissed
:confused:

[formula_2002]

Originally posted by Dave Schwartz
Formula,

Yes, I have lots of thoughts on this subject.

I have posted the simplest ones here. The deeper ones need to remain proprietery at this point.


Dave
Dave, if you would

The simple one increase your profit from to___?
The deeper ones increase your profit from to ____?

Joe M

[Rick]

JustMissed,

What I meant was that they lose about 10 cents on every dollar bet. Their win % is very high however and I'm sure there's some subset of them that's profitable. I just haven't found it yet. Morning line oddsmakers are usually very good at picking winners (better than 30%) but the public nearly always bets them down so that they lose 20 cents or more on the dollar overall. Actual favorites do a little better than that both ways so it's actually less efficient to eliminate them.

[VetScratch]

Formula & Sjk,

Thanks for the feedback.

I will be saving probabilities/odds before post-time in the future, but like most everyone but a few of you out there, I haven't historically been saving pre-post probabilities/odds (just past performances and race results, including final odds).

Sjk, the fair-value odds I mentioned are really breakeven parimutuel-style odds, and in my own play, I also require a premium like you do. To keep it simple (i.e., no calculators or laptops required at the track or OTB), I look for a minimum of one toteboard increment as a premium for contenders but a double-odds premium for non-contenders (which are usually those horses projected at odds above 9/1).

When I finish my real-time project, since the computer is already involved, a more sophisticated wagering strategy becomes possible, and more precise probability comparisons become possible since Internet pool totals can be monitored.

Formula, your exacta versus win study was an eye-opener, but your results are consistent with the way "smart" barns bet... exotics first, then whack the win pool at the last second.

[JustMissed]

Originally posted by Rick
JustMissed,

What I meant was that they lose about 10 cents on every dollar bet. Their win % is very high however and I'm sure there's some subset of them that's profitable. I just haven't found it yet. Morning line oddsmakers are usually very good at picking winners (better than 30%) but the public nearly always bets them down so that they lose 20 cents or more on the dollar overall. Actual favorites do a little better than that both ways so it's actually less efficient to eliminate them.

WHEW! Thanks, I thought one of us was losing it.

JustMissed

[Dave Schwartz]

Formula,

You said: "Dave, if you would

The simple one increase your profit from to___?
The deeper ones increase your profit from to ____?"

I can't answer this. This is not a database issue. This is a "real bet" issue. I speak anecdotally (sp?).

I can only say that FOR ME it was the difference between winning marginally (i.e. 4-6%) and staggeringly.

Best of luck to you.


Dave

[formula_2002]

Just a brief recap of Sundays picks.

Full analysis asap

Per defined odds structure
'A' picks: 11 plays, 45% wins, 226% profit
'B' picks: 4 plays, 0% wins, 0% profit.



All plays at all odds;

32 picks, 43% wins, 69% profit

Note: One of the more significant wins a due to scratches in the KEE 2nd. Changer became the pick and won at 14.50-1

[VetScratch]

Game Theory,

How can you justify advising value players to score their probabilities with a formula that rewards the smallest payouts?

Simon Magnus was a renowned magician, now known as the Father of Heresy who was exposed by Paul as a false apostle on the floor of the Roman Coliseum. You have alluded to qualities like "magic" when you discussed how your methods for probability scoring and odds blending have been perceived by others.

It occurs to me that you may be a servant of the Takeout Devil, and that you come among value players as a false apostle, like Simon Magnus, to spread heresy and cause financial distress.

If Spanish Inquisitors were to throw you in the Win Pool, would you float or sink?

BTW, when we talk about bad beats... sinking was the sign of innocence, while floating got you burned at the stake!

:) :) :) :)

[GameTheory]

VS --

This is your lamest piece of bait yet. Back to troll-school for you.

Definitely through with you...

[VetScratch]

Game,

How many smilies does it take to imply intent?

Why would I seriously bait you?

In any case, zaniness would not seem like a good choice of bait for anyone trolling for you. :) :) :) :)

[VetScratch]

Game Theory,That's all fine and good, but you're using the definition of entropy from the wrong field of science. In information theory, it is something different. In fact, John von Neumann suggested that Claude Shannon (father of information theory) call it "entropy" because no one really knew what it meant. (The spirit of the meaning is somewhat the same, but entropy in information theory is something that can be quantified precisely.)If fun can't be tolerated, so be it. Let's vow to ban fun from our debates!

In 1948, Bell Labs scientist Claude Shannon developed Information Theory, and the world of communications technology has never been the same.

Another of Shannon's talents was his ability to ride a unicycle. He kept one in his office at Bell Labs, and from time to time would pedal down the hallways, sometimes while juggling.

Information Theory regards information as only those symbols that are uncertain to the receiver.

For information to be useful, it has to be transmitted symbolically to someone or, in a computer's case, to something. At the destination, someone or something has to receive the symbols, and then decode them by matching them against his or her own body of information. Information Theory regards information as only those symbols that are uncertain to the receiver. The amount of information, or uncertainty, output by an information source is a measure of its entropy. In turn, a source's entropy determines the amount of bits per symbol required to encode the source's information.

You have borrowed one of Shannon's concepts for measuring uncertainty, which you express as
Ln(1/probability)/Ln(2) and have called it a measure of "surprise" for horseracing probabilities (http://www.paceadvantage.com/forum/...36443#post36443).

However, information theory examines uncertainty to find ways of expressing and communicating information such that efficiency has a minimal impact on certainty. Exactly what this has to do with handicapping puzzles me. Your assertion that "entropy in information theory is something that can be quantified precisely" does not seem to reflect the definition of entropy (i.e., inconsistency or chaotic tendencies) associated with horseracing.

IMHO, handicapping is better served by methods developed by actuaries. An insurance actuary maintains statistics and evaluates them for the purpose of forecasting liability probabilities, such as my chance of being diagnosed with cervical cancer and dying in 2004. Taken together, many such probability estimates are combined to calculate my current insurance premium payment.

In the long run, the insurance company depends on the accuracy of actuarial estimates to stay in business. In other words, if there is not a high degree of correlation between expected liabilities and actual liabilities, the actuary has done a poor job. Gross variances in degree of correlation will have negative results. If actual liabilities grossly exceed expectations, the company will be unable to pay claims. If expectations grossly exceed actual liabilities, the company may falter in a competitive marketplace because lower premiums will be offered by other companies with more astute actuaries.

In my opinion, actuarial methodology serves handicappers better than information theory methodologies.

Now, we can all choose our mentors. In my case, I would prefer a good actuary as a handicapping partner, as compared to a drinking companion, where I might well choose to party with a juggling unicyclist!

[Larry Hamilton]

The difference is that Game has a reputation of being helpful and easy to get along with since his first day. You, on the other hand, came in here flame-throwing from almost your first post. Fun my ass, go away.

[VetScratch]

Well, Larry, if you find nothing amusing about a debate between two horseplayers that manages to invoke the names of Claude Shannon (Father of Information Theory) and Simon Magnus (Father of Heresy), then I doubt the preposterous associations required by the game "Six Degrees of Kevin Bacon" would amuse you either.

[Fastracehorse]

Anyway, based on many years of experience with different methods, I still think that the easiest way to get genuine overlays with a good method is the throw out the ML favorite. Try it with whatever you're doing now. It's almost sure to improve the bottom line.

+++++++++++++++++++++++++

Thanx for the answer.

So, if you pick the M/L fave you won't bet it to win??

Also, do you like to make alot of win bets??

fffastt

[azibuck]

I don't know about the rest of you, but when I read a VetScratch post, I always imagine it being said by Diane Chambers, the character on Cheers. Pompous, pseudo-intellectual, makes sure everyone knows how smart she is.

[JustMissed]

Originally posted by JustMissed
MV, You are very literate and have a way of explaining concepts that is easy to understand.

You said earlier "Like others have stated in this thread, truly understanding the concept of value, and embracing it, has been a breakthrough for me".

Would you please explain:

A. What is value?

B. How is value used to select bets?

C. How did your embracing value cause a breakthrough for you?

Thanks,

Hey, maybe MV missed my post or quite possibly he chose not to respond. Hum, I wonder why.

Every time these dudes post some sort of psycho-statistics babble and you ask them to explain how that relates to winning races or making money they suddenly fall deaf and silent. Very interesting.

JustMissed
:confused:

[MV McKee]

In doing so, I will hopefully answer some of the questions posed earlier:

I had hoped the above statement, which I prefaced my follow-up posting with, would suffice in relating my answer(s) to your questions. I apologize if I was not concise enough.

A summarization of the answer I gave to all three of your questions is that understanding the concept of value has brought me to a point where I am finding value in selections other than my top selection. In other words, my bottom line is improved because I now understand that the potential value in a race may not necessarily be a product of the most likely winner. While my R.O.I. has deflated a bit, my overall profit has increased, due to the fact that I now play more races.
A more detailed version of the above paragraph is part of the posting I composed in response to your questions (as well as questions posed by others).

How I went from being very literate to posting psycho-statistics babble without uttering (or typing) a word is beyond me. May have been a by-product of the Helen Keller syndrome that apparently also affects us "dudes". I have not noticed the same trend you have in which people suddenly go deaf and silent after being asked to relate abstract concepts to their practical use. Then again, I generally try not to categorize something as babble simply because I do not understand it.

[VetScratch]

Is someone after my job?:D

[JustMissed]

Maybe it is just me but I might answer questions in this manner:

EXAMPLE:

A. What is value?

To me, value means receiving winning wager payoffs that equal or surpass my expectations.

B. How is value used to select bets?

I use a mechanical method to determine value where I prepare my own blended value odds line and compare the value odds of my contenders to the tote odds. If I would expect the horse or horses to win and the tote odds are greater than my value odds, I consider a bet.

C. How did your embracing value cause a breakthrough for you.

I may have been a little bit too exuberant in my statement. I did not mean to make it sound like I had discovered a cure for cancer or had embrassed Zen Budism. What I meant to say was that since I had added value consideration to my betting decision, my bottom line has increased substantially.

END OF EXAMPLE

You know I could be wrong but that is how I might answer a direct question.

You know, when I was in junior high a dude asked the teacher how he would know if he answered a question adequately. Old Mrs. Clark, our English teacher, said that "an answer is adequate if it is understood by the reader".

Just a thought,

[GameTheory]

Just Missed --

Not everyone is a lawyer...

[alysheba88]

Mark Cramer, Barry Meadow, Steve Fierro do a better job of describing value then I could ever hope to.

Is must reading for the serious player

[VetScratch]

Originally posted by Netcapperazibuck
I don't know about the rest of you, but when I read a VetScratch post, I always imagine it being said by Diane Chambers, the character on Cheers. Pompous, pseudo-intellectual, makes sure everyone knows how smart she is. Both Diane and Carla were great character roles and must have been enjoyable to project for both the writers and the actresses. I never felt Rebecca's character offered as much enjoyment to the writers, cast, or viewers.

I'm flattered that you would compare me to Diane, but don't overlook Carla... she's much closer to the personality I try to project when I am with horsemen or players at the track.

[Larry Hamilton]

What amazes (there's that word again) me is that someone could put words together so well would not demonstrate an understanding of their impact. Rather than trying to sort out the inuendo and double meaning phrases, I'll pass. You just made my ignore list along with amazin, lbj, canuck and derek.

[JustMissed]

alysheba88 mentioned three authors and how good they were at explaining "value".

I nicked this from the turpedia.com site:

Begin Quote:
UNDERLAYS AND OVERLAYS

If a handicapper thinks Horse A has a 25% chance to win, the fair-value odds would be 3-1 (calculated by dividing 100 by 25 and subtracting 1)

If the bettors offer odds of 2-1 and the handicapper bets, that's a mistake, because the handicapper has bet on an underlay. An underlay is simply an overbet horse

Horse A should win one of every four races (25%). If Horse A wins and pays only 2-1, the payoff will be $6.00. After four bets of this kind, the handicapper has bet $8.00 and collected $6.00, a $2.00 loss. The loss equals 25% of the $8 investment

If the bettors offer 4-1 on Horse A, however, that's an overlay.
An overlay is an underbet horse

Horse A still is expected to win one of four races. If Horse A wins and pays 4-1, the payoff will be $10.00. After four bets of this kind, the handicapper has bet $8.00 and collected $10.00, a $2.00 profit. The profit equals 25% of the $8 bet.

Any individual bet is part of a pattern that will be positive or negative as many bets accumulate. This is a game of percentages and probabilities, not certainties. Betting overlays and avoiding underlays is crucial. It's nothing less than the first axiom of pari-mutuel wagering.To estimate the fair odds on any horse, ask yourself how many times the horse should win in 100 attempts.

Divide that number into 100 and subtract 1.
That number is the fair-value odds to $1.00. If the actual odds are higher, the handicapper might bet. If the actual odds are lower, the handicapper should not bet.
End Quote

Am I missing something? Is there more to this "value" business than is discussed in the above turfpedia.com article.

I always thought betting overlays was Basic Handicapping 101.

Am I missing something here?

JustMissed
:confused:

[alysheba88]

There is more to it that that Just Missed. There becomes which horse do you bet. For example, if you make an odds line on four horses and two or three are overlays how do you bet? Fierro goes into this alot. People will have to read that.

Also, as Barry Meadow writes, how do you construct your bet. Should you bet a flat amount or not (the answer is no).

But what you have written is some of the basic stuff for sure. It is not, nor should it be a complicated subject. People understand the idea of getting a "good deal" on a house or car, but dont seem to when it comes to horses.

[Dave Schwartz]

JM,

Although I think this thread has run its course and I really don't want to get it started again, I think you have stated the obvious: Value is based upon betting horses that have a greater chance of winning than the public thinks they do.

The Catch-22 is producing a good set of probabilities to make your line from in the first place.

When I first responded to this thread it was because my experience in talking with "conventional" value players has been that the majority of them failed quite nicely. The problem is that the odds lines they are making just do not hold up to the litmus test of reality.

My own experience with value took a strong turn towards improvement when I began "smoothing" ("blending" seems to be the popular word in this thread) my probabilites toward the public's opinion (tote board). A handful of others came forward and said that doing something similar worked for them.

Probably the most astounding of them all is Formula 2002 - he has recently posted picks and seems to be applying his own version of the A-B-C grades and his A's are just kicking butt. (As I recall it is only a few months ago he was telling us why the game could not be beaten.)

Other people in this thread essentially told us why it cannot work.

I am astounded at how many people I speak with that tell me they do not win (long term) at the races yet steadfastly defend what they are doing as the "right" way.

In my humble opinion, success at the window is a result of synthesizing other people's ideas with/into your own and making the whole thing work. If it works, use it. If it doesn't fine... go find something that does.

As that old saying from the '50s says, "Don't knock it 'til you've tried it." <G> If your outcome needs changing, who knows? Maybe this will change it.


Just my opinion.

Dave Schwartz

[Show Me the Wire]

Do you mean buying a horse or wagering on a horse? Value is what something is worth. If I am buying a car, a house or a horse I can get a pretty good idea what the property I am buying is worth. I cannot relate the idea of value, as understood in the purchase sense, to wagering on a horse.

The difference, which leads to the dissimilar application of value, is investing in a tangible versus an intangible. The car, house or horse purchase results in the ownership of a tangible item. Wagering on a race results in the outcome of a probability based on unknown factors influenced by randomness. Defining value in this manner is much more difficult and open to varied interpretations.

Return on the wager is determined upon probability, but not true probabilities. Generally, each horse competing is not physically equal to each other meaning each horse does not have a natural probability to win. For example in an 8-horse field each contestant does not have 1/8 of a chance to win. This is generally true in the majority of races on a daily basis.

Basically, the probabilities are determined by a pool of uninformed, educated guesses and informed wagerers. The uninformed wager on names, numbers, public selector choices and the like, the educated guesser wagers on information from past events, and the informed wager on known information, such as a horse went lame in-between starts or a horse has recovered from an infection or possibly a horse is being administered a prohibited performance enhancing substances. Further the real probabilities are dependent on the randomness or chaos of the race. The most informed wager is a loser if the horse stumbles badly, loses a jockey, falls, etc. The method of setting odds is manufactured and not based on natural or set probability i.e. coin flips.

The question is how do you define value based on manufactured probabilities. In other words what is the worth of manufactured probabilities?

I believe value is created when the uninformed and educated guesser set the probabilities, instead of the educated guesser and the informed wagerer, because uninformed money is more likely to manufacture an inferior set of probabilities.

Regards,
Show Me the Wire

Perception is reality

[Dave Schwartz]

SMTW,

I found one question in there that I could address...

You said: "The question is how do you define value based on manufactured probabilities. In other words what is the worth of manufactured probabilities?"

... but, alas, with another question:

All probabilities are "manufactured," aren't they?


Dave

[Show Me the Wire]

Dave:

I guess you can take that stand, I can manufacture a three-sided coin and manufacture 3 to 1 as opposed to 2 to 1. I wanted to say most games of chance based on probabilities equate to natural odds, once again my reference to coin flips. A two-sided coin has natural odds of 2 to 1, while a three-sided equal coin has natural odds of 3 to 1. Generally, if I create a race with any specific number of horses the natural odds would not equate to the number of sides manufactured.

To answer your question according to my definition, which may be incorrect, is no.


Regards,
Show Me the Wire

Perception is reality

[Dave Schwartz]

Okay, so what kind of probabilities are not manufactured?

Yours? Someone else's? Output of a pick service?

Dave

[GameTheory]

Even scientists and statisticians can't agree on what probability is. I take a Bayesian view of probability, which is pretty much in-sync with what normal people usually mean when they talk about it, although they might not realize it.

That is, probability is a measurement -- a statement of belief. An event is either going to happen or it isn't. There are no probabilities without expectation. If no one is betting (in the general sense) on something to happen or not happen, it doesn't make sense to speak of probabilities.

Probability is a conditional measure of belief (or uncertainty) that a certain event will occur, given the information we have and the assumptions we are making. You can't leave out the second part, which is why you can't talk about "absolute" probabilities because the conditional part is always associated with the person (or computer, or whatever) assigning the probability. So probabilities are things you determine, not things you discover.

[Show Me the Wire]

Originally posted by Dave Schwartz
Okay, so what kind of probabilities are not manufactured?

Yours? Someone else's? Output of a pick service?

Dave

I think we are invoved in semantics and I probably should have not used the word manufactured. I attempted to clarify my statement by relating value through comparing probability based on games with natural odds versus games lacking natural odds. Also my use of the word random is wrong as randomness applies to games of fixed probabilities. The better word is choas, which is more akin to a 40 mph gust of wind happening during one toss coins in a serious of coin tosses.

In summary horse racing is not probability based on natural odds or what I call fixed probabilites. I define fixed as being limited by the number of sides on dice, coins, cards in a deck, spaces on a wheel, etc.

Here I have to lean on GT as he said probability is based upon expectations. I am saying in horse racing the expectations are developed by combining different levels of knowledge versus knowing with certainity an event will happen 1/6 of the total time.

That is why value in wagering is difficult to define as wagering is intangible. The value of an intangible is subjective to the user.

Hope this clarifies my opinion.

Regards,
Show Me the Wire

Perception is reality

[Dave Schwartz]

SMTW,

Thank you for clarifying.

I submit that many people are applying value (though not all). It all starts with what I call either "direct" (i.e. "I make this horse 2/1 and he is 4/1 therefore he is a good bet.") or "indirect" ("My top horse has proven to be profitable at 4/1 or higher.") value. A "spot play" angle is another way of finding "indirect value."

All I attempted to do in this thread was display a way that has made the direct value approach work more effectively FOR ME. I contend that since it worked for me it might work for a few other people.

I have no need to trumpet it as the end-all or Holy Grail for everyone. I can tell you that it is turning into that FOR ME, but how could I possibly know if it will have that result for someone else?

If anyone is beating the game now (especially if they are beating it solidly) I would encourage them to make any changes carefully, if at all.

On the other hand, if things aren't going so well, then perhaps this might be the one thing that pushes them on to success. It was FOR ME.


Dave

[Fastracehorse]

<It all starts with what I call either "direct" (i.e. "I make this horse 2/1 and he is 4/1 therefore he is a good bet.")


You must have a very decent win % with your methods. Maybe you would do well without making a line at all.



fffastt

[Dave Schwartz]

ffast,

How would you infer that from what I said?

Dave

[Fastracehorse]

It's just that if it were me, I don't think I could win on your direct method approach.

I just bet horses I feel have a good probability of winning and are good odds.

So, in order for your direct method to work I inferred you must have a very good selection process - about 30 % winners.

fffastt

[Dave Schwartz]

You inferred incorrectly. I have a good contender process and an excellent value analysis process.


Dave

[JustMissed]

Dave, Thanks for putting this in your post, I think it will give some of us an accurate perpective of where you are coming from.

[YOU WROTE]:All I attempted to do in this thread was display a way that has made the direct value approach work more effectively FOR ME. I contend that since it worked for me it might work for a few other people.

I have no need to trumpet it as the end-all or Holy Grail for everyone. I can tell you that it is turning into that FOR ME, but how could I possibly know if it will have that result for someone else?[END QUOTE]

I think all would agree that our horseplaying is a two-fold activity:

Contender selection
Bet selection

1. Contender selection usually is completed by labeling our contenders as follows:

a. A,B,C,D
b. 1,2,3,4
c. 4-5, 6-5,2-1,6-1,for example
d. Tick marks, pluses or any other ranking symbol

2. Bet selection is so varied and complex I could not even begin to list them other than to offer these two methods:

a. Value plays(comparing your odds line to tote odds or some variation thereof).
b. Any other method

For me to infer that the A,B,C,D method for contender selection is superior to the odds line method and has "CHANGED MY LIFE" would be foolish and, at best, arrogant. The same would be true for others who have described their use of "value plays" as superior to other betting methods and as such have used adjuctives to infer that it was some sort of spritual gift bestowed to them by God, and only to them, because of their special understanding of psycho-statistics.

Anyway, as far as your posting, I know your input as been very helpful to those who do make their own line and has helped peak the interest for those who might benefit from an investigation of the subject.

JustMissed
:)

[Fastracehorse]

I don't think I inferred incorrectly - you said exactly what I said.

I said I couldn't win with a value analysis process per se.

Your friend,

fffastt

[Rick]

There are some methods that are very good at estimating the probability of the top ranked horse but not so good at estimating the probabilities of lower ranked horses. If that's the case then you'll only find true overlays on the top ranked horse. Of course, "blending" may help with the others but it's not a sure thing. And making adjustments so that all horses have the best fit to their actual probabilities may cause you to wind up with a line that's not accurate enough for any of them. I'm not saying that this is always the case, but it is a common occurence and might explain why some are so adamantly opposed to the idea of calculating odds and looking for overlays. They probably tried it in the past and it didn't work with their method. That could be true even though there may be other good ways for them to identify overlays. My motto is "whatever works".

[Rick]

Just for the record, I don't think that "contender selection" is necessary and don't understand why so many people handicap this way. I have only one "contender" in each race and look for an overlay on that one horse only. In the past, with less accurate methods, I've found overlays on the 2nd ranked horse too but never any lower ranked than that. The better my accuracy gets on the top ranked horse, the worse it gets on the 2nd and lower ranked horses.

[Fastracehorse]

<The better my accuracy gets on the top ranked horse, the worse it gets on the 2nd and lower ranked horses.

My 2nd place horse gets me at least as much as my 1st ranked horse gets me there. I need a little honing - lol.

I just find it difficult to understand how people can make money with a value method. I mean I understand the logic but it wouldn't work for me.

===========================================

<I have only one "contender" in each race and look for an overlay on that one horse only.

This is what I do too. I have methods that lead to alot of solid over-lays but I realize how difficult it is to win consistently. And a value-line process would only bury me. Just because a horse is an over-lay doesn't mean it is a good bet.

===================================

Rick,

If guys are making money with win betting at shorter prices, say 5/2 - 9/2, they must have a very strong win %.

Do you know any guys like that??

fffastt

[VetScratch]

Originally posted by Rick
Just for the record, I don't think that "contender selection" is necessary and don't understand why so many people handicap this way. I have only one "contender" in each race and look for an overlay on that one horse only. In the past, with less accurate methods, I've found overlays on the 2nd ranked horse too but never any lower ranked than that. The better my accuracy gets on the top ranked horse, the worse it gets on the 2nd and lower ranked horses. I think your approach recognizes the distinction between handicapping and value play tactics. Overlay opportunities exist throughout the odds spectrum, but overlay wagers that work for individual "playing styles" do not work.

For value plays, you choose to drop "smart bombs" while other prefer "carpet bombing." For commercial handicapping products, I have noticed that "recommended value play tactics" tend to appeal to "carpet bombers" because the "typical customer" finds more "satisfaction" from "plenty of plays that produce marginal results" than "less frequent plays that may produce better results."

Many players "really" don't want a system that restricts overlay plays to the top one or two contenders. However, these are also my "key horse" plays, especially in exactas and other exotics.

Racing must be a frustrating industry for many product vendors...

If a product vendor can show even a marginal profit with "high velocity wagering tactics" that provide plenty of plays, this will often be the approach that is recommended or advertized. Since marketing priorities rule the roost, a product vendor may NOT be encouraged to "explicitly" recommend or advertize their "best empirical wagering tactics." Apparently, even the mention of "passing some/more/many/most races" is perceived as a negative "sound bite" in the marketplace. Of course, marketing priorities are generally empowered because "perception" is what attracts new customers.

[VetScratch]

Correction... I deleted the end of the first paragraph while trying to eliminate an extra paragraph break after "work."
I think your approach recognizes the distinction between handicapping and value play tactics. Overlay opportunities exist throughout the odds spectrum, but overlay wagers that work for individual "playing styles" do not workto eveyone's satisfaction.

[Fastracehorse]

<For value plays, you choose to drop "smart bombs" while other prefer "carpet bombing."

My value is in good probability over-lays. I don't look for value for the sake of value.

=============================

I was asking Rick what he thought was a successful strategy for some winning horse players he might know.

Especially what their avg. win pay-off and win % might be.

fffastt

[VetScratch]

Good for you Fasty, but I was replying to Rick's post... NOT yours.
Example:
My quoted salutation was Rick's post. I did NOT quote your post.

[alysheba88]

I dont know it seems some are making this more complicated then it needs to be.

The whole process is very simple to me.


You select your contenders (in my case three or four horses)

You make a line based on YOUR opinions.

You compare it to the "line" made by the betting public (the actual betting lines).

You bet accordingly.

I look for at least a 50% overlay, although 100% is really what pays the bills in the end.

If your highest rated contender is an overlay you bet. If not, you look at your second choice, and so on.

Its a very easy. almost mechanical process. All the hard work is done the night before. When its go time there is very little thinking to do.

You will lose money if you consistently bet underlays. There is no other way around it. Its not about picking winners.

[Rick]

fast,

I don't regularly associate with any other horseplayers these days, except here of course. Since I don't go to race books or the track much I don't meet anyone interested in the subject anymore. In the past I've known some who had a very high win % and ROI but they were playing spot plays, maybe 2-5 per day at all tracks. I doubt if I could make money at 5-2 odds but 3-1 or 7-2 isn't difficult to show a profit at. Of course, most of the profits come from those at 4-1 and above and I think that would be true for almost anyone's method.


VS,

Actually, my 2nd choices do pretty well overall, but for some reason the overlays come in races where I already have one on the 1st choice. I guess I could play two horses in those races, but that's really not my style. In races where the 1st choice is not an overlay it still loses less than the average at very short odds. That situation makes it much less likely to find an overlay on other horses. So, the only way I can play every race is to bet exactas or angles in those other races. I'm sure there's some subset of those short odds horses that are overlays but I haven't found a way to identify them yet.

[Rick]

So, I guess it would be more accurate to say that I've never had more than two contenders in any race. My 3rd choices and below have always had a horrible ROI. I suppose it's possible that a 3rd choice at really high odds might be an overlay but it probably wouldn't happen more than once in a hundred races.

Actually, it makes sense when you do the math. The top two choices with a good method win a little better than 50% of the time. If those are break even (playing every race), then the others would have to lose 40% in order for there to be a 20% loss overall betting every horse. In my opinion, that's about as good as I should expect to do at separating winners from losers. That's why I'm astounded that people would claim to be showing a profit betting two horses in every race with better than 60% wins. My calculator starts smoking every time I try to figure out a way that could be possible.

[GameTheory]

You have to consider that if betting multiple horses the best thing to do will be to bet different amounts on them, and often the extra horse(s) will actually be taking a slight loss overall but smooths out losing streaks and allows you to increase churn for more overall profit...

[alysheba88]

I don't vary my bet size by contender. But can understand those that do. I bet my second choice if he is the play the same as I would my first contender. However, I will vary the amount of the bet by the % of edge. Ie; I bet more when I have a 100% edge then 50%. Not talking dramatic differences, but different.

[GameTheory]

I mean you would vary it by edge, exactly.

[Rick]

Here's an example of why the math doesn't work out:

Lets use 64% winners on the top two since that's the most conservative figure mentioned in the Sartin books.

64% winners at 25% profit is possible only if the other 36% lose 100%!

.64 * .25 - .36 * 1.00 = -0.20

Since they usually claimed more than 64% wins (up to 80% actually) and up to 50% profit they must have been betting into some pools that returned more money than people bet. Sorry but it just don't add up.

I realize that most of you wouldn't make claims like that, but there is that one guy who said it in print.

[VetScratch]

Rick,
64% winners at 25% profit is possible only if the other 36% lose 100%!

.64 * .25 - .36 * 1.00 = -0.20If you show a 25% profit while hitting 64% winners, I think the best expression for this would be:

1.25 = (.64 * 2.515) - (.36 * 1.00)

which demonstrates that you need a 151.5% profit on your 64% winners to achieve a net profit of 25%.

[Rick]

VS,

I wasn't talking about the individual bettor but rather the overall pool, which should lose about 20% after breakage at most tracks. The amount won by the 64% winners has to be made up by the other 36% losers. I just thought I'd throw this out there and see if anyone can find an error in the logic.

[VetScratch]

Rick,

Oops, I thought you were describing the typical Sartin devotee who routinely hits at least 64% winners and shows at least a 25% profit. :D

[Rick]

VS,

Yeah, I was talking about the Sartin types but just demonstrating that there aren't enough losers to go around to support their claims. After all, you can't lose more than 100% of your money. One guy that we all know wrote in a book that most of us have read that he got 70% winners and 50% profit. I don't think it's possible because of what I've said, but I'm willing to listen to anyone who can show that my logic is wrong.

[Dave Schwartz]

Can you explain to me why it would be IMPOSSIBLE to get 70% winners and 50% profit?

(Understand I am making any claims, I just do not understand your logic.)

Dave

[Tom]

Originally posted by Rick
VS,

Yeah, I was talking about the Sartin types but just demonstrating that there aren't enough losers to go around to support their claims. After all, you can't lose more than 100% of your money.

I know a guy who regularly lost all his money then lost half of mine every time we went to the track. :mad:

[Rick]

Dave,

On further review, maybe I can't. I was doing the calculation with 70% of the money bet rather than 70% of the winners. I still think the conclusion is true but it would require more than that to prove it. Sorry to confuse you with the fuzzy math.

[Dave Schwartz]

Rick,

No problem.

Dave

[Red Knave]

Originally posted by Rick
I still think the conclusion is true but it would require more than that to prove it.

Rick -
I don't think the conclusion is true.
The percentage of winning races has no direct relationship to the amount of money in the pool for those races.
Here's how I do the math (maybe it's fuzzy too!)
64% winners, 25% profit
64 wins out of 100 bets. The average Sartin bettor bets on 2 horses per race.
So 100 bets at $2 per horse at 2 horses per race = $400 bet.
25% profit means 25% of $400 or $100.
Therefore $400 bet + $100 profit = $500 total return from the wagers.
$500 returned from 64 winning bets means that the average winner paid $500 / 64 = $7.81 each (approx).
I think that's all the math we need. We don't care about how much the losers in these races put through the machines.

[VetScratch]

Red Knave,

Maybe Rick's challenge makes sense as follows: If a selection system wins 64% (or even more) races playing 2 horses, it can only produce a profit until too many players adopt it and drive the average payoff beneath the threshhold needed to make a profit. Then, if we assert that players will flock to a winning system, the catch is that any winning system will be undermined by popularity.

[Red Knave]

Maybe Rick's challenge makes sense as follows: If a selection system wins 64% (or more) races playing 2 horses, it can only produce a profit until too many players adopt it and drive the average payoff beneath the threshhold needed to make a profit. Then, if we assert that players will flock to winning systems, the catch is that any winning system will be undermined by popularity.

Not only does that make sense, that seems to be the way of it.

[VetScratch]

Yup, most all of us have given winners to friends who changed their minds in the mutuel lines and bet another horse. When this happens, the friend should be praised for making the right moral choice by not playing in a way that could have diluted the payoff. :D

[Rick]

Well, the problem is really much more complicated than I originally thought. It depends on the distribution of odds on the winners and how possible it is to select higher odds winners and still maintain a high win %. 64% winners on the top two would probably work out to about 40% on the 1st choice and 24% on the 2nd. How does one do that without picking a large percentage of favorites? And if you do, then your saying that the rest of the favorites need to lose an enormous amount to make up for your +50% ROI. I may not be able to prove that it's impossible, but if you look at the actual numbers I'll bet that it looks so ridiculous as to be virtually impossible. We can see what the average odds are on the 64% of winners with the longest odds and see what the maximum theoretical ROI could be but nobody seriously believes that you could pick 64% winners without using a large majority of lower odds horses.

My point is that there has to be some reasonable limit on what can be achieved and these claims always go way beyond what anyone has ever done in a controlled test. I was hoping to find a claim that could be proven impossible in the strictest sense so that people would stop believing these exaggerations. If it turns out that it's also necessary to use some common sense to reach the same conclusion then I guess those who want to ignore it will still have something to hope for and those who make the claims can continue to sell the dreams.

Anyway, I don't really intend to spend a lot of time thinking about it but would be interested if anyone else has the information and desire to find out what the limits really are.

[Red Knave]

It depends on the distribution of odds on the winners and how possible it is to select higher odds winners and still maintain a high win %.
Yes, I think this is true.
64% winners on the top two would probably work out to about 40% on the 1st choice and 24% on the 2nd. How does one do that without picking a large percentage of favorites?
What you say has merit. In order to win this high a percentage of races many favorites must be in the mix.
But favorites lose 2/3 of the time and second choices even more. Is it not possible to pick 3rd, 4th or 5th favorites that win? And it's not necessary to bet every race, just the ones where it looks like a higher priced selection might win.
It doesn't take too many $10 or higher winners to raise the average payoff.
And if you do, then your saying that the rest of the favorites need to lose an enormous amount to make up for your +50% ROI.
No, I'm not saying that at all. In fact, I don't care how many favorites lose. A 50% ROI just means that, in the long run, for every $10 I bet I get back $15.
I may not be able to prove that it's impossible, but if you look at the actual numbers I'll bet that it looks so ridiculous as to be virtually impossible. We can see what the average odds are on the 64% of winners with the longest odds and see what the maximum theoretical ROI could be but nobody seriously believes that you could pick 64% winners without using a large majority of lower odds horses.
This would be a good test.
My point is that there has to be some reasonable limit on what can be achieved and these claims always go way beyond what anyone has ever done in a controlled test. I was hoping to find a claim that could be proven impossible in the strictest sense so that people would stop believing these exaggerations. If it turns out that it's also necessary to use some common sense to reach the same conclusion then I guess those who want to ignore it will still have something to hope for and those who make the claims can continue to sell the dreams.
This would also be a good test. I think any results would be biased by the person doing the test.

[cato]

Pizzolla, former Sartinista, has said many times that while he was winning a big % of his win bets--60-70%, betting 2 horses a race, he made very little money on those bets. BECAUSE, the Sartin plan at that time was to bet the top two horses and pretty much ignore odds.

I am clueless as to why people accept the Sartin numbers as a given. For the chosen few who may have hit 50%, 60% or 70% of bets (betting 2 horses a race)
:cool: ---and I'm betting that was a small small small % of the Sartin folks---it had to be a lotta chalk and, in turn small, or no, profits.

Lets say you hit 60% betting 2 horses:cool: and average a mutual of $6.00 (2-1). You are still losing money.

Cato

[Rick]

I don't doubt that you can get a high percentage of winners playing two horses in very few races, but not in all or almost all races as was claimed by some. The most irritating claim was one of 70% winners and 50% profit along with the comment that anyone who thought it couldn't be done was just a loser.

To those who still believe this drivel, just try sometime to bet two horses in every race. You'll probably get a little better than 50% wins and if you have a very good selection method you may break even. You won't be able to get 60% wins without losing money and you won't be able to make any money without lowering your standards for win %. That's the way EVERY method that I've ever seen has worked, including pace oriented methods. In fact, if anything, pace methods usually have a LOWER win % than speed methods. Their appeal was the higher average odds that existed 15-20 years ago. Today, all public handicappers include pace as a factor and there win percentages have not risen much if at all. Don't you find it unusual that NO public handicapper has been able to master the Sartin techniques well enough to win at those higher rates when all of their "secrets" are now public knowledge?

[Rick]

For those of you familiar with physics, there's a phenomenon I like to call the "Heisenberg pace handicapping principle". In a similar manner as the famous uncertainty principle, pace handicapping methods perform differently when they're actually observed by someone. If you're not watching they have amazing success, but as soon as you look they somehow lose their effectiveness.

[Red Knave]

This thread has turned interesting for me.
Rick and cato pose some interesting questions and it caused me to start thinking about some things that I had previously taken for granted.
Following are some numbers I generated for Aqueduct, Belmont and Saratoga for this year using races at distances of 5f to 10f. The numbers come from a database at http://www.tracstar.com/test/dbaseform3.cfm . I am unfamiliar with these people but I am taking the data as reliable. The results are from 1682 - 1697 races depending on which favorite you ask for. I think this is due to field size and ties. At any rate a decent size recent sample.

Here is how the top 4 public choices fare in all these races.
Fav#---% won------------------av price
1 35% (610 of 1697) $ 4.88
2 20% (337 of 1696) $ 8.10
3 15% (260 of 1693) $11.10
4 10% (165 of 1682) $15.06
So, the top 4 public choices win 80% of the time (35+20+15+10). These percentages are rounded down by a bit. The actual total came to 80.99%. The top 3 choices win at over 70% and I expect that a handicpper with some method to seperate them could maybe dope out 64%.

Now, suppose that our 2 horse bettor only bet on horses in the top 4 choices and that his wins were distributed among them in the same percentage as they occurred at the track. So out of 64 wins in 100 races
35/80 * 64 or 28 wins were favorites
20/80 * 64 or 16 wins were 2nd favorites
15/80 * 64 or 12 wins were 3rd favorites
10/80 * 64 or 8 wins were 4th favorites.
Total winners 64 (28+16+12+8). I am open to other suggestions on how the 64 winners were arrived at but I know they can't all be favorites since we are betting 2 horses per race and whatever selection method we choose can't always pick the 1st and second choices.

If we do the math to see what the return was we get this
28 * 4.88 = 136.64
16 * 8.10 = 129.60
12 * 11.10 = 133.20
8 * 15.06 = 120.48
Total return 519.92 after betting 400.00. So profit is 119.92 or 119.92/400.00 = 30%
So maybe it is possible.

[Rick]

Red Knave,

Excellent analysis. That's the kind of thing I was looking for. NY tracks have a considerably better win % for top choices than most of the others, but still it's interesting.

I can't look into it much right now though, since I'm moving this week in order to get an "overlay" on my cost of living.

[formula_2002]

RED, for 4300 RACES AT BEL,AQU AND SAR

ODDS RANK, win %, AVERAGE WIN PRICE

1 30% $4.28
2 19% $7.34
3 14% $9.94
4 11% $12.42

[cato]

Red Knave: Your analysis was that the top 2 choices win 55% of the time (which could be a tad high according to formula); but going on, let's just assume that we will go 4 deep, instead of two deep and will increase our hit rate to 64% by betting on higher odds horses...and therefore we can make a profit.

That assumption is the equivalent of the cartoon where the two scientsists are pouring over a lenthy mathematical equation and they get to the final result with the use of "and then a miracle occurs"

Your primary assumpion is faulty and therefore your analysis and argument is faulty...but I like the discussion

Cheers, Cato

[pmd62ndst]

Red,

If I'm not mistaken, aren't the average payoffs you mention for $2 win bets?

If that's the case, then wagering on those 400 races will require $800 and not $400.

$519.92 - $800 = - $280.08

I knew it couldn't have been that easy.

PMD

[hdcper]

Red Knave,

I am about ready to call it a night, but quickly I believe there is something wrong with looking at this as you have suggested.

I agree that the top 4 picks win about 80% of the races, but to win those 80 races out of a 100 we would be required to make 400 wagers (4 plays per race times 100).

However, using the 2 bet per race criteria, we are only making 200 wagers and yet we are making the assumption that we will win 64 of those 100 races. I guess it is possible, but still hard to believe day in and day out.

What do you think, am I making sense or not?

Bill

[Fastracehorse]

It is too difficult a game for me to base my wagers on an odds line - that is what I'm saying.

For me, just because a horse is longer odds than he should be, even if he is my top pick, does mean the horse warrants a bet.

What I am saying is that the limiting reagent is the probability based on my selection method.

Many top picks are fickle.

I prefer the spot play method.

fffastt

[Fastracehorse]

<If you're not watching they have amazing success, but as soon as you look they somehow lose their effectiveness.


I'm not sure I know what you are getting at here.

fffastt

[formula_2002]

Originally posted by pmd62ndst
Red,

If I'm not mistaken, aren't the average payoffs you mention for $2 win bets?

If that's the case, then wagering on those 400 races will require $800 and not $400.

$519.92 - $800 = - $280.08

I knew it couldn't have been that easy.

PMD

I gather he is randomly selecting 2 horses from the top four odds ranked horses. If the horses were jelley beans in a jar, the win % for the two selected horses should reflect the win %'s within the sample.

"if it looks too good to be true, it is."
In any small sample, just about anything can happen.

[Red Knave]

Originally posted by formula_2002
I gather he is randomly selecting 2 horses from the top four odds ranked horses. If the horses were jelley beans in a jar, the win % for the two selected horses should reflect the win %'s within the sample.


I am not selecting ANY horses, randomly or otherwise.
The original premise was that it is impossible to have a 64% win percentage and show a 25% profit. The point was raised that if your win percent was that high then you obviously had many favorites. I just wanted to know whether or not it WAS possible. Maybe for a good enough handicapper it is possible to pick 2 of the top 4 choices and win 64%.
I know that I'm not doing it. <g>

And thanks for the update on the results for the top 4. I should have asked you first, although I did state where and how I got those numbers.

[Red Knave]

Originally posted by cato
Red Knave: Your analysis was that the top 2 choices win 55% of the time [...], let's just assume that we will go 4 deep, instead of two deep and will increase our hit rate to 64% by betting on higher odds horses...and therefore we can make a profit.

That assumption is the equivalent of the cartoon where the two scientsists are pouring over a lenthy mathematical equation and they get to the final result with the use of "and then a miracle occurs"

Your primary assumpion is faulty and therefore your analysis and argument is faulty...but I like the discussion

What I'm saying is that a good handicapper might be able to do this. I know that I never have and apparently neither have you.
You state that the primary assumption is faulty yet you offer no proof that it is. You just believe that it is.

I believed the opposite and set about to prove it. I may not have done that but I know I did not disprove it either.

[hdcper]

I agree Red Knave that it is possible, but continuing to think about this keeps making me doubt the probability of it happening.

Lets again look at your excellent analysis and see how difficult this actually is. As you suggested, to meet a profitability of around 30%, the so called handicapper would have to hit 28 favorites, 16 second favorites, 12 third choices and 8 going off fourth in post time odds. Thus, lets assume further that this requires playing 50 horses in each category results in the following win percentages by category:

28/50 = 56% winners

16/50 = 32% winners

12/50 = 24% winners

8/50 = 16% winners

Now if we look at the other 50 horses in each category, the following win percentages based on the remaining expected winners by category we see this:

7/50 = 14%

4/50 = 8%

3/50 = 6%

2/50 = 4%

Extremely below the respective win rates based on the public's long time success rate.


Possible yes, likely no!!!

So is anyone up to the showing us the proof of your great handicapping magic? Prior to the running of the races that is.

Bill(hdcper)

[Red Knave]

Originally posted by hdcper
Extremely below the respective win rates based on the public's long time success rate.
hdcper
I think the premise rests on the ability of the handicapper to get 64% winners in 2 horse bets. I have read many articles by handicappers claimimg these numbers over the years (Pizzola, Hambleton, Bradshaw) so I believe the high win % is possible. All I wanted to do was see if you could get a positive ROI if you DID win at that rate. There were a few posts in this thread that suggested it was not possible but I convinced myself otherwise.
Possible yes, likely no!!!
Not likely if you bet and win at the same rate as the public.
In order to have a positive ROI you must do what the public does not. Either exploit situations where you have a legitimate edge in the odds or win at a higher rate than the public (or both <g>).

[VetScratch]

Red Knave,
I think the premise rests on the ability of the handicapper to get 64% winners in 2 horse bets. I have read many articles by handicappers claimimg these numbers over the years (Pizzola, Hambleton, Bradshaw) so I believe the high win % is possible. All I wanted to do was see if you could get a positive ROI if you DID win at that rate.After accepting 64% winners as attainable, those who posted database figures should easily be able to figure out the worst case ROI for the races in the sample.

IOW, what would be the return if the lowest 64% of the payouts were collected betting 2 horses per race? One has to assume that a very high win percentage is probably biased towards betting favorites, so worst case ROI for 64% wins might be revealing.

Along similar lines, for the same database, what is the breakeven percentage of wins betting 2 horses if we assume the shortest prices are always in the subset of winners? This might give chalk players a rough estimate of how well they would have to perform to reach breakeven betting 2 horses.

[formula_2002]

I think this formula would help this note;


edge =(odds x win %)- (1-win%)

to attain a 10% edge (profit) while winning 64% of the races, the min odds will be

.10=odds x .64 -(1-.64)
.10 = .64 odds -.36
.64 odds= .46-1

odds= .72-1


Joe M

[VetScratch]

Formula,
edge =(odds x win %)- (1-win%)Yes, this gives us the average odds required for any profit/loss target. Now, based on the actual databases used in this discussion, how likely is it that a high percentage winner (e.g., 64%) could approach breakeven?

0 = (odds * .64) - (1 -.64)
odds = .36 / .64 = .5625/1
Then, for 2 horses bet per race, 1.125/1 average odds are required to break even.

Now, for the 64% actual winners with the lowest final odds in the actual database, what were the actual average odds?

The variance from 1.125/1 would be interesting since very high percentage selection methods "tend" to identify relatively low odds starters.

No scientific conclusion can be expected, just an interesting exercise for players trying to reach the high win percentages published by some system authors.

[Red Knave]

0 = (odds * .64) - (1 -.64)
odds = .36 / .64 = .5625/1
Then, for 2 horses bet per race, 1.125/1 average odds are required to break even.
Vet Scratch -
For 2 horse betting, this is not correct. It should be
0 =(odds * .64) - (2 - .64) = 2.125 i.e. $6.25 to break even.
You can prove this by dividing the total bet in 100 races by 64, $400 / 64 = $6.25

The variance from 1.125/1 would be interesting since very high percentage selection methods "tend" to identify relatively low odds starters.
Formula-2002 posted previously that the average favorite in his NY sample paid 4.28 and the 2nd favorite pays $7.34. Not enough on the favorites to break even. If we ONLY use the top 2 favorites and they win at the same rate as in the sample (i.e. 30% and 19%) then the favorite would win
(30 / (30 + 19)) * 64 = 39 times and the 2nd favorite would win (64 - 39) = 25 times. Doing the math to get total return would give us
39 * 4.28 = 166.92
25 * 7.34 = 183.50
Total return $350.42, a $49.52 loss or (12.38%).
Still not break even.
This is quite interesting. For this win percent, it looks impossible to make a profit betting the top 2 public choices. This, of course, makes sense.
So, in order to get a positive ROI, any time your own top 2 choices are also the top 2 public choices (on average), I guess you have to pass the race. I must admit that this looks pretty difficult.
No scientific conclusion can be expected ...
And here we are trying to be scientific! <g>

[VetScratch]

Red Knave,

For 64% wins betting 2 horses, odds of 1.125/1 are correct!

For every $100 wagered at odds of 1.125/1 you collect $136 on 64% wins and lose $36 on the 36% losses (i.e., they still give back your wager when you win at the tracks I play.) :)

Formula's formula calculates odds, not payouts.

[Red Knave]

Originally posted by VetScratch
Red Knave,

For 64% wins betting 2 horses, odds of 1.125/1 are correct!

For every $100 wagered at odds of 1.125/1 you collect $136 on 64% wins and lose $36 on the 36% losses (i.e., they still give back your wager when you win at the tracks I play.) :)
Vet -
Do the math again.
2 horses per race for 100 races at $2 each is $400. In order to break even on 64 winners you need $6.25 average payoff. At my track this comes out to odds of 2.125/1 since they do, in fact, return the original wager <g>.
You're forgetting that $100 wagered means only 25 races bet, at $4 per race, of which we win 16 (64% of 25). So,
16 * $4.25 (i.e. 1.125 * 2 + $2.00) = $68 returned.
On the other hand,
16 * $6.25 (i.e. 2.125 * 2 + $2.00) = $100 returned. Break even.
The difference is due to 2 wagers per race.


Formula's formula calculates odds, not payouts.
I was referring to another post from page 13, as follows.
Originally posted by formula_2002
RED, for 4300 RACES AT BEL,AQU AND SAR

ODDS RANK, win %, AVERAGE WIN PRICE

1 30% $4.28
2 19% $7.34
3 14% $9.94
4 11% $12.42

[VetScratch]

Red Knave,

Oops, you're right, doubling breakeven odds of .5625/1 doesn't work because betting 2 horses sacrifices one wager, so you do need 2.125/1 odds to bet 2 and win 64% at breakeven. :eek:

For every $100 bet you need to collect $168 on your 64% wins because you will lose $68 on your 36% losses.

[VetScratch]

Red Knave,

Looking at Formula.2002's results in your last post, the average payoff for winners that went off as the two post-time favorites was only $5.47. Since they win about 50% of all races, getting an average payout of $6.25 (at 2.125/1) while winning 64% of the time does not "seem" like an inviting challenge.

I have never "intuitively" liked the idea of flat-betting 2 horses because the effective rate of takeout becomes about 64% at NYRA tracks (50% + 14%). Psychologically, surrendering at least half my bets takes most of the zing out of winning (with a single bet). If there is a grind-it-out niche that shows a profit by always betting multiple horses in every race (perhaps with dutching), I can't imagine sticking with it unless it was a hands-off totally automated procedure. Without completely delegating both handicapping and wagering to a computer, at some threshhold of boredom, I would be inclined to give up playing the horses to look for a more engaging pastime... like going back to work on a farm or at a racetrack.

[formula_2002]

Here is something you may find interesting. Scroll down to the bottom of page and click on "Public's Estimation Od The Odds"

You may need excel to read the report.


http://globalwinningpicks.homestead.com/GLOBALWINNINGPICKSX.html

[Speed Figure]

What program do you need to open the file?

[formula_2002]

should be a spread sheet like excel.

I can e-mail a "picture" of the file. If you like you can send an e-mail request.

Joe M

[Red Knave]

Originally posted by formula_2002
Here is something you may find interesting. Scroll down to the bottom of page and click on "Public's Estimation Od The Odds"

That is interesting.
I think I'm reading it correctly that the 1st column is odds rather than a ranking (like 1st fav, 2nd fav). It is instead, 0 to almost 1/1, 1/1 to just under 2/1, 2/1 to just under 3/1 etc. Is this how you read it?

[formula_2002]

YOU GOT IT.
EXAMPLE;
ROUNDED ODDS OF 3 REPRESENTS ODDS OF >=2.5 AND.<=3.4

[Red Knave]

Originally posted by VetScratch
I have never "intuitively" liked the idea of flat-betting 2 horses because the effective rate of takeout becomes about 64% at NYRA tracks (50% + 14%).
I have never had a philosophical problem betting multiple horses per race. The proviso is that I need a high win % to go along with it.
But, the results we have all shared in this thread has made me rethink how I will go about it in the future.

[VetScratch]

Formula_2002,

Very interesting XLS file.

What was the source database? Any screening, or just the actual mix of all days, all races, all weather conditions, etc.?

Have you considered tabulating finer odds-range subdivisions between 0.00 and 5.00 (like parimutuel toteboards and the original Quirin example)?

If you add a column for the odds average in each range, we could compute expected winners and compare them to actual winners. For example, without the actual average, we have to assume even distribution of odds between 2.50 and 3.49 for range 3.00, which may not be true, especially in the lowest ranges if you add more ranges.

I really imagine you have already done expected versus actual comparisons for your own purposes, and have probably also added statistical confidence tests... so what you publish is, of course, up to you. :)

[formula_2002]

Vet, as far as various screens go.. Al jolson said it best, "I got a million of 'em".

The data base is a 3 year accumulation of bris/all-ways data.

Public's expected winners;

1-(actual winners/expected winners )= % profit

[VetScratch]

Formula_2002,
Originally posted by formula_2002
Vet, as far as various screens go.. Al jolson said it best, "I got a million of 'em".
The data base is a 3 year accumulation of bris/all-ways data.
Public's expected winners;
1-(actual winners/expected winners )= % profit I can't find my copy of Al Jolsen's math primer for handicappers. :)

Please explain "1-(actual winners/expected winners) = %profit"

Since "(actual winners/expected winners)" is equal to Impact Value (IV), I don't see how you get "%profit".

Everything else about your fine research presentation makes perfect sense, but I don't get what this formula means.

Let's look at an example:ODDSRN WNET PNET SNET #HORSES #WINS %PROFIT %WINS
5 -1,382.05 -1,441.15 -1,301.40 7,806 1,088 -0.18 0.14
Everything about your 5/1 ODDSRN group (4.50 to 5.49 odds) makes sense. On win bets, the public lost 18%, which is roughly equal to the national average for takeout+breakage (17%+1%).

In a broad sense, this loss (-18%) means horses in this category won just about as expected.

Also, from data in the row, we can calculate that the 1,088 winners went off at average odds of 4.904/1 (slightly below 5/1).
((7806 - 1382.05) / 1088) - 1 = 4.904 (or roughly 4.90/1).

What we can't determine are the actual average odds of the 6,718 losers.

However, if we assume 5.00/1 was the average actual odds for all horses (winners and losers), the expected number of winners would be:
expected winners = #HORSES * ((1 - takeout) / (odds + 1.05))
expected winners = 7806 * ((1 - .17)/(5.00 + 1.05))
expected winners = 1071

Now, looking back at your formula as it applies to the 5/1 odds group:
1-(actual winners/expected winners) = %profit
1-(1,088/expected winners) = (-.18)
expected winners = 922

I can't see how this formula is meaningful, since 922 could not be right for expected winners, given the other figures.

What your results suggest to me is that the public lost 18% because the number of actual winners slightly exceeded the estimate for expected winners while the actual average odds on winners was 4.90/1 (slightly below 5/1).

Since everything else makes sense, what am I missing about the significance of this formula? It seems that what it really calculates is the variance above or below a par impact value (1.00).

[formula_2002]

Vet, the analysis is done by rounded odds, therefore the winners and losers have the almost the same average odds. In this sample all horse ave odds was $4.93-1 while the winning horses was $4.92.

Joe M

[VetScratch]

Formula_2002,

I don't see from your own data how you can refute 4.904/1 for the winners:
((7806 - 1382.05) / 1088) - 1 = 4.904 (or roughly 4.90/1).

Nevertheless, if we accept 4.92 for winners and use 4.93 for all horses, let us still look at your formula

1-(actual winners/expected winners )=%profit
1(1088/expected winners)=%profit
1(1088/expected winners)=(-.18)
expected winners = 922

Since your data gives actual winners and %profit, I assumed you were showing how expected winners could be calculated.

However, if odds were 4.93 for all horses, expected winners would be:
expected winners = #HORSES * ((1 - takeout) / (odds + 1.05))
expected winners = 7806 * ((1 - .17)/(4.93 + 1.05))
expected winners = 1083

Your formula produces 922 for expected winners.

What did you mean us to calculate with your formula? I can't see how it is anything else but a variation on impact value calculations because it produces the variance above or below a neutral impact value (1.00). If actual and expected winners are known, it does not produce %profit.

Given actual winners and %profit, it certainly does not allow us to calculate expected winners.

[formula_2002]

Vet, the formula is intended to compare expected winners in a oddsrange to the actual SYSTEM winners in that same odds range. If the system has any value a/e should be greater then 1.

The most meaningful figure in each odds range is the z-score.

paceadvantage.com/forum/showthread.php?s=&threadid=7667&highlight=significance

Joe M

[VetScratch]

In other words, the formula has no significance for the data in the actual XLS file that you encouraged us to download and look at. :)

[formula_2002]

if thats what you get out of this

[VetScratch]

For your XLS file, expected winners could be determined only after you posted average odds (4.92/1), and SYSTEM winners were not included in the data (or in the formula). How did you expect that throwing us an unexplained formula would add to our understanding of the significance of results in your XLS file?

Neither the XLS data or the formula included SYSTEM winners, system value, or z-scores.

I quit... going out for a significant breath of fresh air before my collection of Nobel Prize Bobbleheads starts talking to me. :)