how many overlays on a 9 trace card

[formula_2002]

To anyone that is making at least 10% in the win pool.

Just how many overlays do you find on a 9 race card??

OK, if you are not making 10%, how many overlays do you think are on a card?

Caution, the information you provide could help prove the impossible.. :)

[cj]

Here is the answer:

True overlays: I have no idea

I generally find 5 or 6 horses on a 9 race card I think are overlays. The thing is, I could be dead wrong on all but one and still make a nice profit if I am right on the other one.

This is what I don't understand about those who say you have to statiscally prove your betting line, that 2-1 shots should win 33% of the time, 3-1 shots should win 25% of the time, 4-1 shots should win 20% of the time, etc.

I will always stand by the fact that you don't have to prove this for your line to be a good tool. The real test is this - do your OVERLAID horses win often enough at high enough odds to show a profit? When you are wrong, it doesn't matter by how much you are wrong. The key is to have your line point you towards horses the public will underbet.

[GR1@HTR]

I'd say 2 to 4 per race...About 27ish per card...Now that doesnt mean they will all win. Contending overlays win about 12% on average...But if you have 3 in a race that is about a 36% win percentage you will have in the race...Most will find this post ludicruis, but it works for me...

[Jeff P]

I generally find about three horses on a nine race overlaid enough to justify (IMHO) making a win bet.

How many of those are TRUE overlays? All of them of course!!! Or is it just the ones that win?

I'll admit to being dead wrong somewhere between 83 and 85 percent of the time. But when I am right, those that do win pay more than enough to make the whole excercise wonderfully worthwile. :cool:

[JustMissed]

Hey, why did you guys reply to Joe's question. Didn't you get the memo?

Based on Joe's thread "trolls of the board take heart", over in the Off Topic section, we are not suppose to reply to Joe anymore in order to help him curb his evil 'trolling' ways.

Hope it helps, I think Joe is a nice guy when he is medicated.

JM
:D

[Steve 'StatMan']

The number of true overlays on a race card? No one could know for certain it depends on the contestants in each race and how the betting goes. What may be an overlay with 5 minutes to post could end up being an underlay when the race starts, or vice-versa.

What one can do, however, is make their best reasonable estimate of the true odds, based on the factors they are able or willing to identify. Bets are made on those going off at the higher odds. Allow for enough room with the odds so that you have room for error in your judgements but still have a margin for profit.

Example: Betting horses with a 33% chance to win (2/1 fair odds) at actual odds of 3/1.

If you are exactly right in your estimates (and you always got paid $8.00 for those 3/1 shots), the profit margin is:

(.33 * 3) - .67 = .99 - .67 = .32 - a 32% thoretical profit.

If you don't actually win 33%, but do win, say, 28%, then the results are:

(.28 * 3) - .72 = .84 - .72 = .12 - Still a 12% profit. For having an error in your judgement of (.33 - .28) / .33 = (.05/.33) = .15555... a 15.5% error in one's judgement.

You could still break even by actually being right only 25% of the time

(.25 * 3) - .75 = .75 - .75 = 0

and could still be (.33 - .25) / .33 = .242424.... about 24% off in one's judgement.

Only being worse than 24.24% off in one's judgement would result in a loss, under these circumstances.

The hard part - being hard-working enough, patient enough, and actually finding enough worthy factors and situations and wait for those odds. Personally I'm still working on achieving the right personal/work balance to succeed on this part.

Total rigidity isn't necessarily required for actual success. In fact, the complexity of racing makes rigid testing by paper or computer impossible to prove to the 100% level, and since the odds can differ for similar situations later on, there are guarantees of future success.

But the therory does work. Like my example in another thread (Off Topic - Trolls Take Heart) you have to find the situations, the subsets, the 'Antiques Among The Junk'. If you're correct enough in your 'judgements on value', it can be done.

For those who have never read it, Barry Meadow's "Money Secrets At The Racetrack" is an absolute must read. It certainly inspired me, and many of us here (Thanks Barry!)

[Overlay]

I think that the distinction that GR1@HTR makes about contending overlays is useful. Because the public tends to overbet low-odds horses and underbet high-odds horses, I usually run into a fairly high frequency of overlays in my betting line, but many are only slight overlays on horses to which I have assigned a very low true probability of winning. When measured as potential wagers in terms of the edge provided by the disparity between my calculated true odds and their tote-board odds, they don't offer enough of an advantage to warrant any play as a win bet. I find that I perform best in win betting when I focus on overlaid horses with true odds which I calculate to be lower than their random chance of winning based on the field size of the race (for example, horses with true odds lower than 7-1 in an eight-horse field).

[toetoe]

Thank you so much, Statster, for impeccable math, the sine qua non for any shot @ actually making money. Of course I checked it this way:
3/1 horse pays $8.00; 3/1 horse "wins" 33% of the time.
8.00 --------x-------- .33 = $2.64 or
64 cents profit per $2.00 or
32 cents profit per dollar.

[Valuist]

Are we talking strictly to win? There may only be a few legit overlay win candidates per card but there's a decent number who are overlay candidates to hit the underneath slot(s) in exactas, tris or supers.

[Steve 'StatMan']

Are we talking strictly to win? There may only be a few legit overlay win candidates per card but there's a decent number who are overlay candidates to hit the underneath slot(s) in exactas, tris or supers.

Thanks toetoe!

I agree Valuist! Plenty of ways, pools and areas. I'm just trying to show the basic 'Win' premesis - we've been having some long discussions on whether or not the races are even beatable. I'm trying to make the case that, finding methods to distinguish individual bets and horses at odds, success is possible, vs. Joe/Formula's long running preminse that the races are not beatable because betting all 3/1 shots is a losing wagering proposition (Naturally! If one doesn't make any distinctions between them!)

[thoroughbred]

Thank you so much, Statster, for impeccable math, the sine qua non for any shot @ actually making money. Of course I checked it this way:
3/1 horse pays $8.00; 3/1 horse "wins" 33% of the time.
8.00 --------x-------- .33 = $2.64 or
64 cents profit per $2.00 or
32 cents profit per dollar.

Doesn't a 3/1 horse win only 25% of the time?

[formula_2002]

Thank you all for your replys.
Yes I was talking about win pool overlays.

The hard numbers I got from the replys were ;
5-6 per crad

2 to 4 per race ( I have not done the math, but I dont think you can have 2-4 10% profit overlays in 1 race) so I'll go with 1 per race and per race .

On my own, I would say 1 per race card.

So I'm going to assume 2.5 overlays per race card.

say 8 horses x 9 races =72 horses

2.5/72 = 3.5% overlays per race card

Very rough figure.

In another thread, I post that in the odds range >=2.5 to <=3.4 (average odds of 3-1)
you would have to play 270 3-1 horses, win 27.5% of the time at a 15% take-out track in oder to establish a significant system to pick 3-1 horses for a 10% profit

3-1 horses represent 8.85% of al the horses that run.

say only 3.5% of those horses are overlays.

Therefore;

270/.035 x .0885=81716

Which means I would have to handicap 81716 horses to find 270 3-1 "system" picks.

81716/( 8 horses per race x 9 raced per card)=
1134 track card days, or perhaps 3 years worth of racing.

Anyone agree with the math?

[Steve 'StatMan']

Doesn't a 3/1 horse win only 25% of the time?

On average, horses end up at 3/1 win a bit less than 25%, because of the track take, and as probably all odds groups, a flat bet loss.

The goal for win betting is to find horses whose chances to win are better that their win odds.

So betting horses you feel should be 8/5 that are really going off at 5/2 or better, or maybe your '5/1 shots' are going off at 8/1. Those are potentials for value and profit - assuming one's judgement is good enough. (Big assumption, goes toward handicapping skill).

The goal from the first example is to find a horse you think that should be 2/1 (wins 33% of time) that has final tote odds of 3/1.

[formula_2002]

Actually 3-1 horses win % tend to correlate to the track take-out.

In a 7200 race sample in tracks with a take of about 15%, they won 21% of the time for a loss of 16%.

[toetoe]

Tbred,
No, we always forget to add 1 to the denominator when we estimate probability. 25% = 1/4. Bet one to gain 4, incl. orig. bet, which isn't yours anymore, anyway (like playing with "the track's money"). We call that 3 to 1.
Always add one to every ratio and you'll know the correct relationship between prices. Ex.: 3 to 1 is not 50% better than 2 to 1. It is 33% better:(3 + 1)/(2 + 1) = 4/3 = 1.33.

[Steve 'StatMan']

I'm not questioning the math, just wondering why one's method has to pick 270 horses at 3/1, although it is just an example, of course.

If you're using 'a method' to select the horses, they could be at a wide variety of odds.

Unless were talking about only betting 3/1 shots. But unless there is a reason to prefer some 3/1 shots over others, the long run future for the bets is dim.

[thoroughbred]

Tbred,
No, we always forget to add 1 to the denominator when we estimate probability. 25% = 1/4. Bet one to gain 4, incl. orig. bet, which isn't yours anymore, anyway (like playing with "the track's money"). We call that 3 to 1.
Always add one to every ratio and you'll know the correct relationship between prices. Ex.: 3 to 1 is not 50% better than 2 to 1. It is 33% better:(3 + 1)/(2 + 1) = 4/3 = 1.33.

All I was saying was, (and let's forget the track takeout). if the odds on anything are 3/1, the win rate is 25%. In other words it wins one out of four times. The 1 in the denominator you speak of doesn't apply to this simple odds question.

[Jeff P]

Anyone agree with the math?

Joe, your math looks okay but I tend to thimk your basic premise of only looking at horses within a very narrow odds range close to 3-1 will make the game harder to beat not easier.

I'll say this: I think it's far easier to beat the game by handicapping for and win betting the types of horses that win as overlays in the odds range of between 6-1 and 15-1 than horses going to post at 3-1.

[Brian Flewwelling]

When i was much younger i was told that 1 plus 1 is 2.


Was that true? and is i still true?


Fleww
:p

[Dave Schwartz]

I am joining this discussion late but I would liek to give an answer.

The answer would be about 78% of the races contain one or more overlays.

Now, whether or not they are "playable" (or even visible) from the handicapper's unique perspective is another matter.


Regards,
Dave Schwartz

[InsideThePylons-MW]

Doesn't anybody here ever HATE a favorite?

Don't really bet much just to win anymore and don't have time to form a rock solid odds line. Isn't the right idea to form a line based on your personal opinion and not what you think the horse will go off at?

If you hate a 8/5 favorite and you give him an actual 20-1 chance to win in a 8 horse field, probably all 7 of the other horses in the race are going to be an overlay..........or am I missing something? :rolleyes:

It takes a strong opinion to exploit value and extract the most money possible from a race.

Obviously dealing with mostly the unopinionated in this thread.

[GR1@HTR]

IMHO, for each underlay in a race, it creates one or more overlays. A heavy pounded favorite that might only have a 40% chance of winning that goes off at 3/5 creates 3 to 4 strong condending overlays.

Im not a poker player, but I was explaining to some fellow co workers the other day about horse racing: That each race is like a hand of poker...you either fold when you don't have an advantage or you play through it knowing you might have an some favorable possibilities (odds, inside information...) on your side.

[formula_2002]

JP,
Thanks for the math review.

[hurrikane]

this, i think, is the issue with Joe's thinking.

He is assuming all 3-1 shots are created equal. They are not.
And that premise is the starting point for a winning horse player.

If we can get joe to that place maybe he will turn.

And maybe tomorrow when I wake up the snow will be gone and it will 80 degrees outside. Right.

[formula_2002]

this, i think, is the issue with Joe's thinking.

He is assuming all 3-1 shots are created equal. They are not.
And that premise is the starting point for a winning horse player.

If we can get joe to that place maybe he will turn.

And maybe tomorrow when I wake up the snow will be gone and it will 80 degrees outside. Right.

H.. you are correct about me thinking that all 3-1 shots are created equal (within a certain statistical range).
But I also think there are a groupe of people that don't.

I'm simply attempting to determine what is necessay to prove them RIGHT.

[hurrikane]

that's easy Joe,

everyone has been tell you..

it's winning bets.

[alysheba88]

If I make lines on all 9 races, typically will find an overlay on 7-8 of them. I just make lines on 4 contenders usually.

[Dave Schwartz]

Alysheba,

Your approach is much like mine.

And I see it exactly as you do.

GR1 said about the same thing and it is very logical.

I think of it this way: The only race without overlays is the race where we agree so closely with the public's position that all the horses suffer from the disadvantage of the takeout. Very few races per day fall into that category.

Now, some races only provide overlays among the non-contenders and are typically not betable (by me). At least not in the win pool.


Regards,
Dave Schwartz

[alysheba88]

Absolutely Dave.

The one thing I am fine tuning is the idea of not all overlays being equal. For me, that means my track record on the race in question. Ie; my best race type is maiden special weights. I also do better on routes compared to sprints. Do better on certain odds categories than others and so on. Right now compiling a fairly large sample first before I start making rush judgments. Using the Bettor Keep Track program. Wish I had been using this for years.

[GR1@HTR]

A88,

In better keep track, i set up and tested senarios of the number of overlays I had per race...
Working off memory here...but it went something like this.

1 overlay= -.02 ROI
2 overlays = 1.08 ROI
3 overlays = 1.20 ROI
4 overlays = 1.50 ROI (and thats betting all 4 horses to win)

Not to say that the above will work for everyone...

[JackS]

Many authors have reasonably noted that many horses going off at odds in the longshot range (20-1? and above), are underlays.
Just a thought but wouldn't it be practical to identify these horses that have very little chance of winning and are taking action (although seemingly minimal) that should be going off at 2 or 3 times posted odds?
Causes for this phenom could be factors such as leading jock or trainer stats or even the past glory that the horse one possessed but no longer deserves due to age, injury ect.
Identify the underlays and you may have overlays on the horses that are the true contention in this race.

[Steve 'StatMan']

Note: The more experienced bettors, cynics, and the time concious, may not want to read this post. At least I've given everyone fair warning ;)
-----------------------------------------------------------------------
I did some more thinking on the sample size from Joe's examples earlier (I didn't want to repeat the post another time).

I wanted to note that, on 1 circuit that races 5 days a week, 52 weeks a year, and a generous 10 races per card, there are only 5 x 52 x 10 = 2600 races per year.

I wasn't sure in Joe's example if he wants to pick 270 3/1 shots from 11,000 races, or was looking from a group of 87,000+ 3/1 shots.

In any event, that's a lot of races. If one is testing an angle, to find enough 3/1 shots with a specific angle, it may take 5, 10, 20, 40 years worth of data for the one circuit, if ever. As for national data, we may only be interested in how an angle does in NY, or Chicago, since we may only be betting races from those tracks, against others who bet those races, in the same pools. What happens to 3/1 shots with a similar angle, or for that matter all 3/1 shots, at Santa Anita, Fonner, or Portland Meadows really shouldn't matter in our case.

For most angles, you might not find 270 occurances of an angle, total, let alone 270 that went off at exactly 3/1 odds. Oh, you'll find more than 270 overall 3/1 shots. But we may never find 270 bets with a specific angle at 3/1 (or all oods) in 5, 20, 40 years worth of data that applies to our circuit. Heck, by then, much of the game and the information will have changed, and the participants who were a part of that data (the horses, trainers, jockeys, and the bettors who set the odds) will be retired if not dead and burried.

If we wait 20-40 years, or even just 5 years to prove something before placing a bet, we're probably never going to make a bet. There's other things in life.

Successfully betting horses has got to be about taking reasonable risks at a price (they do call it Gambling, whether we're careful with our bets, or not). Yes, we don't want to take too many lumps learning the game and testing methods, and pay too high a tuition price for the education. But if a person is too risk averse for wagering without 'the ultimate statistical proof', hey, maybe horse race wagering isn't right for them - there's no shame in that.

Yes, people need to be careful not to jump in too deep into the pool before they learn how to swim - but if you're going to learn to swim someday, then you've gotta get in the water at some point before you proven you can swim. Start in the shallow end, get a teacher, and practice.

Once a person regularly feels they are finding enough good bets compared to the odds, and they do the right things, and keep results and records, they can pretty much tell if they are successful at it yet or not. Some things just can't be total proven in advance by stats.

I've have had some good runs before, so I know I'm close. Many others who've posted here are already there successful. Will any of us be successful after 20-40 years? With the confidence we've gained over time, let's try it/keep doing it, and see what happens. It's a exciting experience, if nothing else! :)

[Dave Schwartz]

Steve,

Excellent post!

Your points are very well made, but you haven't proved anything. <G>

Seriously, what I hear you saying is that "I can't mathematically prove anything (with a high degree of confidence), but I can say with some degree of certainty that I have an approach that is "worthy of a shot."

You know, I think this is where most so-called database handicappers misuse their database. The shear statistical sample size necessary to prove anything negates whatever advantage the databaser has if they insist on statistical certainty.

BTW, this is what has Joe all turned around as well.

Think of it this way... Suppose I am convinced that 1000 plays are necessary for the confidence level I desire. Let us also assume that I have a BIG database (which I do). And I mean BIG.

It all comes do to how I segment that data. How will I do that? Obviously a 4f race for 2yr old fillies should not be grouped with a graded stakes at 1 1/4 miles on the turf. That is easy to see. But what about the not so obvious?

Let us consider the permutations or categories.

Surface
Dirt is certainly far different than turf. We'll have 2 categories there.
2 categories


Distance
For now, let's settle upon 5 categories:
mini - dash - sprint - route - marathon
2 x 5 = 10 categories

Race Type
This is a tough one. On the simple side, you have maiden and non-maiden.Are MSWs different than maiden claimers? Certainly.

Are claimers different from Alw? Yes. And claiming races, non-winners of 2 lifetime are also different.

Are NW allowance races different from classified allowance? Absolutely. In fact, the are actually more like claiming races than NWs.

Handicaps and stakes can go together, but graded races should probably be on their own.

I think we can lump optional claimers in with claimers pretty safely.

Starter races need to go on there own.

So, we have: (MSW)-(Md)-(Clm/Alw/OClm)-(ClmN2l)-NWAlw-(Hcp/Stks)-GR-ST

Okay, that gives us 8 categories. (And, yes, I know we could argue, uh, discuss this topic all day.)

2 x 5 x 8 = 80 categories


Age
Age is certainly a factor. Let's keep it real simple here - Young and old. (A young horse is a 2yr old or a 3yr old before May.) 2 categories.

2 x 5 x 8 x 2= 160 categories


Track Condition
This is a big one. But let's just handicap fast tracks for now.

2 x 5 x 8 x 2 x 1= 160 categories


Sex
Well, there are really only the two of those, aren't there? And males DO run differently than females.

2 x 5 x 8 x 2 x 1 x 2= 320 categories


Field Size
This one is very important. Want to know the win percentage of favorites? Do you suppose favorites only win 33% of the time in 5-horse fields?

First, we will ignore races with less than 5 horses. Then we will break down the others thusly:
(5-6)-(7-8)-(9+)

2 x 5 x 8 x 2 x 1 x 2 x 3= 960 categories


Class of Race
Is a $50k claimer different than a $5k claimer? You bet! Let's try this:

Cheap: <$5k
Low: $5k-$10k
Medium: $11k-$29k
Allowance Grade: $30k-$69k
Stakes Grade: $70k+

2 x 5 x 8 x 2 x 1 x 2 x 3 x 5= 4,800 categories


Pace Mix of Race
This could be expressed as high, medium or low pace pressure. Or it could be a count of "E" horses, or a range of Quirin ES points. Let's assume that we can come up with a scheme that addresses this factor in 3 categories.

2 x 5 x 8 x 2 x 1 x 2 x 3 x 5 x 3= 14,400 categories


Class of Track
Does a $15k claimer run differently at Belmont from Wyoming? Sure it does.

Let's assume 5 "tiers" of tracks:
WorldClass - (i.e. SoCal/New York, CD, WO, Fla in the winter, etc.)
2nd Tier - NoCal, Chicago, Louisiana, MD, MTH, PHA, etc.
3rd Tier - TuP, CalFairs, Mnr, CT, etc.
Barrell Bottom - Small tracks that run long meets.
Tiny Tracks - Those that run just a few days per year.

We'll not even bother with the 5th tier.

2 x 5 x 8 x 2 x 1 x 2 x 3 x 5 x 3 x 4= 57,600 categories

So, there you have it - Over 50,000 categories!

And forget being track specific!

Now, admittedly, some of these permutations (categories) cannot exist - (When was the last time you saw a 2f turf race for older males?) As a good example, in our software's RaceReports we designate 144,000 "race configurations," of which about 20,000 actually exist.

to be continued...

[Valuist]

Macao at Aqu today definitely was an example of an overlay. 5-1 vs that weak field?

[Dave Schwartz]

So, what is a databaser to do?

A typical circuit has about 2500 races per year. (Recall that we only used fast/firm tracks in our database but let's don't quibble about that.)

So, if a guy is downloading 10 circuits, he will wind up with about 25,000 races per year.

And if he downloads for 3 years to build a database he will have around 75,000 races.

So, let's divided 75,000 races by only 14,000 categories. We get an average of 5.36 races per category! Hardly worthy of even being called a "sample."

In our giant database (375,000 races) and 20,000 categories we average 18.75 races per category, but the largest category has over 3,500 races. That would be, 6f, claiming, older males, fast tracks, 8 horses, etc. (Note that we segment just a little differently.)



The Point
(You knew there'd be one, right?)

Actually there are two.

First, the database guy says, "Well, I'm going to have to break the database into larger pieces - less categories."

So, he wants a miniumum of 1,000 races per category so, in a 75,000 race database he can have 75 categories, right? Wrong! Some of those categories will still have 3-race samples! He may wind up with 20 "statistically significant" (according to his definition) categories and the other 55 categories may well not have enough sample size to study.

Fine. So he only plays (a guess here) 1/2 the races.


Second, by lumping "less-than-like" races together (He reduced the permutations but not slicing and dicing so fine, remember?) his predictive accuracy MUST suffer.

Think of it like this - As a category is studied, he is, essentially, building a approach for the average race in the sample. Consider that some races within the segment are closer to "average" than others.

We might, for example, say that there are 3 "sub-segments" within the data: Average, Close-to-Average, and Far-from-Average. (You could just as easily break it down with 4 or 12 or 87 sub-segments if you wish, but this seemed logical to me.)

So, when he handicaps an "average" race, he is getting the best performance possible from his sample: It was meant for this kind of race.

But what about when he handicaps the Close-to-Average group? He is not exactly handicapping apples with oranges. It is more like tangerines with oranges. But no matter how you juice them (pun) the accuracy of the predictions should be weaker with this group than with the Average group.

Finally, the Far-from-Average group is like handicapping avocados with oranges and we would expect the performance to suffer.


One More Thing
It isn't just that the other two groups will perform below the "average" group. Consider also that those other two groups are being included in the analysis of the sample; effectively, the avocados are being used to determine the juice of the oranges!

In other words, the handicapping process of the Average group is being mislead (for lack of a better word) by the Far-from-Average and Close-to-Average groups!


Next: Solutions...

[alysheba88]

All very true Dave. I think sometimes people just need to think these things out more. There are so many variables and different race situations. I think people go to two extremes. The one where judgements are made based on a very very small race sample (or even one day) and the other side where they insist there has to be thousands of interations of something for it to be worthwhile. Not even understanding what "it" is that that needs the iterations, since as you point out there are so many variables that affect them.

There is a definite and real danger in micromanaging things. Prefer looking at general trends.

[Dave Schwartz]

BTW, this is why stats guys like Joe will typically fail miserably at horse racing. What is most important to them is proof and proof demands sample size even at the expense of accuracy!

Compare Joe's position with StatMan Steve's - I'll wait for his comments but I imagine that he starts with the premise that "likeness of sample" must come first.


I am going to take a break here, but there is a solution coming.

[formula_2002]

come first.


I am going to take a break here, but there is a solution coming.

Dave, I feel your pain..Whew!!

[alysheba88]

Mark Cramer handles the whole sample size issue very well in his books.

[formula_2002]

Mark Cramer handles the whole sample size issue very well in his books.
Tell us a bit..I have never read any of his stuff..

thanks

[Hosshead]

Just how many overlays do you find on a 9 race card??



Caution, the information you provide could help prove the impossible.. :)
The correct answer is: exactly 9 !

[JustMissed]

To anyone that is making at least 10% in the win pool.

Just how many overlays do you find on a 9 race card??

OK, if you are not making 10%, how many overlays do you think are on a card?

Caution, the information you provide could help prove the impossible.. :)

You cannot accurately identify a item that is the comparison of a subjective item with a quantitive item.

An underlay is mathematically impossible to quantify and therefore cannot be verified.

Examples:

Quantitive: She got the most votes in the beauty contest. She won the beauty contest.

Subjective: She is the most beautiful girl in the contest. She should win the beauty contest.

Quantitive: The horse went off at 4-1. The horse won and paid $10.

Subjective: That horse is better than 4-1. That horse SHOULD go off at 5-2.

JM

[formula_2002]

[QUOTE=JustMissed]You cannot accurately identify a item that is the comparison of a subjective item with a quantitive item.

An underlay is mathematically impossible to quantify and therefore cannot be verified.

/QUOTE]


Then how do you determine when the reward is worth the risk ?

[JustMissed]

[QUOTE=JustMissed]You cannot accurately identify a item that is the comparison of a subjective item with a quantitive item.

An underlay is mathematically impossible to quantify and therefore cannot be verified.

/QUOTE]


Then how do you determine when the reward is worth the risk ?

Joe, just compare ticket cost with expected payoff. Would you play a $24 trifecta combination if you expected a $47 payoff-of course not. Would you play a $48 trifecta combination that you expected to collect $390-that's your answer.

I have nothing against booksellers or software vendors using "overlays" to sell their product but the average horseplayer should be able to identify this concept as not valid for betting decisions.

Hey, if making an oddsline helps a guy with his selections, more power to him. But to use that oddsline to make betting decisions is simply nuts.

People like to talk about horseplaying as an investment or a business but then seem to ignore the concept of sells price less cost or return on investment less amount invested. Have you ever heard of a merchant basing his sells price on what he thinks he should pay for a jar of pickles-heck no, he bases everything on his acutal cost compared to expected sells prices.

You don't have to believe me. You can easily prove it yourself.

Take any commerical software that produces an oddsline. If that program uses the same criteria to make an oddsline for a $2500 maiden claiming race as it does for the Kentucky Derby, then you know that oddsline is as worthless as tits on a boar.

JM

[thoroughbred]

You cannot accurately identify a item that is the comparison of a subjective item with a quantitive item.

An underlay is mathematically impossible to quantify and therefore cannot be verified.

Examples:

Quantitive: She got the most votes in the beauty contest. She won the beauty contest.

Subjective: She is the most beautiful girl in the contest. She should win the beauty contest.

Quantitive: The horse went off at 4-1. The horse won and paid $10.

Subjective: That horse is better than 4-1. That horse SHOULD go off at 5-2.

JM

I assume you are making this point to, in some way, try to show that determining an underlay, (or overlay) is impossible because of the comparison of the subjective to the quantitative.

But, if so, you are forgetting a key point. It is possible to calculate, quantitatively, what the odds for each horse should be. In CompuTrak, for example, an oddsline is calculated based on quantitative factors from race performance. Based on the quantitative values that determined such an oddsline, a comparison with the final actual odds will, quantitatively show, whether a horse is going off as an underlay or overlay.

[Dave Schwartz]

JM,

>>>Hey, if making an oddsline helps a guy with his selections, more power to him. But to use that oddsline to make betting decisions is simply nuts.<<<


I am not sure if you meant precisely what you said or not. If you did, from experience I must vehemntly disagree.

Now, before launching into why, I must absolutely agree with the following statement that you made:

>>>Take any commerical software that produces an oddsline. If that program uses the same criteria to make an oddsline for a $2500 maiden claiming race as it does for the Kentucky Derby, then you know that oddsline is as worthless as tits on a boar. <<<

Not only would I agree with that statement, I will add:

"Any odds line that is made without considering the public's opinion as part of the line-making process is likely flawed."


I can only address the issue of "risk-reward analysis" from the standpoint of how we address them.

First, we believe playing a race is a two-step process:


Handicapping (i.e. assigning probabilities)
Exploitation of the race


In step 1 you assign the probabilities to each animal and in step 2 you decide how to make money from the race.

Technically, we do not "make an odds line." We assign probabilities. While that may appear to be the same thing, in our application it is not.

You see, in the first step we have created a probabilty for the race. If the second step is merely deciding if the horse is a good bet then it is done by $net and optimum bet. The current tote odds determine those things.

In other words, we don't usually say, "I will bet this horse if he is over 3/1." We are more likely to say, "I will bet this horse if he has a $net of over $2.40 and/or an optimum bet of 2.3%."

What makes our approach so much stringer is that the public's opinion also modifies the probabilities of the horses dynamically!

I have used this comparison before: You have 2 horses that your software has made 2/1 - (We prefer 33%, but that's okay.) One of these horses is going off at 1/1 and the other at 20/1. Those two anaimals do not have the same chance of winning! The 1/1 horse is probably significantly underestimated and the 20/1 horse is probably way overestimated.

Now, the 20/1 horse is still the better wager but when you bring his probabilities into line, he drops from being tied as your best probability horse to being (perhaps) your 3rd or 4th best horse. Still profitable, and probably playable, but to be played differently.



Regards,
Dave Schwartz

[Dave Schwartz]

JM,

BTW, you are right that you cannot prove that your oddsline is correct. (We are starting to sound a little Joe-like here. <G>)

But you CAN prove that a particular weighting method works better than another. We are looking for improvement here.


And improvement is good. <G>

Dave

PS: Where is that thread I was expounding on yesterday? I can't seem to find it.
PPS: Should I bother? Is anyone listening? Sorry if I am being a boor.

[formula_2002]

So, what is a databaser to do?

A typical circuit has about 2500 races per year. (Recall that we only used fast/firm tracks in our database but let's don't quibble about that.)

So, if a guy is downloading 10 circuits, he will wind up with about 25,000 races per year.

And if he downloads for 3 years to build a database he will have around 75,000 races.

So, let's divided 75,000 races by only 14,000 categories. We get an average of 5.36 races per category! Hardly worthy of even being called a "sample."

In our giant database (375,000 races) and 20,000 categories we average 18.75 races per category, but the largest category has over 3,500 races. That would be, 6f, claiming, older males, fast tracks, 8 horses, etc. (Note that we segment just a little differently.)



The Point
(You knew there'd be one, right?)

Actually there are two.

First, the database guy says, "Well, I'm going to have to break the database into larger pieces - less categories."

So, he wants a miniumum of 1,000 races per category so, in a 75,000 race database he can have 75 categories, right? Wrong! Some of those categories will still have 3-race samples! He may wind up with 20 "statistically significant" (according to his definition) categories and the other 55 categories may well not have enough sample size to study.

Fine. So he only plays (a guess here) 1/2 the races.


Second, by lumping "less-than-like" races together (He reduced the permutations but not slicing and dicing so fine, remember?) his predictive accuracy MUST suffer.

Think of it like this - As a category is studied, he is, essentially, building a approach for the average race in the sample. Consider that some races within the segment are closer to "average" than others.

We might, for example, say that there are 3 "sub-segments" within the data: Average, Close-to-Average, and Far-from-Average. (You could just as easily break it down with 4 or 12 or 87 sub-segments if you wish, but this seemed logical to me.)

So, when he handicaps an "average" race, he is getting the best performance possible from his sample: It was meant for this kind of race.

But what about when he handicaps the Close-to-Average group? He is not exactly handicapping apples with oranges. It is more like tangerines with oranges. But no matter how you juice them (pun) the accuracy of the predictions should be weaker with this group than with the Average group.

Finally, the Far-from-Average group is like handicapping avocados with oranges and we would expect the performance to suffer.


One More Thing
It isn't just that the other two groups will perform below the "average" group. Consider also that those other two groups are being included in the analysis of the sample; effectively, the avocados are being used to determine the juice of the oranges!

In other words, the handicapping process of the Average group is being mislead (for lack of a better word) by the Far-from-Average and Close-to-Average groups!


Next: Solutions...

here it is...
Looking forward to the solutions..

[Exactaman]

some random responses and an actual opinion

the word "proof" should never be used in association with horse racing. there are so many variables in the equine equation, the word is clearly not applicable. unless you want to prove a horse has 4 legs. hell there are even mares and geldings.

nothing is alike about any particular 3-1 shot except that the crowd will lose an average of 17% on them. realizing this is as good as the entire point of betting on horses! attempting to prove it would be like trying to prove that urine is the best thirst-quencher.

once again, just missed misses the point of value betting spectacularly. if you can pick a horse just missed for god's sake, why can't i assign a probability to a horse winning? if i do that, my kid sister could make a betting decision based on it. i can only assume from your statement that in fact you have no ability to pick a horse. now this seems provable. in any event, i defy you to refute the former part of this statement.

one can prove that one's oddsline is correct, unless you are a mathematician. LONG TERM RESULTS DO NOT LIE unless you figure someone could be very very lucky

if you make a horse 2-1 and it goes off 20-1 something is a lot more likely SERIOUSLY wrong with your handicapping than with the crowd's opinion

ok now for the opinion :) don't hold your breath :)

let's all try to catch some winners out there. oh yeah a horse doesn't have 6 sides :):):)

[sjk]

I don't see why you would want to subdivide your dataset of races into subgroups to evaluate the statistical effectiveness of your process. I don't see any value in subdividing by odds groups as Formula insists on doing or on subdividing in the myriad of ways that Dave suggests.

Say you bet 10,000 races based on the decisions that your handicapping process leads you to. At the end of the year you either have more money in your account than you started with or you don't. If you don't I think it is fair to judge that your process is not very effective. If you do have more, you can use some simple statictics (on Joe's page) to determine that probability that your success is due to luck vs. the effectiveness of your process.

Why divide into subcategories? You often read about medical trials with 10,000 participants which prove the effectiveness of some new medicine or procedure. What if every time they completed such a successful trial they categorized the participants by height or weight to see if statistical significance occurred in each subgroup? It never would.

If you take the dataset of all sunrises in recorded history and subdivide into subgroups based on their calendar date you would be unable to demonstrate a likelihood that the sun will rise in the morning. What's the use of that?

[Steve 'StatMan']

I took awhile in creating this post - hope I don't repeat new stuff or come back in a wrong spot.

Let me try to clarify my own position a bit in light of some of the recent posts.

I'm not comfortable with a computer assigning an odds line, and I'm not a slave to an odds line where I'd bet my 4th or worse choice in a race. But I use my handicapping, lots of stats, speed figures, etc. as a guide to judge the talent in the race, and get a feel for how likely the horses are to actually run their good races. It helps to make an estimate of what my chances are. If I like a horse and he's 1/1, I'm not interested, esp. since I don't think I'll be hitting more than 50% of those.

A person may not need to know that a horse has a 25% chance to win, but if you're going to be a 3/1 shot, you'd better at least feel you're going to win more than 25% when you make the bet. Otherwise, that particular bet is a bad deal. Where I and others may get sloppy, is either not going deep enough in the analysis, or assuming we're better at judging what is or isn't a good bet than we really are. That's why it can help to either quantify, or at least make a reasonably good estimate.

The overlay therory (without a line) works for Trifectas too. If you're boxing 4 horses in a $2 Tri (Cost $48), you better be able to hit at least 20% of them if you're going to bet one that will pay $240.

(($240 - $48) X 20%) - ($48 X 80)
($192 X 20%) - ($48 X 80)
38.4 - 38.4 = 0 = Break Even

You may very well not need to keep records, but it sure might help to know you hit 20% of your trifecta boxes, and that this one seems more likely to hit than usual, so that if it does pay $240, I still would be getting a good deal.

The odds line style of analysis just helps one ensure the situation is potentially as good as one feels it is.

Yes, you can know 1, 2, 3 years later if you're ahead that you're doing it right. For those of us who struggle with feeling we make good wagers, yet keep losing money over time, with no records to go back and track what when wrong, or determine which of our 'feelings' was more correct than others. For many, this is an important tool to help turn the corner.

I know when I followed it, I had success. When I let myself get too distracted and blew it off, this is when I've stuggled.

[thoroughbred]

Dave,

I don't agree with your statement:

"Any odds line that is made without considering the public's opinion as part of the line-making process is likely flawed."

Let's for the sake of argument take a case where someone can make an absolutely accurate oddsline based on true probabilities as if they were known. (I know this is not realistic, just, as I said, it's for the sake of argument.)

With such an accurate oddsline, one would be able to establish true and accurate risk to reward ratios by comparing such an oddline with the actual final odds.

If acutal odds were taken into account, in setting up the oddsline, that oddsline, in this idealistic example, could be made inaccurate due to whatever errors were made by the public in establishing the actual odds. Further, and probably even more important, using the public odds as part of the oddsline and then comparing it to the public odds can interfere, in some substantial way, with the attempt to establish true odds as opposed that of the public.

While this example used an ideal hypothetical example, I believe the principle is correct.

[cj]

Here is my problem with factoring the public odds into a personal odds line: No matter how much you factor in the odds, you won't change the fact that the horse you are studying is an overlay or an underlay.

Let's say I think a horse is 4-1, or 20%. The public makes him 2-1, or 33%. No matter how I merge these two, the horse will still be an underlay. Now, if the horse is 9-1, or 10%, no matter how I merge these two lines, the horse will still be an overlay. Of course you can change the degree of underlay or overlay, but never the side the horse falls.

I think the best use of public odds is to make you reevaluate when there are huge differences. Who is making the error? If you study carefully and find it was you, you can save some big bucks. If its the public, you can make some big bucks.

[GameTheory]

Here is my problem with factoring the public odds into a personal odds line: No matter how much you factor in the odds, you won't change the fact that the horse you are studying is an overlay or an underlay.

Let's say I think a horse is 4-1, or 20%. The public makes him 2-1, or 33%. No matter how I merge these two, the horse will still be an underlay. Now, if the horse is 9-1, or 10%, no matter how I merge these two lines, the horse will still be an overlay. Of course you can change the degree of underlay or overlay, but never the side the horse falls.

I think the best use of public odds is to make you reevaluate when there are huge differences. Who is making the error? If you study carefully and find it was you, you can save some big bucks. If its the public, you can make some big bucks.

If your betting method is to flat bet the overlays, then you're right, it doesn't matter. (Most the time -- if you do a lot of adjusting and then renormalize sometimes something can switch from an underlay to overlay or vice-versa.) If the relative rankings of the contenders affect your betting, or if you vary your bet size based on perceived edge, then it matters a lot.


NOTE: Mumbo-jumbo averse may skip the rest of this message.


Thoroughbred's comment about oddslines brings up a common misconception -- that there is some ideal or "true" probability for each horse. It doesn't exist, even in theory. Probabilities are subjective, and I really mean subjective -- they are not subjective approximations of some objective "true" probabilitity, they are subjective period. There is no such thing as true probability, because something either happens or it doesn't.

One person can give a horse a probability of .10 and be just as true/perfect as someone else who gives the same horse a probability of .60. What we are really talking about with probabilities is degrees of belief about the relative winning chances of the horses. Those beliefs are informed by the information/data we possess about the race, and are interpreted by a computer or a person's mind or both. So GIVEN a certain person at a certain point in time with a certain set of data, you can say there is a *optimal* probability to be assigned by them, but there is no absolute correct probability for everybody except to be completely right all the time -- give 0's to all the losers and 1's to all the winners, which of course no one can achieve. [Although I can't prove it with statistics! :) ]

And if we want to take the "interpreter" of the data out of the question (the person, the computer, whatever) and just talk about the data, we can say that there is an optimal probability to be assigned GIVEN THIS EXACT SET OF DATA. In other words theorhetically there is some exact amount of information to be gained (which can inform our probability) from an specified set of data. But since in the real world different people are using somewhat different sets of data and interpreting them in different ways it doesn't really make sense to speak of objectivity or of "true" probabilities.

[formula_2002]

[QUOTE= But since in the real world different people are using somewhat different sets of data and interpreting them in different ways it doesn't really make sense to speak of objectivity or of "true" probabilities.[/QUOTE]

I know you have written alot but the above jumped out at me.

for basis of calculation, the following represents the public's
estimate of the true odds.

((1/(odds+1)/ (1/(1-track take)).

Skewed from short priced horse to long.

If you can beat that enought times, you are what I would call a winner.

thats why 3-1 shots win 21% of the time at a 15% take.

win % =(1/(3+1))/ (1/(.85))
= .25/(1.17)
= 21%



After hitting an exacta one day, my friend got excited for me a said, "Wow you won $200 for your $2.00 bet"
I turned to him and said, "no, I actually lost 40 cents" true story. He also thought I was nuts.

[Overlay]

I agree that, in terms of any one race, there can be only one winner, and thus in retrospect one of the horses in the race will have had a true 100% chance of winning the race, and all the others will have had none. But wouldn't the assignment of a percentage probability have real meaning in the context of a large sample of races, where you could determine whether horses to which you assigned any given probability of winning in their individual races did in fact win that percentage of the total number of races in the sample? And if the sample is large enough to pass statistical tests designed to show whether the "fit" of the projected probabilities versus the actual winning percentages is too close to be accounted for by random variation, wouldn't that serve as the most reliable basis that can obtained for determining whether or not a horse offers betting value in future races (short of perfect foreknowledge of the races' outcomes, which is impossible)?

[JustMissed]

I assume you are making this point to, in some way, try to show that determining an underlay, (or overlay) is impossible because of the comparison of the subjective to the quantitative.

But, if so, you are forgetting a key point. It is possible to calculate, quantitatively, what the odds for each horse should be. In CompuTrak, for example, an oddsline is calculated based on quantitative factors from race performance. Based on the quantitative values that determined such an oddsline, a comparison with the final actual odds will, quantitatively show, whether a horse is going off as an underlay or overlay.

Thoroughbred, That was a thoughtful reply but I think you are thinking wrongly about overlays.

Quick story: A little girl was helping her mother cook Christmas dinner. The mother cut off the bone end of the ham and put it in the roasting pan with the ham. When the little girl asked her mother why she cut the end off the mother replied "I don't really know, my mother taught me to cook and that is the way she always did it".

When the grandmother came over Christmas Day, the little girl asked her about cutting off the bone end of the ham. The grandmother told her that she had a small roasting pan and the only way she could get the ham into the pan was to cut off the end.

Can you see that the mother prepared her hams wrongly for all those years without thinking about what she was doing or questioning that particuliar preparation.

a. Some book writer years ago wrote that you should figure out what the odds of a horse should be and if they were less than the tote odds, it was an overlay-if greater-it was an underlay.

b. Handicapping software came around and none of them had oddslines. Some sharpie figured out if he would have his software calculate an oddsline to compare with the track handicappers oddsline-he could sell more software.

c. Later on another sharpie came along and figured since the oddline was just sitting there doing nothing, why not use it to calculate over/underlays and make our betting decisions based on that.

There you go Thoroughbred. Pace and speed figs are based on quantitative factors from race results yet how different are these figures from vendor to vendor. Why are they different? Because they are subjective in nature. Same can be said for prime power numbers, ranks, ratings, sheet figs or any other figure.

As you well know, the only way to be profitable at the track is to have the teller give you more money than you give her. Period.

There are only three factors that determine your profitabilty:

1. Your win percentage.
2. Your average mutual price collected.
3. The volume of your wagers.

There are many factors which will aid you in improving 1,2 & 3 above but calculating the amount of perceived overlay is not among them.

JM

[GameTheory]

I agree that, in terms of any one race, there can be only one winner, and thus in retrospect one of the horses in the race will have had a true 100% chance of winning the race, and all the others will have had none. But wouldn't the assignment of a percentage probability have real meaning in the context of a large sample of races, where you could determine whether horses to which you assigned any given probability of winning in their individual races did in fact win that percentage of the total number of races in the sample? And if the sample is large enough to pass statistical tests designed to show whether the "fit" of the projected probabilities versus the actual winning percentages is too close to be accounted for by random variation, wouldn't that serve as the most reliable basis that can obtained for determining whether or not a horse offers betting value in future races (short of perfect foreknowledge of the races' outcomes, which is impossible)? Yes, absolutely. But you shouldn't be fooled into then thinking that those are the only set of oddslines you could have used to get such a fit. Here is an example I've often used. Let's say we have a short series of races, each with 4 horses, and we've got three different handicappers assigning probabilities:


RACE #1 Hdcp A Hdcp B Hdcp C
Horse A* 30.00% 10.00% 20.00%
Horse B 20.00% 40.00% 10.00%
Horse C 40.00% 20.00% 30.00%
Horse D 10.00% 30.00% 40.00%


RACE #2 Hdcp A Hdcp B Hdcp C
Horse A 10.00% 30.00% 40.00%
Horse B* 40.00% 20.00% 30.00%
Horse C 20.00% 40.00% 10.00%
Horse D 30.00% 10.00% 20.00%


RACE #3 Hdcp A Hdcp B Hdcp C
Horse A 20.00% 40.00% 10.00%
Horse B 30.00% 10.00% 20.00%
Horse C* 10.00% 30.00% 40.00%
Horse D 40.00% 20.00% 30.00%


RACE #4 Hdcp A Hdcp B Hdcp C
Horse A 40.00% 20.00% 30.00%
Horse B 10.00% 30.00% 40.00%
Horse C 30.00% 10.00% 20.00%
Horse D* 20.00% 40.00% 10.00%

* = winner of the race


Our three handicappers don't agree about a single horse in any of these races. But no matter how you cut it, all three handicappers achieved EXACTLY the same performance in terms of assingning probabilities.

So while the public line is indeed very good, it is not the "true" line because there is no true line except 1 & 0's. There are an infinite number of lines that can be assigned for each level of performance. Now in reality, given the information we have and the actual abilities of the horses, the number of REASONABLE lines that we can come up with shrinks as our performance increases (as we assign higher highs and lower lows and still do it accurately) and it is indeed quite tough to create a line that is both reasonably accurate and significantly different than the public's. So while there are some very real obstacles to creating great lines that are different from the public, we can at least be assured that we are not theorhetically chained to their assesments in order to acheive high accuracy.

[thoroughbred]

Game Theory

Please explain this statement of yours a bit more:

"There is no such thing as true probability, because something either happens or it doesn't."

I'm puzzled by it. Would you say that if you toss one die, which has six sides, like in a game of craps, that the true probability of it coming up, say, a four, is not 1/6 th just because it either happens or it doesn't?

[sjk]

Just Missed,

The odds line/overlay approach works well for me. When I bet a horse (or exacta) the only thing I know about the horse is his number and that my computer thinks he is an overlay.

Usually after I make the bet I look up his name so that I can follow him while watching the race.

[JustMissed]

Dave, You made a lot of really good points.

I really wish you would give using an oddsline to base betting decisions another look. I really think you would conclude that it is some sort of Voodoo.

As far as using the publics oddsline(tote) to adjust your own oddsline, I can see where you are coming from. Of course this is not rocket science-is it? If you really like a horse and the tote shows him at 20-1, I don't know about you but I am going back to see what I missed.

Likewise, if I like a horse that is at 6-1, I probably already know what I know that the public doesn't know.

JM

[Kreed]

Probability is FAR FAR more complicated than it appears. Coin tossing &
card gams --- get real. When a set of conditions get closer to a Pr of 1
(or zero) you're into functional relationships, not Pr. And Pr's change with
more information as in, "What is the Pr of having natural Red Hair?" vs "What
is the Pr of having natural Red Hair if you're Irish & Freckled?"--- most cappers
don't specify (in advance) the Population that they're trying to understand from
their Samples. A large mistake.

[thoroughbred]

Thoroughbred, That was a thoughtful reply but I think you are thinking wrongly about overlays.

Quick story: A little girl was helping her mother cook Christmas dinner. The mother cut off the bone end of the ham and put it in the roasting pan with the ham. When the little girl asked her mother why she cut the end off the mother replied "I don't really know, my mother taught me to cook and that is the way she always did it".

When the grandmother came over Christmas Day, the little girl asked her about cutting off the bone end of the ham. The grandmother told her that she had a small roasting pan and the only way she could get the ham into the pan was to cut off the end.

Can you see that the mother prepared her hams wrongly for all those years without thinking about what she was doing or questioning that particuliar preparation.

a. Some book writer years ago wrote that you should figure out what the odds of a horse should be and if they were less than the tote odds, it was an overlay-if greater-it was an underlay.

b. Handicapping software came around and none of them had oddslines. Some sharpie figured out if he would have his software calculate an oddsline to compare with the track handicappers oddsline-he could sell more software.

c. Later on another sharpie came along and figured since the oddline was just sitting there doing nothing, why not use it to calculate over/underlays and make our betting decisions based on that.

There you go Thoroughbred. Pace and speed figs are based on quantitative factors from race results yet how different are these figures from vendor to vendor. Why are they different? Because they are subjective in nature. Same can be said for prime power numbers, ranks, ratings, sheet figs or any other figure.

As you well know, the only way to be profitable at the track is to have the teller give you more money than you give her. Period.

There are only three factors that determine your profitabilty:

1. Your win percentage.
2. Your average mutual price collected.
3. The volume of your wagers.

There are many factors which will aid you in improving 1,2 & 3 above but calculating the amount of perceived overlay is not among them.

JM

Hi JM,

An interesting story. Yes, we should be careful not to be using a procedure just because it has been used before and we go along with it for that reason without thought. I heartily agree wtih you on that.

But the final odds, are calculated from probabilities, i.e. in this case from the amounts the public has bet on each horse, (subjective), and it is those actual odds that determine the amount of the payoff. As we all know, with the track takeout, the actual odds are lower than they would have been without the takeout. So to make a good profit in the long run, we have to look for overlays that will pay enough to overcome the long term effects of the takeout.

You are correct, that many oddlsines, and the factors that oddslines depend on, differ from vendor to vendor, and often they do base their calculations on subjectiie facors. So in truth our discussion comes down to the accuracy of determiining the probability of each horse winning. Just because we cannot achieve perfection in determining an accurate probability, does not mean that we should not keep on trying, nor should we discard the basic principle that there is a probability for each horse winning, even though it may be difficult to determine it.

[Steve 'StatMan']

Dave, You made a lot of really good points.
Likewise, if I like a horse that is at 6-1, I probably already know what I know that the public doesn't know.

JM

And if you hit more than 14-ish% of those at 6/1, you'll make money.

You probably have a good feeling about your handicapping, and probably will hit more than 14-ish% of your win bets. So you don't need a detailed odds line. You know enough that is value, so you're betting it.

You (and hopefully all of us) know not to bet 1/5 shots to win, because you know you're not going to hit nearly all of them.

Many of us on the 'pro-Line' side are just trying to be sure our judgement is good about the 6/1 shot being a good bet. Some may feel they need a computer program to do it. I'm skeptical, because it can only handle the factors it is given.

We're also trying to convince Joe that just because a person bets 6/1 shots, they aren't automatically going to lose, because bets on every odds range lose money. You just have to have a good enough reason to back that 6/1 shot.

[Dave Schwartz]

JM,

>>>I really wish you would give using an oddsline to base betting decisions another look. I really think you would conclude that it is some sort of Voodoo.<<<

WHat do you mean by "voodoo?"


Dave

[JustMissed]

JM,

>>>I really wish you would give using an oddsline to base betting decisions another look. I really think you would conclude that it is some sort of Voodoo.<<<

WHat do you mean by "voodoo?"


Dave

voo·doo ( P ) Pronunciation Key (vd)
n. pl. voo·doos
1. A religion practiced chiefly in Caribbean countries, especially Haiti, syncretized from Roman Catholic ritual elements and the animism and magic of slaves from West Africa, in which a supreme God rules a large pantheon of local and tutelary deities, deified ancestors, and saints, who communicate with believers in dreams, trances, and ritual possessions. Also called vodoun.
2. A charm, fetish, spell, or curse holding magic power for adherents of voodoo.
3. A practitioner, priest, or priestess of voodoo.
4. Deceptive or delusive nonsense.


JM :D

[Jeff P]

Posted by GameTheory-

Thoroughbred's comment about oddslines brings up a common misconception -- that there is some ideal or "true" probability for each horse. It doesn't exist, even in theory. Probabilities are subjective, and I really mean subjective -- they are not subjective approximations of some objective "true" probabilitity, they are subjective period. There is no such thing as true probability, because something either happens or it doesn't.
-------------------------------------------------------------


True probability may not exist, but I think one can get a pretty good idea of probability itself by carefully measuring historical occurances to arrive at a historical probability.

What I'm about to outline is a simple process by which a historical probability (and perhaps a profitable betting method) can be derived.

Let's say for the sake of argument that we develop a mechanical set of selection criteria that we can apply to every horse in a race. Ultimately, that mechanical set of criteria (no matter what that set of criteria happens to be made up of so long as it is based on causal factors) shakes out one horse from each race as that race's highest rated horse.

For me, the question arises, how does the entire group of our mechanically selected top selections perform historically?

Let's say we confront our set of mechanical selection criteria with a set of several thousand fresh races and notice that the top choice produced by our mechanical selection criteria produces 25 percent winners. Let's further say that we perform many subsequent tests using the same selection criteria on additional samples of several thousand fresh races and get very similar results each time. At some point, doesn't it seem plausible that our mechanically selected top choices should have a base probability somewhere very close to 25 percent?

In theory, the horseplayer should be able to bet those 25 percent horses whenever they go to post at odds of 3-1 or greater and turn a profit. But it doesn't always work that way. If we are using causal factors, then sure, we can expect that we will continue to pick 25 percent winners. But if we are using the same or similar factors that the betting public uses, then a very high percentage of the time our 25 percent horses are going to be underlays. In order to be bet profitably, our 25 percent horses are going to have to be selected using factors that are not known or understood by our competition- the betting public.

Also, as Dave alluded to, and this is important the odds DO give us some basis for changing our probability estimate. The odds reflect the cumulative opinions belonging to the thousands of players that happen to be looking at any one race. If one of our 25 percent horses is sent to post at 30-1, chances are very strong that our selection doesn't have a true 25 percent chance of winning the race we are looking at. Conversely, when one of our 25 percent selections is sent to post at 3-5, chances are very strong that our 25 percent horse really has a much higher probability of winning than 25 percent.

Can we change our probability estimate and make it more accurate based on the odds created by all those other players? Fortunately, yes.

We can do this by using the same process that got us to the 25 percent number in the first place. If we have enough data, we can break our data into segments according to post time odds and ask the ourselves questions along the following lines:

How do our 25 percent horses do when going to post at 6-5? How do our 25 percent horses do when going to post at 6-1? At 12-1? At 20-1?, etc. Then after sampling our 25 percent horses in this revised odds based context, we can get a much better idea of what probability is- perhaps even refining it to the point that we gain an understanding of the historical impact that the odds have on our selections.

Where is all this leading to?

What if somewhere along the line we happen to notice that the factors in our mechanical selection criteria are unique enough so that horses mechanically selected by us are ignored by the public often enough to be bet profitably?

Doesn't it follow that we should be betting them?

It does. But at what odds ranges?

Traditional thinking has us betting our 25 percent horses at 3-1 and up. Like Dave, I think this approach is flawed to some degree. Yes, if we have done our homework, and are using unique causal factors, and are hitting 25 percent winners overall, then the entire set of horses we bet at 3-1 and up should be profitable.

The good stuff My own research keeps telling me one thing: The traditional type of thinking related to using an odds line can be improved upon a great deal. When studying groups of profitable win bets broken out by odds ranges I very often see the following pattern: There often appears to be a "Sweet Spot" in the odds where maximum profits can be achieved. Just as there is a minimum odds range below which it is detrimental to bet (and this represents traditional thinking) there is also a maximum odds range above which it is detrimental to bet. As the odds climb upward, the win rate drops off. When the odds hit a certain point going upward, the win rate becomes too low to warrant sensible play (at least for me.) And as the odds climb even higher, very often no winners exist in that part of the sample at all.

[JustMissed]

Hey Jeff, that was a real fine post but weren't you really saying:

1. When you bet low price horses you hit more but lose money.

2. When you bet higher priced horses you hit less and still lose money.

3. When you bet along with the public you are in the "sweet spot".

What is that about?

Like I said before, this ain't rocket science.

All you have to do is have more hits at higher prices or have higher prices on your hits.

JM

[GameTheory]

Game Theory

Please explain this statement of yours a bit more:

"There is no such thing as true probability, because something either happens or it doesn't."

I'm puzzled by it. Would you say that if you toss one die, which has six sides, like in a game of craps, that the true probability of it coming up, say, a four, is not 1/6 th just because it either happens or it doesn't?Yes. However, when discussing something as well-understood as a 6-sided die, there is no practical value to my assertion. But on a theorhetical level, yes.

But it is a good example because it illustrates my point -- you are thinking that because your probability is accurate over a large sample (i.e. if you roll a die many times, 1/6th of the time you'll get a four) that you've arrived at the "true" probability. In other words, you are saying, "Since I am right and the frequency of 4's is 1/6th, then that's the true probability -- if you don't choose 1/6th as the probability, you are wrong."

But you're assigning a probability of 1/6 for a single roll -- a roll which will either come up a 4 or not. Now let's explore why you assigned 1/6th in the first place. If we go along with the idea that a probability represents a degree of belief in whether an event will occur or not, we need to ask what those beliefs are based on. Presumably, you assigned 1/6th because of what you know about the die -- namely it has six sides and you presume it is fairly weighted, etc. So GIVEN what you know about dice in general (and what you are assuming about a particular actual die), you come up with the quite accurate probability of 1/6th, a probability that is as accurate as it can be GIVEN WHAT YOU KNOW.

Now since you've chosen dice, which we humans have a pretty good handle on, we have to get a little fanciful here to make the rest of the point. Anyway, now imagine there is some superbeing that can see the molecular structure of the die to see EXACTLY how it is weighted, that he can somehow calculate the Newtonian physics of how it is going to bounce based on the throwing motion and habits of the thrower, etc etc. In other words, he has more information about each particular throw than a normal mortal. When this superbeing picks which side is going to come up, he is so confident that he assigns a probability of 3/6 or 50%. And let's say he is right -- his prediction comes up correct 50% of the time. So what's the true probability? 3/6 or 1/6? You're both correct frequency-wise.

How is the super-being able to be right more than 1/6th of the time? Because he has superior information -- information not available to you or not interpretable by you.

An event either happens or it doesn't. The probability we assign is the representation of the degree of belief we have that it will happen GIVEN the information we have/know/think-we-know.


Look at my example with the handicappers -- each of them was right exactly as much and the other ones yet they didn't agree on anything. Horse-racing is much more complex than rolling a single die. We don't necessarily have to be God or supermen to imagine that we can gain an information advantage (which will be translated into a probability advantage) over the public. So being right (meaning your probabilities match observed frequencies) doesn't mean you have found the true probability, because someone else can be just as right with a different probability (and different information).

I have also often given the example of simply assigning every horse in the race an equal chance just as you would like to assign an equal chance to every side of a die. And your probabilities will be perfect frequency-wise just as they are with the die, but you would never argue they are true, would you? You would look at such a handicapper as moron because of his poor use of the information he has. Just as the superbeings who can pick dice 50% of the time would look at you as a moron for only being right 1/6th of the time.

[Dave Schwartz]

JM,

LOL - I KNOW what voodoo IS. Really I meant what is the alternative you are suggesting to this voodoo?

Dave

[GameTheory]

SNIPYes, yes, yes. Good stuff.

I use the odds to find factors that will "work well" with the odds. I first take a look at each factor in isolation and determine its predictiveness alone. Then I look at the correlation of its predictions with the odds. The most predictive factors are usually highly correlated and fairly worthless. But a *mildly* predictive factor -- a factor that is weak but real that is mostly uncorrelated with the odds can become killer when in combination with the odds.

In this way, I use the public line as my starting point, and then improve upon it, shoving it this way and that with my weak but uncorrelated factors. And I end up with a combination line that is dynamite -- better than the public's.

[Dave Schwartz]

GT,

And what you just described is how statistics should be used in horse racing.

Excellent!

Dave

[JustMissed]

Yes. However, when discussing something as well-understood as a 6-sided die, there is no practical value to my assertion. But on a theorhetical level, yes.

But it is a good example because it illustrates my point -- you are thinking that because your probability is accurate over a large sample (i.e. if you roll a die many times, 1/6th of the time you'll get a four) that you've arrived at the "true" probability. In other words, you are saying, "Since I am right and the frequency of 4's is 1/6th, then that's the true probability -- if you don't choose 1/6th as the probability, you are wrong."

But you're assigning a probability of 1/6 for a single roll -- a roll which will either come up a 4 or not. Now let's explore why you assigned 1/6th in the first place. If we go along with the idea that a probability represents a degree of belief in whether an event will occur or not, we need to ask what those beliefs are based on. Presumably, you assigned 1/6th because of what you know about the die -- namely it has six sides and you presume it is fairly weighted, etc. So GIVEN what you know about dice in general (and what you are assuming about a particular actual die), you come up with the quite accurate probability of 1/6th, a probability that is as accurate as it can be GIVEN WHAT YOU KNOW.

Now since you've chosen dice, which we humans have a pretty good handle on, we have to get a little fanciful here to make the rest of the point. Anyway, now imagine there is some superbeing that can see the molecular structure of the die to see EXACTLY how it is weighted, that he can somehow calculate the Newtonian physics of how it is going to bounce based on the throwing motion and habits of the thrower, etc etc. In other words, he has more information about each particular throw than a normal mortal. When this superbeing picks which side is going to come up, he is so confident that he assigns a probability of 3/6 or 50%. And let's say he is right -- his prediction comes up correct 50% of the time. So what's the true probability? 3/6 or 1/6? You're both correct frequency-wise.

How is the super-being able to be right more than 1/6th of the time? Because he has superior information -- information not available to you or not interpretable by you.

An event either happens or it doesn't. The probability we assign is the representation of the degree of belief we have that it will happen GIVEN the information we have/know/think-we-know.


Look at my example with the handicappers -- each of them was right exactly as much and the other ones yet they didn't agree on anything. Horse-racing is much more complex than rolling a single die. We don't necessarily have to be God or supermen to imagine that we can gain an information advantage (which will be translated into a probability advantage) over the public. So being right (meaning your probabilities match observed frequencies) doesn't mean you have found the true probability, because someone else can be just as right with a different probability (and different information).

I have also often given the example of simply assigning every horse in the race an equal chance just as you would like to assign an equal chance to every side of a die. And your probabilities will be perfect frequency-wise just as they are with the die, but you would never argue they are true, would you? You would look at such a handicapper as moron because of his poor use of the information he has. Just as the superbeings who can pick dice 50% of the time would look at you as a moron for only being right 1/6th of the time.

Absolutely Brilliant GT,

That was probably the best thing I have ever read on the board.

It appears that a lot of horseplayers have been DUPED by those that try to apply casino probablities to a single horse racing event as well as the MYTH of the preceived OVERLAY.

Bravo,

JM

[Dave Schwartz]

Jeff,

Good post. I basically agree with you.

However, the part that the public's opinion plays (i.e. the tote board) is far more than trivial. (Not that you said it was.)


WHat I mean by that is that when you take that 25% system and remove the 5/2 and below horses that theoretically would be losers, you no longer have a 25% system.

In fact, you may well have a 15% or 10% system.


Perhaps the best example of this is what happens if the factor you are looking at is "Favorite? yes/no."

When we total up the "yes" horses we get 33% (or thereabouts) as a win percentage. Using the "25%-overlay-is-3/1 up" theory all we have to do to win is to bet all the favorites that go off at over 2/1.

Of course, it does not work that way because we find that as the tote odds go up the win probability goes down and vice-versa.


Well, why would it be any different with any other factor?

If top Beyer speed rating in the last race produces 25% winners, why would that be any different? As the odds go up or down, the probability moves with it. Of course, the relationship between the tote and the up/down movement would not be as closely correllated but there would still be a correllation.

So, my point is that there are no "25% factors" really. That is a non-issue. The reality is that there is a relationship between the public's opinion (i.e. pool percentage) and probability.

We have been calling it "PIV" (pool impact value) for almost 10 years. It is the computed by dividing the actual winners by the expected winners (just like Impact Value) except that "expected winners" are determined by the percentage of pool wagered on the horses. In other words, a horse with 20% of the pool should produce 20% of a win. (And 20% of the pool is not 4/1.)


Dave

[JustMissed]

JM,

LOL - I KNOW what voodoo IS. Really I meant what is the alternative you are suggesting to this voodoo?

Dave


I would suggest the alternative to voodoo is non-fraudulent and truthful belief.

JM

[Dave Schwartz]

JM,

What do you mean by "non-fraudulent?"

Are you sugesting someone is committing a fraud here?


Dave

[GameTheory]

It appears that a lot of horseplayers have been DUPED by those that try to apply casino probablities to a single horse racing event as well as the MYTH of the preceived OVERLAY.
Oh, I believe in overlays, and I believe in assigning probabilities as best you can. But they are subjective, and you can have two methods that have wide disagreement about the probabilities of individual horses but still with both performing about equally well on average.

The practical point behind all the "no true absolute probability" stuff is that I've seen many people get discouraged because they've heard so often, "The public odds represent the true probabilities, or are at least closer than anything else" and they think well what's the point -- you can't beat the public if they've already got the "truth" nailed down.

[formula_2002]

GT writes
" So being right (meaning your probabilities match observed frequencies) doesn't mean you have found the true probability, because someone else can be just as right with a different probability (and different information)."

So, if I am right in saying a coin has a 50/50 chance of head or tails, I have not found the true probabilty...
because in another, lets say, in a 10th dimension,
I have determined the coin behaves in a different manner thereby altering the probability.

I'll buy into that so long as we agree that once we have determined the various probabilities we are right.

ps.
It gt wants to reply..please address it to the PA's 10th dimension board ..I'll pick it up there..

[GameTheory]

GT writes
" So being right (meaning your probabilities match observed frequencies) doesn't mean you have found the true probability, because someone else can be just as right with a different probability (and different information)."

So, if I am right in saying a coin has a 50/50 chance of head or tails, I have not found the true probabilty...
because in another, lets say, in a 10th dimension,
I have determined the coin behaves in a different manner thereby altering the probability.
You are right if you say a fair coin tossed many times will have a frequency of roughly 50% heads and 50% tails. But you are assigning a probability for one toss only that you are going to bet on -- it is going to be 100% heads or 100% tails, not 50% of both. I am making a distinction between probability and frequency.

[InsideThePylons-MW]

After reading this thread, I am glad that I spent my time "learning how to extract money from races" and that over time, the probability of this is 100%.

[formula_2002]

You are right if you say a fair coin tossed many times will have a frequency of roughly 50% heads and 50% tails. But you are assigning a probability for one toss only that you are going to bet on -- it is going to be 100% heads or 100% tails, not 50% of both. I am making a distinction between probability and frequency.

GT, if I understand you;

probability = favorable outcomes/ total outcomes.

probability = 50 tails/ 100 tries= 50%

frequency = 50 % of the time it is 100% heads
frequency = 50 % of the time it is 100% tails

correct?

If that is correct I will go further.

[JustMissed]

JM,

What do you mean by "non-fraudulent?"

Are you sugesting someone is committing a fraud here?


Dave

Dave, You should buy yourself a dictionary and keep it by your computer to look up words you don't understand.

You asked me the meaning of voodoo and I explained to you one of the meanings was:

4. Deceptive or delusive nonsense.

Then you asked me: "Really I meant what is the alternative you are suggesting to this voodoo?"

Deceptive means "an illusion or fraud", an alternative to that would be non-fraud.

I would never accuse anyone of fraud but I certainly do believe that using perceived overlays to make betting decisions is a fraudulent notion.


JM

[hurrikane]

I'm of the opinion that some of you need to get out of the house once in a while. :D

[JustMissed]

I'm of the opinion that some of you need to get out of the house once in a while. :D

Hey, I getting ready to head down to Tampa Bay Downs right now.

Today is hat day and I don't want to be late.

The first 6,500 people to show up get a FREE TBD Hat.

Just cost $2 for admission and you get a free hat to boot-Now that is what I call a VALUE PLAY.


JM

[Exactaman]

i have been making an oddsline and betting off it for 4 years so i hardly consider it to be voodoo or fraudulent. no vendor sold me on it, i use no program. i came to this method through hard experience and information freely given by others.

just missed, i read through all your comments, and the only one i felt was not an automatic mindless gainsaying of the value of an oddsline was this:

Like I said before, this ain't rocket science.

All you have to do is have more hits at higher prices or have higher prices on your hits.

how ironic, i could not think of any better explanation of the benefit of an oddsline! all i am trying to do is ensure i get better prices on the horses i select that hit. where is the voodoo, much less the fraud? indeed this is not rocket science, and certainly, CERTAINLY no more rocket science than saying "i think number 1 will win so i will bet him." each one is a leap of faith based on personal opinion. either method may succeed or fail, but the value of using an oddsline to bet is PRECISELY to ensure higher prices are received. bravo! you have grasped it perfectly.

i use a simple mathematical basis for my bets. the day you can ascertain 100% which horse WILL win a race i will admit you are right, until then i am just as right to say, these 2 horses have a similar chance of winning, i will bet on the one with the higher odds, as you are to say, this one will win and i will bet it. again, i defy you to disprove this statement.

[Dave Schwartz]

JM,

I am trying to understand this... Are you saying that my suggestion is fraudulent?


Dave Schwartz

[formula_2002]

Today, 1/22/2005, Bris list about 20 tracks that will have racing today.

Prediction

20 tracks x 9 races x 8 horses = 1440 horse will run today

1440 X .09 = 129 of those wil be 3-1 shots

129 x .21 = 27 of those horses will win..

(actually the win % will conform to the formula I posted previously)

The accuracy of those figures will increase as day after day after day goes by..

Anyone here dispute that?

[cj]

You finally understand Joe, if you bet the same as the public, you will lose around 15-20%. Bravo, but I could have told you this without the cost of all those All Ways files. The good news is, you don't have to bet ALL the 3-1 shots, only the ones of your choosing!

[GameTheory]

GT, if I understand you;

probability = favorable outcomes/ total outcomes.

probability = 50 tails/ 100 tries= 50%

frequency = 50 % of the time it is 100% heads
frequency = 50 % of the time it is 100% tails

correct?

If that is correct I will go further.

Frequency is correct. Frequency is the % of the time some specified event happens (a coin lands on tails, the favorite wins, etc.).

I'm saying probability is subjective, and you don't assign probabilities for sets of events, but for each individual "trial" of the event (a single coin toss, a single horse in a single race). In other words, if you say the probability of landing on tails is .5, you mean it is .5 FOR THIS NEXT TOSS OF THE COIN. Now you may want to use the same .5 value for every toss, but it still applies to one toss at a time.

The coin is just like the die -- .5 is the best we can do not because .5 is the "true" probability (for a single toss it is either 1 or 0), but because that is the best guess our skills will allow us to make. A single coin toss is not a random event -- the laws of physics apply, after all, but we are not skilled enough to predict what is going to happen on a single toss. Dice were invented to be able to simulate randomness, but they aren't actually random -- they are random TO US. What I'm getting at is that probability values are ABOUT US (the people that assign them) just as much as they about the event itself. They represent OUR DEGREE OF BELIEF THAT A CERTAIN EVENT WILL OCCUR GIVEN THE INFORMATION WE HAVE.

Now imagine our coin flipper is a robot that for each "toss" simply picks up the coin and then places back down on its opposite face. So now we have heads-tails-heads-tails-heads-tails. Frequency is unchanged -- 50% each side, but an observant guesser given the task of assigning the probability of the coin coming up heads before EACH toss will quickly acheive 100% accuracy as soon as he picks up on the pattern. Why? Because he now has more information he can use to inform his belief -- the tosses don't have the usual chaos of flipping around in the air and bouncing around on the ground that normally would preclude us from making a better guess.

[formula_2002]

so how does a complete understanding of "frequency" help in picking a 3-1 shot that wins 27% of the time 100% of the time?

Or is that not the issue here?

[JustMissed]

Exactaman, Maybe I don't understand exactly what Value Betting is and I am wrong in my assessment.

Below are some examples from Gordon Pines very fine web site, netcapper.com.

These examples are what I understand value betting to be. If I am wrong, please let me know.

Yours truly,

JM



Track Tracts

Betting Line Wagering Simplified
by Gordon Pine

The first two races of this year’s triple crown illustrate the power of wagering with a betting line. With decent, if less-than-brilliant, handicapping, I was able to take money from both races. How is that possible? In the Derby, my fourth pick won. In the Preakness, my third pick won at 2/1 odds. In both races, my top pick was out of the money.

My meat and potatoes betting includes win, place, show and exacta wagers. If there are overlays in a race, I will generally bet them in some or all of these pools. My wagering method is mechanical, easy-to-implement, and well-worn – I’ve been happily doing it this way for years. The overarching principle is to place bets that involve overlays as defined by my betting line. Here’s how it works:

•Win Betting: Bet if the horse’s odds are equal to or greater than the Bet Odds.
•Place Betting: Bet Win/Place if the horse’s odds are 3+ times the Fair Odds.
•Show Betting: Bet Win/Place/Show if the horse’s odds are 6+ times the Fair Odds.
•Exacta Betting: Wheel the overlay(s) top and bottom with the top two favorites. If there are more than one overlay, box them.

Using the betting line I posted for the Kentucky Derby in Capper’s Corner the night before the race (assuming $20 win/place/show bets and $5 exacta bets) here’s how this betting method worked

Fair Bet Track
P# Horse Odds Odds Odds
3 Balto Star 5/2 4/1 8/1
8 Congaree 7/2 5/1 7/1
17 Point Given 4/1 6/1 9/5
16 Monarchos 7/1 9/1 10/1

$20 Win 3,8,16 = $60
$20 Place 3 = $20
$5 Exacta, 18 combinations = $90
(3+8+16 with 10+17, 10+17 with 3+8+16, and 3-8-16 box)

Total Bet $170
Collected $230
Profit $60

I didn’t expect to get my 9/1 Bet Odds on Monarchos, but they’re generous on Kentucky Derby Day, and it ended up being a bet. I missed my big exacta payday when Congaree got nosed out of the second spot, but the $23.00 win payoff for Monarchos made it a profitable day.

Before the race, the Preakness looked like a passer, unlikely to offering much value. I expected one of the top three favorites to win the race. I did give an outside shot to A P Valentine or Richly Blended to pull off an upset. Here’s the betting line I posted in Capper’s Corner the night before:

Fair Bet Track
P# Horse Odds Odds Odds
7 Monarchos 3/1 9/2 2/1
5 Congaree 4/1 6/1 5/2
11 Point Given 6/1 8/1 2/1
4 A P Valentine 7/1 9/1 10/1
6 Richly Blended 7/1 10/1 17/1

$20 Win 4,6 = $40
$5 Exacta, 10 combinations = $50
(4+6 with 7+11, 7+11 with 4+6, and 4-6 box)

Total Bet $90
Collected $203.50
Profit $113.50

As I and just about everyone else figured, one of the favorites won the Preakness. What allowed me to make money in the race was one of my overlay’s driving finish to get second. Despite a troubled trip, A P Valentine rushed up to complete a bounteous $81.40 (for $2) exacta.

The best explanation of how to use a betting line to make money at the track is contained in Barry Meadow’s book Money Secrets at the Racetrack. My personal betting method is a bastardization of his methods which I developed in the search for the easiest possible method which still incorporated the power of the betting line.

The idea behind my place wagering is, if a horse is going off at three or more times its Fair Odds, it’s likely to pay its Fair Odds payoff or more in the place pool. And that’s for an event that’s much more likely than finishing first – finishing first or second. The same logic applies for show bets when a horse is going off at six or more times its Fair Odds. This is based on something Mark Cramer wrote about – he posed this question: If you’re willing, for instance, to take 4/1 on a horse in the win pool, why wouldn’t you bet it to place if you’re likely to get that price in the place pool?

The notion behind my exacta wagering method is, if a horse is overlayed in the win pool, it’s likely to be overlayed in the exacta pool. Combine that with the fact that one of the top two favorites is going to be in the exacta more than 80% of the time, and you’ve got my method. If any of my overlays runs at least second, I’m probably going to make money in the race. If I have multiple overlays that run one-two, I’m going to have a bonanza. And I don’t have to look at exacta payoffs for 12 different combinations with three minutes to race time – I just watch the win odds, mark my bets, and place them.

I’ve never been a trifecta player, but I’m thinking of testing some ideas. (If anybody has experience with profitable, easy-to-use trifecta wagering methods with a betting line, I’d welcome suggestions.) Here’s what I plan on testing: Trifecta as a Show Bet. I’m going to look at how I’d do if I bet all my betting line contenders (horses with Fair Odds of 7/1 or less) in the first spot, to all my betting line contenders in the second spot, to just my overlay(s) in the third spot. So, it’d be a Contenders-Contenders-Overlays trifecta ticket. A typical one might be 4-4-2 or a $12 ticket. If this works, I could take money from the race if one of my overlays ran in the money.

Keep in mind that each type of bet (win, place, show, exacta, trifecta, etc) should be profitable on its own in the long run. You shouldn’t bet exactas and trifectas just because you feel the need to cash every possible race. No matter how you structure your bets, you’re going to lose sometimes. Strictly speaking, there’s no such thing as a good saver bet if it cuts into your edge. However, you should try to structure your betting so that you focus on overlays, and you get paid if the overlay runs well, even if it doesn’t win. Getting paid if my horse finishes third or better – that’s the kind of leeway I like. Hmm – with the superfecta, maybe I can make it pay if my overlay finishes fourth or better. NC

Copyright ©2001 NetCapper Inc. All rights reserved.

[GameTheory]

so how does a complete understanding of "frequency" help in picking a 3-1 shot that wins 27% of the time 100% of the time?
It doesn't.

But it does let you know that since there is no single "right answer" probability value for each horse in a race, you need not be discouraged from diverging from the public with your assessments because of the fear that they've already got the "truth" nailed down. It also lets you know that two different handicappers could have widely differing sets of oddslines, yet both perform equally well.

[formula_2002]

It doesn't.

But it does let you know that since there is no single "right answer" probability value for each horse in a race, you need not be discouraged from diverging from the public with your assessments because of the fear that they've already got the "truth" nailed down. It also lets you know that two different handicappers could have widely differing sets of oddslines, yet both perform equally well.

I can agree with everything you say here.
I can even understand both of them making 10% profit and not playing the same 3-1 horse.
My point is, I would require them to win at least 75 of any 270 bets, in order to statistically prove that they won by means other then pure chance.

If this is Ok, I can go further.

[Kreed]

I couldn't find the NEW THREAD button but why not hold a Team contest --
say 2 or 3 Team members?

[GameTheory]

I can agree with everything you say here.
I can even understand both of them making 10% profit and not playing the same 3-1 horse.
My point is, I would require them to win at least 75 of any 270 bets, in order to statistically prove that they won by means other then pure chance.
Yeah, so would I if I wanted to prove that. Where are we going with this?

[Exactaman]

Exactaman, Maybe I don't understand exactly what Value Betting is and I am wrong in my assessment.

Below are some examples from Gordon Pines very fine web site, netcapper.com.

These examples are what I understand value betting to be. If I am wrong, please let me know.



the article certainly describes one approach to value betting. i don't agree with everything there and one or two things are perhaps not clearly explained, but it is a good enough example. fwiw my method is quite different, surely there are many.

i would describe value betting as assigning a probability to a bettable event and making a bet when the odds on offer are deemed to exceed that probability. the handicapping skill is in assigning the probability, as far as the handicapping aspect of the game goes, i really don't think this is different than simply attempting to select a potential winner. the problem is simply approached from a different angle.

you mentioned getting higher prices jm, do you try to ensure that? a one-all cutoff price maybe? or consider it when you handicap?

But it does let you know that since there is no single "right answer" probability value for each horse in a race, you need not be discouraged from diverging from the public with your assessments because of the fear that they've already got the "truth" nailed down.

that is it in a nutshell. god bless anyone who can make an oddsline and then blend the crowd's opinion into the mix before making the bet. it's interesting as crowd action can often point towards performance, the whole concept is just too complicated for me though, the whole point of a line being to go against the crowd. i feel comfortable living and dying with my line, though in(relatively rare) cases where i was for one reason or another unsure about a horses assessment during the handicapping process i may make a quick adjustment/pass decision based on crowd action.

so how does a complete understanding of "frequency" help in picking a 3-1 shot that wins 27% of the time 100% of the time?

i feel quite sure it won't :)

[formula_2002]

Yeah, so would I if I wanted to prove that. Where are we going with this?

Not much further.

say all the 3-1 overlays are equally distributed over all the various odds horses.

Now it's just a matter of determining how many races you have to review in order to find those 270 races and win 27% of them in the 3-1 odds range.


270/.275= 981 in the singular odds range >=2.5 through <=3.4).

ALL 3-1 horse represnet 9% of the all the horse that run.

981/.09 = 10,900 total horses off all odds ranges.

say 8 horses per race, 10900/8= 1,362 races.

1362/9 race per day per track= 151 track days.

say a guy bets 3 tracks a day, 151/3= 50 days of non stop handicapping.

So then it is quite simple to prove one can make a 10% profit, at a 15% tak-out track playing 3-1 horses.

If the take-out and or the profit increase, the sample size of 270 decreases.

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

To prove a 10% profit at a 15% take-out track playing 10-1 horses;
2.5% of the horse are 10-1, and 850 plays are required with 10% wins.

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

850/.10= 8500 in the singular odds range >=9.5 through <=10.4).

ALL 10-1 horse represnet 2.5% of the all the horse that run.

8500/.025 = 340,000 total horses off all odds ranges.

say 8 horses per race, 340,000/8= 42500 races.

42500/9 race per day per track= 4722 track days.

say a guy bets 3 tracks a day, 4722/3= 1574 days of non stop handicapping.

If you can check the math for me, and it's ok, I'll post the number of races one would have to check to prove a 10% profit at a 15% takout for all the odds ranges 1 through 10.

[Kreed]

Guys, this thread is VERY keen & I agree with almost all the discussions, but
not with the conclusions. #1, as CJ typed, no matter what the odds, even
3:1, there is NO GUN to your head to BET. #2, ALL Pr arguements into racing
and capping methods ARE NOT THE SAME as in coin tossing experiments.
Shocking? WE ALL know the true Pr in coin tossing, but not so in race
results. #3, There are GROUP stats (like, "Whats the Pr of winning a NW1 race
if the horse has raced >10 times"?) --- but sometimes (maybe rarely) you will
bet on such a horse GIVEN the right set ups. Inclusion in any Group is no
guarantee of winning or losing; in fact, its the art of capping that looks for
the weird situations that buck the Group stat. The real situation of playing
races is that the variables are many & the meaning of those variables is not
always clear. A purely stat approach to picking winners is too simple and
would be very hard to develop, imo.

[formula_2002]

. #2, ALL Pr arguements into racing
and capping methods ARE NOT THE SAME as in coin tossing experiments.
Shocking? WE ALL know the true Pr in coin tossing, but not so in race
results.

Of course we do. Horses win in accordance with the formula I previously.
Craps, and coin tossing also win according to their odds.

Biased coins and dice win in accordance with the biase.
Once the biase is determined by the house, the odds will be adjusted accordingly. (in reality, they would eliminate the biase).
Same with horses.

[Dave Schwartz]

Joe,

I need to tell you that you are becoming a bit of a pita here.

Whenever a thread goes in a direction that might lead to something useful for someone you jump in to say why none of that makes any sense.

You do not believe that racing is beatable and nothing, not even successful players, are going to change your view. I, for one, get it. I think everyone else gets it as well.

My point?

Get out of the way and let the rest of us try to learn something!


Regards,
Dave Schwartz

[formula_2002]

Joe,

I need to tell you that you are becoming a bit of a pita here.

Whenever a thread goes in a direction that might lead to something useful for someone you jump in to say why none of that makes any sense.

You do not believe that racing is beatable and nothing, not even successful players, are going to change your view. I, for one, get it. I think everyone else gets it as well.

My point?

Get out of the way and let the rest of us try to learn something!


Regards,
Dave Schwartz

what meaningful gem am I over looking here.
.

And in addition I started this thread.
I encourage sensible debate, not the "I know this guy" type of bluster.

Now what the hell is a "pita". It took me a year to figure out what a "troll" was, only to find out the guy was talking about some kind of fisherman!!

[Pace Cap'n]

Now what the hell is a "pita".

Think "acronym".

[hurrikane]

Oh Dave,

you are sooooooooo right on my friend

[GameTheory]

Not much further.
SNIP Joe, I will stipulate you any math you want me to stipulate, just not your general conclusions.

So whatever the number of races you say it is, we'll go with that. My question is why should I care? Since I don't see any particular point in proving what you want to prove, why should I care how difficult it is or how long it might take?

First you've got to tell us WHY it is important to attempt to prove it at all.

[GameTheory]

Joe,

I need to tell you that you are becoming a bit of a pita here.

Whenever a thread goes in a direction that might lead to something useful for someone you jump in to say why none of that makes any sense.
It would actually be somewhat refreshing if Joe have us some WHYs instead of just a collection of irrelevant numbers. I want to know why he thinks this stuff is important.

[Kreed]

Formula ... now get real, I WAS talking of an unbiased coin. But even not,
the Pr's are FIXED & KNOWABLE in advance. I think coin tosses have
obscured Pr knowledge, not advanced it, since after solving a series of coin
problems (or card probs too) many think they now know Pr. But the concept
of coins transcend Where the Coins are being tossed -- it doesn't matter if
your tossing them in Times Square or Iraq. Yet, all the ideas that matter do
not involve such simplicity. Your ideas are clear but I think they're a start --- by themselves they have no predictive value. BTW, is a 3:1 horse, in a race
where thats the Fav, the same as a 3:1 with two others of lower odds?
I don't think so. No matter how fine your increments are, they are always
out of context. Yet, I do think that if you take ALL 3:1's (and lower) you
will find large overlap in how they look.

[formula_2002]

First you've got to tell us WHY it is important to attempt to prove it at all.


I thought that was obvious.
It is an analysis to help me determine the value of playing horses. Not just for me, but anyone.

Why would I continue playing a game where (for example)
I could not win 27.5% of 3-1 horses for a 10% profit.

Lets say I would be happy just to break even playing 3-1 horses, all I would need to do is win 25% of 800 races.

These are the thing people need to prove to themselves.
Some see the need to do so, some not.

Does it help prove if people can beat the take-out. I would say it does.
Can you find enough races to play and have that success, I think not.

[sjk]

Joe,

I don't know anyone who restricts their play to 3-1 bets. What you need is to analyze performance based on an AVERAGE odds amount (say 10-1). Happily, the formula on your web page applies perfectly well in this circumstance. Thanks for making it available.

You have suggested that it is important to look at separate odds ranges. Is this based on any accepted principal of statistics? Can you offer a reference (statistics, not handicapping) so that we can understand this? As I pointed out earlier in the thread this flies in the face of how statistics is used every day where an analysis of the ENTIRE dataset is of utmost importance.

The reason that I could not (any I suspect this applies to others as well) do a separate analysis of the 3-1 bets is that while I know which winners paid 3-1, I haven't a clue which one of my losers would have paid 3-1. Who would keep this sort of detailed records on losers (if you bet a lot of exactas it would take considerable amount of time) when taking the time to do so would get in the way of betting the next race? If you bet in the blind pools how would you even know what the losers would have paid?

Many of us are doing just fine at this. You might be money ahead to devote as much energy (and use of statistical anaalysis) to your handicapping as you have devoted to trying to prove that we can't do what we do all the time.

[cj]

Why would I continue playing a game where (for example)
I could not win 27.5% of 3-1 horses for a 10% profit.


This from the guy that recommends playing craps?

[Exactaman]

Can you find enough races to play and have that success, I think not.

certainly I can. why can't I? that's crazy. most can't but a few surely can. nothing could be more sure than that.

[JustMissed]

the article certainly describes one approach to value betting. i don't agree with everything there and one or two things are perhaps not clearly explained, but it is a good enough example. fwiw my method is quite different, surely there are many.

Exactaman, I few things are not clear to me also. I want to find a few more examples of VALUE PLAYS before I come up to a final conclusion.

In the interim, would you mind answering a few questions for me? I have copied gordon pines first race example for your reference and hope anyone else would expound also.

1. How do you calculate the BET ODDS. In the example they are 1.5,1.5,2 & 2 odd points higher than the respective FAIR ODDS.

What is the purpose of the BET ODDS?

2. Did you notice that the only UNDERLAY was the favorite, #17 Point Givens.The bet was constructed using the favorite, 2nd favorite(not a contender) and the other 3 horses which were OVERLAYS(also long odds).

I think Inside the Pylons mentioned this fact in another thread. The fact that if your selected contenders include the Favorite, then the other contenders are more likely to be OVERLAYS.

Common sense tells me this is true. If you selected them as a contender then you are not going to assign them a high FAIR ODDS are you? If one of your contenders is the Favorite then the others are almost all likely to be OVERLAYS so why go to the trouble to make a line?

3. In the exacta plays, he used very common tickets-the favorite and second favorite over and under the long odds runners. Every dog player on the planet knows this play by heart and you certainly don't need an odds line to come up with this one.

Besides that, the tote board gives a running tally of exacta payoffs so you don't need a line to figure value on those plays anyway, just look up at the tote.

4. Did you notice that with the Exacta plays,except for the box, the favorite and co-favorite were used with the 3 overlays. In effect, 2 out of 5 runners were OVERLAYS, or 40%.

I would suspect that some players are under the false impression that they should not play any runner that is not an OVERLAY. Obviously that would be wrong thinking. Had any of Gordon's exactas hit they would have been boxcar payoffs even though they contained 40% UNDERLAYS.

Anyway, I won't bore you any further but hope you will look this stuff over and let me know what you think.

JM




Form netcapper.com:
"Using the betting line I posted for the Kentucky Derby in Capper’s Corner the night before the race (assuming $20 win/place/show bets and $5 exacta bets) here’s how this betting method worked

Fair Bet Track
P# Horse Odds Odds Odds
3 Balto Star 5/2 4/1 8/1
8 Congaree 7/2 5/1 7/1
17 Point Given 4/1 6/1 9/5
16 Monarchos 7/1 9/1 10/1

$20 Win 3,8,16 = $60
$20 Place 3 = $20
$5 Exacta, 18 combinations = $90
(3+8+16 with 10+17, 10+17 with 3+8+16, and 3-8-16 box)

Total Bet $170
Collected $230
Profit $60"

[GameTheory]

I thought that was obvious.
It is an analysis to help me determine the value of playing horses. Not just for me, but anyone.

Why would I continue playing a game where (for example)
I could not win 27.5% of 3-1 horses for a 10% profit.

Lets say I would be happy just to break even playing 3-1 horses, all I would need to do is win 25% of 800 races.

These are the thing people need to prove to themselves.
Some see the need to do so, some not.

Does it help prove if people can beat the take-out. I would say it does.
Can you find enough races to play and have that success, I think not.

A) Even if you do prove it, you haven't proved anything because you can't prove that you can DO IT AGAIN. It is a dynamic game. Your methods may work, and then stop working. The public may bet on this, and then they bet on that. (The public goes through cycles in what they emphasize in their betting -- take a meet that lasts 6-8 weeks and track ROI for each factor on a weekly basis. Also track the ROI for each factor across all tracks on a weekly basis for a whole year. You'll see certain factors maintain similar winning performance but the ROI goes up or down in a cycle. Other factors change both winning performance and ROI.)

B) Statistical signficance has little to do with practical importance.

C) You can only prove it by analyzing the betting records of successful bettors or by becoming a successful bettor yourself. Since you presumably don't have access to the records of others, and you don't put any time or energy into becoming successful yourself, you will never prove it.

D) You insist on this incremental odds stuff, which only makes it harder. Why not try what sjk suggests and analyze the whole dataset? Even if you think that is not good enough, it will nevertheless be a start. If you can't prove significance on the dataset as a whole first, then you're not going to be able to prove it in all the incremental ranges anyway. So prove it on the whole set, and then see if it holds up in the subsets.

E) Even better, just give it all up and as Dave suggests stop annoying people who are actually trying to learn something useful.

[Exactaman]

Anyway, I won't bore you any further but hope you will look this stuff over and let me know what you think.

no one bored here jm, could talk about this stuff for hours.

in fact, you mentioned a few of the main things i disagreed with/thought unclear.

1. How do you calculate the BET ODDS. In the example they are 1.5,1.5,2 & 2 odd points higher than the respective FAIR ODDS.

What is the purpose of the BET ODDS?

fair odds are the horse's handicapped probability of winning. bet odds are the tote odds the bettor demands as a minimum needed to make a bet. most value bettors will insist on higher bet odds than fair odds, as a margin of error, or a means of narrowing down the number of bets and concentrating on higher value plays, take your pick.

how he calculated his bet odds is not clear. I for one use a straight percentage.

2. Did you notice that the only UNDERLAY was the favorite, #17 Point Givens.The bet was constructed using the favorite, 2nd favorite(not a contender) and the other 3 horses which were OVERLAYS(also long odds).

I think Inside the Pylons mentioned this fact in another thread. The fact that if your selected contenders include the Favorite, then the other contenders are more likely to be OVERLAYS.

Common sense tells me this is true. If you selected them as a contender then you are not going to assign them a high FAIR ODDS are you? If one of your contenders is the Favorite then the others are almost all likely to be OVERLAYS so why go to the trouble to make a line?

i think if you say "i have four equal contenders, i will bet the ones that are not the favorite" that is a form of value betting. i think the equation is not always that simple though. one may grade contenders unequally. favorites may be heavy or slim. there may be more or less contenders in a race--in my system, in a 10-horse field, a contender might have fair odds of as high as 9/1. i want to say that in a 17-horse field that would be even higher, but as a harness player i don't have any experience betting such large fields. anyway, i could go on. given these factors, it is not a given for me that non-favored contenders will be overlays. the article is also not clear if there were other contenders in his line that were underlays. i guess invisible ink wasn't :)

3. In the exacta plays, he used very common tickets-the favorite and second favorite over and under the long odds runners. Every dog player on the planet knows this play by heart and you certainly don't need an odds line to come up with this one.

Besides that, the tote board gives a running tally of exacta payoffs so you don't need a line to figure value on those plays anyway, just look up at the tote.

i disagree with his approach, though it may work for him and statistical analysis could possibly show it to be sound. but now we're talking exactas, this is my territory :)

let me explain what i do now, you could even use this as an example of value betting if you are so inclined :) i make a win line and compute a fair-odds price for every exacta in Excel with a straightforward formula. (note: this means in fact that my win line is not actually the chance of winning but the average of the chances of winning or coming second. i do adjust somewhat on the principle that horses with a high win chance will not have as high of a chance of coming second.) i pull the tote odds into excel in real-time, which points out the overlays, which i bet. i want exacta prices 50% higher than the fair price before i bet. for each additional exacta i bet, i increase the price i require by a corresponding amount.

so i have some experience with this. for one thing, in my experience all the exactas with underlayed favorites are underlays as well, except with very highly overlayed horses. i would NEVER throw in the exacta with the fave as a rule. i look at each combination individually.

i disagree with the latter part of your statement, if you have a 4 or 5 contenders going in a 10 horse field (as i often do--and i will throw in some a bit higher than 9/1 for second) and bet in medium sized pools where much of the money comes in late making a betting decision on which of 20 exactas to bet with 0 min left is not as easy as just looking at the tote :)

4. Did you notice that with the Exacta plays,except for the box, the favorite and co-favorite were used with the 3 overlays. In effect, 2 out of 5 runners were OVERLAYS, or 40%.

I would suspect that some players are under the false impression that they should not play any runner that is not an OVERLAY. Obviously that would be wrong thinking. Had any of Gordon's exactas hit they would have been boxcar payoffs even though they contained 40% UNDERLAYS.

true, certainly many overlayed exactas will contain horses that are underlays to some or more degree in the win pool. as i said i really believe each combination needs to be looked at individually. there could be merit to a simplified approach like the author mentions, i can't really say. it will certainly be a lot more workable than mine unless one has a computer handy :)

[Valuist]

Formula-

Why are you even interested in race track betting if you are so convinced it can't be beaten?

[formula_2002]

Formula-

Why are you even interested in race track betting if you are so convinced it can't be beaten?


I'm not interested in betting, but it's fun to play with the numbers.

[formula_2002]

A) E) Even better, just give it all up and as Dave suggests stop annoying people who are actually trying to learn something useful.

We at least we have both stated our case.

The only exception I take to your post is as quoted.

I think there are a lot of people I have not annoyed, and those that may have been annoyed, certainly have not gone away without gaining valued information, even if it was limited to your responses (as wrong as I think they are).


Now all they have to do if to figure out how all this fits into their schemes.

[formula_2002]

This from the guy that recommends playing craps?

CJ, If I though I could make money playing horses I would have hundreds of thousands of dollars avialable to play with.

But, since I can not win at horse race betting, I'll take a few thousand a year and spend it in a recreational manner playing craps, and a few $'s on horses.

[formula_2002]

Joe,

I don't know anyone who restricts their play to 3-1 bets.

It's just and example. It's not ment to restrict play to 3-1.



You have suggested that it is important to look at separate odds ranges. Is this based on any accepted principal of statistics? Can you offer a reference (statistics, not handicapping) so that we can understand this? As I pointed out earlier in the thread this flies in the face of how statistics is used every day where an analysis of the ENTIRE dataset is of utmost importance.


Its just that Quirin, Fabricand and about a few dozen others in displines out side of horse racing have done it the same way.

You might be money ahead to devote as much energy (and use of statistical anaalysis) to your handicapping as you have devoted to trying to prove that we can't do what we do all the time.

I spent the last 10+ years doing computer studies of just that. And here I am

[JustMissed]

I don't think the KD is the best race to use for examples of betting techniques as the odds are all over the board, you have a huge fields and the risk-reward ratio is so high one tends to make bets that they normally would not make.

I only used this as an example of Value Betting as it was the only one I could find. I did not intend to question Gordon's betting techniques and like I said that would be unfair to do so on that race anyway.

I wish someone would post some real life examples for us to look at. Not made up ones put in a book but some real examples so we could really examine them to determine if making an oddsline was really that beneficial.


Sounds like you have your computer system set up great to handle the value betting situations.

JM

[Exactaman]

I wish someone would post some real life examples for us to look at. Not made up ones put in a book but some real examples so we could really examine them to determine if making an oddsline was really that beneficial.

the thing is, only long-term results can really show the benefit. and you have to take people's word on those.

Sounds like you have your computer system set up great to handle the value betting situations.

JM

oh it's amazing what these things can do these days :) thanks, i do take a bit of pride in my excel baby.

[GameTheory]

We at least we have both stated our case.

The only exception I take to your post is as quoted.

I think there are a lot of people I have not annoyed, and those that may have been annoyed, certainly have not gone away without gaining valued information, even if it was limited to your responses (as wrong as I think they are).
Trust me, it gets annoying when you say the same thing over and over and over in evey post you make for YEARS. If you only have one thing to say, just say it once, maybe even a few times, but you butt in every thread you can WITH THE SAME THING OVER AND OVER.

[formula_2002]

Trust me, it gets annoying when you say the same thing over and over and over in evey post you make for YEARS. If you only have one thing to say, just say it once, maybe even a few times, but you butt in every thread you can WITH THE SAME THING OVER AND OVER.


Well then it is obvious.
I have discovered the Total Unified Theory of horse racing.
It would seem that all questions can be solved with the one, singular, and simple Theory... ;)

SOLVE FOR NH IN THE EXPRESSION

2.5 = (NW-EW)/(SQRT(EW*(1-(EW/NH))))

WHERE EW=((1/(odds+(1+BREAKAGE))/ (1/(1-track take))

EW= EXPECTED WINNERS
NH= NUMBER OF HORSES TO PROVE 99% CONFIDENCE

Now if all post would include the phrase;

“not to the exclusion of Formula_2002’s “ Total Unified Theory of horse racing””
I can limit my annoying input. ;)

[GameTheory]

Well then it is obvious.
I have discovered the Total Unified Theory of horse racing.
It would seem that all questions can be solved with the one, singular, and simple Theory... ;)

SOLVE FOR NH IN THE EXPRESSION

2.5 = (NW-EW)/(SQRT(EW*(1-(EW/NH))))

WHERE EW=((1/(odds+(1+BREAKAGE))/ (1/(1-track take))

EW= EXPECTED WINNERS
NH= NUMBER OF HORSES TO PROVE 99% CONFIDENCE

Now if all post would include the phrase;

“not to the exclusion of Formula_2002’s “ Total Unified Theory of horse racing””
I can limit my annoying input. ;)I believe I can speak for the board in saying we all unamiously agree that everything we post is so stipulated and you can go ahead and ASSUME that is the case, thereby nullifying the need for you to comment -- ever.

[formula_2002]

I believe I can speak for the board in saying we all unamiously agree that everything we post is so stipulated and you can go ahead and ASSUME that is the case, thereby nullifying the need for you to comment -- ever.

I can save you the trouble. I've included your unamious agreement in my signature

[Buckeye]

This 3-1 Crap (a Casino Game) is simply a hoax.

Who thinks there are many 3-1 overlays?

Not me.

How about between 9 and 11-1 "K" rank of one since August 2003:

Total Bets 829
WIN PLACE SHOW
Total Amount Bet 1658.00 1658.00 1658.00
Wins 88 163 270
Pct. 10.62% 19.66% 32.57%
Amount Won 1902.90 1459.80 1454.80
Profit/Loss 244.90 -198.20 -203.20
Pct Profit Loss 14.77% -11.95% -12.26%
Avg. Payout 21.62 8.96 5.39

That doesn't even BEGIN to address the possible profitable completely mechanical combinations of factors!

Moreover, it doesn't begin to skim the surface of handicapping genius which could also beat the take.

Back to the drawing board, it's 2005.

[formula_2002]

This 3-1 Crap (a Casino Game) is simply a hoax.

Who thinks there are many 3-1 overlays?

Not me.

How about between 9 and 11-1 "K" rank of one since August 2003:

Total Bets 829
WIN PLACE SHOW
Total Amount Bet 1658.00 1658.00 1658.00
Wins 88 163 270
Pct. 10.62% 19.66% 32.57%
Amount Won 1902.90 1459.80 1454.80
Profit/Loss 244.90 -198.20 -203.20
Pct Profit Loss 14.77% -11.95% -12.26%
Avg. Payout 21.62 8.96 5.39

That doesn't even BEGIN to address the possible profitable completely mechanical combinations of factors!

Moreover, it doesn't begin to skim the surface of handicapping genius which could also beat the take.

Back to the drawing board, it's 2005.

Well if you buy into my confidence logic, the numbers you posted produce a confidence level well beyond 99%.


Sounds to me like you have some thing very significant..

Just one question. How many 9-1 through 11-1 horses won in all the races you handicapped ( Whether you played the race or not). Also knowing the average payout would help

[Tom]

Statistically explain to me how this equation can be true:

Amount $ returned > Amount $ Wagered.

The equation is true and has been for years.

Playinfg with numbers is fun, but you should try playing with....money. :eek:

[Exactaman]

Statistically explain to me how this equation can be true:

Amount $ returned > Amount $ Wagered.

The equation is true and has been for years.

Playinfg with numbers is fun, but you should try playing with....money. :eek:

not statistically here, but here's my explanation (not that I thought you needed one Tom :))

paramutuel betting on horses is a game of skill--someone who does research into horses' performance plays against and will likely outperform one who does not. A few determined people are bound to refine their skill until they reach the logical goal of making a profit.

[cj]

Formula insists you can't beat the takeout. I know it can be done. Why? It is quite simple. Many, many people do MUCH worse than the track takeout. They are not very good handicappers, and even worse bettors. Probably even some good handicappers who are very poor bettors. I would assume every member of this board knows at least a few of these people.

A lot of people will lose whatever they bring to the track. 100%, not the track take.

[formula_2002]

Statistically explain to me how this equation can be true:

Amount $ returned > Amount $ Wagered.

The equation is true and has been for years.

Playinfg with numbers is fun, but you should try playing with....money. :eek:


what you posted is not an equation..

[Dave Schwartz]

CJ,

Excellent logic.


Dave Schwartz

[formula_2002]

Formula insists you can't beat the takeout. I know it can be done. Why? It is quite simple. Many, many people do MUCH worse than the track takeout.
A lot of people will lose whatever they bring to the track. 100%, not the track take.

In the long run, no bettor, even me, can lose more then the track take-out (along the long to sort priced odds biase).
If I were to bet $100 per race for each of 100 races, I would probably show a loss of $1500. (Craps would cost me $141 which would most likely be off-set by a free dinner)

Yes, there may one who will lose their entire bank-roll ($100,000) at the track in one one day

And, if here were to bet it on one race, I could not imagine how much money that one race would atrack from more knowledgeable players, such that the odds become more equalized to reality.

A fool and his money will soon be parted, but only at the rate of the track take-out.

I have to look into that.

[cj]

In the long run, no bettor, even me, can lose more then the track take-out (along the long to sort priced odds biase).
If I were to bet $100 per race for each of 100 races, I would probably show a loss of $1500. (Craps would cost me $141 which would most likely be off-set by a free dinner)

First, I don't really care about craps, its a game for suckers. Might as well play the lottery. Now, if you really believe no bettor can lose more than the track take, you are truly living in lala land. You ever heard of Gambler's Anonymous? I've have known people (not one, not a few, but many) that can cash a $2500 trifecta on Wednesday, and on Saturday, they are trying to borrow money again because they've lost it all AT THE WINDOWS by then. How do they do this? They increase their bet size, and bet more, lose, then chase with more!


Yes, there may one who will lose their entire bank-roll ($100,000) at the track in one one day

And, if here were to bet it on one race, I could not imagine how much money that one race would atrack from more knowledgeable players, such that the odds become more equalized to reality.

This is not the person I'm talking about. They don't lose it in one day. Most have jobs, so their "bankroll" is equivalent to their paycheck, or some fraction of it that is left after paying the bills. You don't hear it talked about much, but who do winning players think they are making their profits from?


A fool and his money will soon be parted, but only at the rate of the track take-out.

The good news is, they find ways to replenish the bankroll! And many lose more than the track takeout. This is a false assumption on your part, very false. You are implying that betting horses is purely a game of chance where each individual will lose the track take over time. This is no more true than sitting at a Poker Table and assuming if you play long enough, everyone will break even. The worst players might win a hand or two, but over time, they will leave broke, and it is not by chance.

[formula_2002]

First, I don't really care about craps, its a game for suckers. Might as well play the lottery. Now, if you really believe no bettor can lose more than the track take, you are truly living in lala land. You ever heard of Gambler's Anonymous? I've have known people (not one, not a few, but many) that can cash a $2500 trifecta on Wednesday, and on Saturday, they are trying to borrow money again because they've lost it all AT THE WINDOWS by then. How do they do this? They increase their bet size, and bet more, lose, then chase with more!


This is not the person I'm talking about. They don't lose it in one day. Most have jobs, so their "bankroll" is equivalent to their paycheck, or some fraction of it that is left after paying the bills. You don't hear it talked about much, but who do winning players think they are making their profits from?


The good news is, they find ways to replenish the bankroll! And many lose more than the track takeout. This is a false assumption on your part, very false. You are implying that betting horses is purely a game of chance where each individual will lose the track take over time. This is no more true than sitting at a Poker Table and assuming if you play long enough, everyone will break even. The worst players might win a hand or two, but over time, they will leave broke, and it is not by chance.


CJ, I would say you are correct about "I've have known people (not one, not a few, but many) that can cash a $2500 trifecta on Wednesday, and on Saturday, they are trying to borrow money again because they've lost it all AT THE WINDOWS "

My point was that their "excesses" or even guys like you that may be "wrong" in a race, would not contribute to the benifit of the more knowledgeable players.

The "knowledgeable players" would see the excess in the odds board and make the proper adjustment to the odds.

But since there are many "knowledgeable players" they would adjust the odds to the point where they would have no benefit in the other guy's excess.

Craps, or any gane of chance, for many is a suckers game, only if they dont understand the math of the game.

[Exactaman]

My point was that their "excesses" or even guys like you that may be "wrong" in a race, would not contribute to the benifit of the more knowledgeable players.

The "knowledgeable players" would see the excess in the odds board and make the proper adjustment to the odds.

But since there are many "knowledgeable players" they would adjust the odds to the point where they would have no benefit in the other guy's excess.

Craps, or any gane of chance, for many is a suckers game, only if they dont understand the math of the game.

That's clearly wrong. In your example, more knoweledgable players will be able to benefit at the same odds.

[cj]

Craps, or any gane of chance, for many is a suckers game, only if they dont understand the math of the game.

You'll never understand racing. You seem to think the tote is infallible, which it is not. Over the long term, it is very accurate, but as I've said many times, racing is a game of individual events, and the long term doesn't prove a whole lot. First, you said "suckers" will lose only the track take, which implies everyone will lose the track take. Now, you say, well, maybe they will lose more, but everyone will adjust to this. If some people are losing MORE than the track take, which only a complete moron would say they are not, by definition, some people MUST be doing better than the track take. To further this, some people must be doing better by enough to make a profit. I would think this can be proven mathematically as well, but I'll leave that to you. Let's say that 15% of the players lose 50% of their money. How is this money redistributed among the remaining 85%? We'll never know the answer, but there is no way that none of those 85% are making a profit. It's a game of skill, some will be better than others.

As for craps, its a negative expectation game. I understand the math, so I KNOW its a sucker's game. If you believe otherwise, I'll set up a table in my basement, and I invite you to come and play...with Vegas house rules of course.

[JustMissed]

You may have a serious error with your takeout analysis.

The takeout is already deducted from the toteboard amounts/expected payoffs and therefore TAKEOUT should not be considered when determining individual overlays, edges or profits. TAKEOUT IS ALREADY BAKED IN THE BREAD.

When you select a 2-1 horse for example. The track is guaranting you they will pay you $6 if you selection wins.

They are not saying, I repeat, they are not saying that if your selection that has a 33.3% chance of winning does in fact win, we will pay you $6 less the track take.

The player is guaranteed to collect the tote odds(or often more)but never less than the posted expected payoffs.

In effect, the losers are paying the take, not the winners.

Even Fabricand when he published his study of 93,001 horses showed that the percentage of winners were 71.3% in the 3.10 Mutuel Price Range, he did not say "3.1 Probablity Range before Deducting Takeout".

Probability of winning is not the same as expected mutuel price.

You should get your terminology correct before you continue any of the studies.

JM :D

[Jeff P]

posted by Formula 2002 -In the long run, no bettor, even me, can lose more then the track take-out

Joe, nothing could be further from the truth!!!

The following table shows what I have for all starters sitting in my database for dirt races so far in 2005. It is broken out by rank for Bris Prime Power. A $2.00 win bet on every starter returns somewhere in the neighborhood of 77 cents on the dollar (a little less than track take.) A $2.00 win bet on every horse ranked in the top three of Prime Power returns a little higher than track take. And a $2.00 win bet on those horses ranked low in Prime Power returns nowhere near the track take.

It appears that a notoriously bad handicapper, whose selections consistently rank very low in Prime Power, might lose more than track take.

Conversely, it appears that just by doing a little handicapping, the player can slant long run expectations more to his liking- dosn't it?

Now prime power as a single factor isn't going to turn the corner for you or any other player. It's a causal factor (it has predictive utility) but far too many people see it each day for it to be useful as a single factor to turn a profit.

But it does indicate that a little handicapping can make a difference. All you need to make a profit are some good causal factors that aren't visible to or aren't understood by the betting public.

Data Window Settings:
Data Window Settings:
999 Divisor
Dirt (All*) Distance: (All*) (From Index File: F:\2005\Q1_2005\pl_profile.txt)


Data Summary Win Place Show
Mutuel Totals 16089.90 15710.70 15433.40
Bet -20774.00 -20774.00 -20774.00
Gain -4684.10 -5063.30 -5340.60

Wins 1257 2496 3691
Plays 10387 10387 10387
PCT .1210 .2403 .3553

ROI 0.7745 0.7563 0.7429
Avg Mut 12.80 6.29 4.18


By: Prime Power Rank

Rank Gain Bet Roi Wins Plays Pct
1 -502.50 2500.00 0.7990 353 1250 .2824
2 -247.20 2484.00 0.9005 266 1242 .2142
3 -244.50 2558.00 0.9044 185 1279 .1446
4 -377.80 2546.00 0.8516 159 1273 .1249
5 -451.10 2552.00 0.8232 109 1276 .0854
6 -557.30 2324.00 0.7602 71 1162 .0611
7 -982.80 2042.00 0.5187 41 1021 .0402
8 -584.80 1570.00 0.6275 33 785 .0420
9 -441.30 1038.00 0.5749 15 519 .0289
10 -286.70 622.00 0.5391 11 311 .0354
11 -146.80 366.00 0.5989 5 183 .0273
12 117.80 162.00 1.7272 3 81 .0370
13 -8.00 8.00 0.0000 0 4 .0000
14 -2.00 2.00 0.0000 0 1 .0000

[formula_2002]

The takeout is already deducted from the toteboard amounts/expected payoffs and therefore TAKEOUT should not be considered when determining individual overlays, edges or profits. TAKEOUT IS ALREADY BAKED IN THE BREAD.
JM :D



If you are calculating the total number of horse required for confidence, you must determine the "expected # of winners",

So
At a 17% take-out track
tote board odds 3-1
that horse should win 20.75% to the time

At a 15% take out track
tote board odds 3-1
that horse should win 21.25%
(unadjusted for breakage)
----------------------------------------------------


If you win more then 20.75% of the time at a 15% take-out track you have been playing horses that have a greater chance of winning at the defined odds of 3-1.

If you win more then 21.25% of the time at a 17% take-out track you have been playing horses that have a greater chance of winning at the defined odds of 3-1.

These last two conditions can be defined as "overlay" plays.
I won more then was expected.
--------------------------------------------------
You must also know your actual # of winner and the total number of horses played in 3-1 odds group.

A win % of 27.5 at 3-1 would give me a 10% profit at any track, regardless od the take-out.
That would represent the Number Of Actual winners, but I have to compare that to the number of expected winners, which can only be determined by using the take-out.

[formula_2002]

posted by Formula 2002 -In the long run, no bettor, even me, can lose more then the track take-out

Joe, nothing could be further from the truth!!!



I agree, and I could not have posetd a clarification fast enough, so I replied to cj as follows;;

CJ, I would say you are correct about "I've have known people (not one, not a few, but many) that can cash a $2500 trifecta on Wednesday, and on Saturday, they are trying to borrow money again because they've lost it all AT THE WINDOWS "

My point was that their "excesses" or even guys like you that may be "wrong" in a race, would not contribute to the benifit of the more knowledgeable players.

The "knowledgeable players" would see the excess in the odds board and make the proper adjustment to the odds.

But since there are many "knowledgeable players" they would adjust the odds to the point where they would have no benefit in the other guy's excess.

Craps, or any gane of chance, for many is a suckers game, only if they dont understand the math of the game.

[JustMissed]

Joe, I see where you are going wrong and why everyone is having so much trouble understanding your false analysis.

You are mixing apples & oranges.

A horse's expection of winning(or probability of winning as some say), has no relationship to the expected mutuel payoffs(or the tote board as some say). This is where you are going wrong.

If there is a horse race without any mutuel betting, each horse has an expectation of winning but there are no expected payoffs are there. A take is not considered since there is not a betting pool to take anything out of.

To prove I am correct, I looked at a results chart for Tampa Bay Downs. I found a horse that won a race with final odds of 2-1. The $2 win ticket paid $6. The $6 was 2 x $2 bet = $4 + a $2 return of the original bet. No takeout deducted from the advertised price you see. They paid what they guaranteed that they would pay. Whoever bet that horse made a $4 profit, independent of their estimation of the horse's expection to win.

Tampa has an 18.9 percent takeout. Where is the takeout?

Even Mr. Fabricand was very specific to identify Mutuel Prices and not being the same thing as a horse probability of winning.

Joe, not to be rude, but if you are making these type gross errors in your betting studies--just maybe you are making similar errors with you handicapping and that is why you continue to lose.

You know we have several member in the NYC area. Maybe you could get one of them to go to the track with you and look over your handicapping to see what you are doing wrong. Wouldn't hurt?

Good luck to you.

JM
:)

[cj]

I agree, and I could not have posetd a clarification fast enough, so I replied to cj as follows;;



But, you haven't replied to this. I'm dying to hear your explanation!

First, you said "suckers" will lose only the track take, which implies everyone will lose the track take. Now, you say, well, maybe they will lose more, but everyone will adjust to this. If some people are losing MORE than the track take, which only a complete moron would say they are not, by definition, some people MUST be doing better than the track take. To further this, some people must be doing better by enough to make a profit. I would think this can be proven mathematically as well, but I'll leave that to you. Let's say that 15% of the players lose 50% of their money. How is this money redistributed among the remaining 85%? We'll never know the answer, but there is no way that none of those 85% are making a profit. It's a game of skill, some will be better than others.

[formula_2002]

Will reply to all some time tonight..
Would you believe work is in the way of my replying. :)

[formula_2002]

A horse's expection of winning(or probability of winning as some say), has no relationship to the expected mutuel payoffs(or the tote board as some say). This is where you are going wrong.

JM
:)

Horses win in accordance with their odds. See Fabricand's "horse Sense" page 34.

I think the first time I saw this king of analysis was in the American Racing Manual back in the early 60's. I think Rowe put it together.
I also have a chart on my web pay That I generated from my own data.

[formula_2002]

Even Mr. Fabricand was very specific to identify Mutuel Prices and not being the same thing as a horse probability of winning.


JM
:)

Take at look at page 31
the expected return (E)=take-out

You should read the entire chapter " "Probability at the races"

What formula do you think I got wrong..
Please no words just formulas.

[formula_2002]

Assuming you are talking about The Bris Prime Power rating, I have posted a ranking report on my web page which is generated from my data base. (Barry Meadow published it last year some time)

The more telling aspect of the Bris Prime Power rating, or just about all other factors is that the rankings always wins in accordance with the odds.

If you have the data, sort on the odds and you will find that ranked horses are impacted by the odds.

Foe instance. A top ranked horse wins more at 1-1 then 2-1.

It appears to me that the public is doing the heavy lifting, not the factor.

If the factor was the "DRIVER" why would it not win the same % accross all odds.

Why would the top rank win, say 50% accross all odds?

I put that question to Bris..The answer did not help me.

[formula_2002]

..... but as I've said many times, racing is a game of individual events, and the long term doesn't prove a whole lot.

I looked at a recent race at AQU

A horse had final odds of 2.45-1.

Normalizing the all the odds, the 2.45-1 horse had a win probability of .2483.

To make a 10% profit, that horse must win .319% of his races. A 0.07068 spread.

To make that profit, you must outperform the public by
.319/.2483 = 128%

Somewhere in that race the public had to make a 28% error for you to make a 10% profit.

How is the math?
If it's ok, I'll go on.

[toetoe]

formy,
28 %age POINTS, but 21.9 %.

[formula_2002]

formy,
28 %age POINTS, but 21.9 %.


?

[toetoe]

You have to distinguish between pctg. and pctg. points. 28 as a percentage of 100 is the same as the # of points -- 28. Now 28 as a pctg. of 128 is 21.9 but the points are based on the original 100 -- 28 pctg. points. So the public has to err only 21.9% more, I guess; not 28% more.

[Exactaman]

certainly over the long term the entire crowd doesn't have to be 28% or even 21% wrong for me to scrape off a 10% profit with my relatively insignificant action?

[chickenhead]

I looked at a recent race at AQU

A horse had final odds of 2.45-1.

Normalizing the all the odds, the 2.45-1 horse had a win probability of .2483.

To make a 10% profit, that horse must win .319% of his races. A 0.07068 spread.

To make that profit, you must outperform the public by
.319/.2483 = 128%

Somewhere in that race the public had to make a 28% error for you to make a 10% profit.

How is the math?
If it's ok, I'll go on.

Toetoe is right, I would say if you make a 10% profit betting 2.45-1 shots over time, the public must have under-estimated those horses you bet on by an average of 22.2% (.2483/.319 = Public Odds are 77.8% of "Actual").

[cj]

Again Formula, you have completely ducked my question.

Here was the premise. First, you said everyone loses the track take, in case you don't remember. Then, you recanted, and said I was right, some people lose more than the track take. Some will lose A LOT more than the track take.

So, if this is true, some people MUST win more than the track take, and some even will win enough to make a profit. Where am I wrong?

Please, no more pointless arguments about some 2.45 to 1 shot, which you'll never know was 2.45 to 1 until you see the chart a half hour after the race is over, and has absolutely nothing to do with my premise.

To repeat, so I am clear, as you seem to have trouble answering my question...If some people are losing more than the track take, why can't some people beat the track take? And of course these people exist, its a mathematical certainty. So why can't some beat it enough to show a profit?

[formula_2002]

Again Formula, you have completely ducked my question.

Here was the premise. First, you said everyone loses the track take, in case you don't remember. Then, you recanted, and said I was right, some people lose more than the track take. Some will lose A LOT more than the track take.

So, if this is true, some people MUST win more than the track take, and some even will win enough to make a profit. Where am I wrong?

Please, no more pointless arguments about some 2.45 to 1 shot, which you'll never know was 2.45 to 1 until you see the chart a half hour after the race is over, and has absolutely nothing to do with my premise.

To repeat, so I am clear, as you seem to have trouble answering my question...If some people are losing more than the track take, why can't some people beat the track take? And of course these people exist, its a mathematical certainty. So why can't some beat it enough to show a profit?


What I have been demonstrating is the ability of the public to BEAT the take-out and the improbability of them to prove it, and therefor, most likely not to, because of the vast number of horses they would have to handicap.

You truely want to consider how much better you have to do to beat the public. It would help my case :)
-------------------------------------------------------

I'll leave it up to you if there is a difference between "clarification" and recanting".

[formula_2002]

Toetoe is right, I would say if you make a 10% profit betting 2.45-1 shots over time, the public must have under-estimated those horses you bet on by an average of 22.2% (.2483/.319 = Public Odds are 77.8% of "Actual").

If I made $3190 a week and you made $2483 a week, I would be making $707 more then you (28.47%) more then you.

If you went to your boss and said you wanted to make as much as me, you had better ask him for a 28.47% raise, because I'm making 28.47% more then you (I'm doing 28.47% better then you) (1.2847 x 2483 = 3190).

[formula_2002]

Worth mentioning

"Although the lack of certainty in the scientific approach is bothersome to many (and welcomed by others), no other method has as yet demonstrated any measure of success in the search for knowledge and truth"

Fabricand,"Horse Sense" page 12.

With the above in mind, I have tried to demonstrate my position on the ability of the bettor to beat the take-out.

I find no arguments to the contrary worth mentioning.

[cj]

What I have been demonstrating is the ability of the public to BEAT the take-out and the improbability of them to prove it, and therefor, most likely not to, because of the vast number of horses they would have to handicap.

You truely want to consider how much better you have to do to beat the public. It would help my case :)
-------------------------------------------------------

I'll leave it up to you if there is a difference between "clarification" and recanting".

Of course the public, as a whole, cannot beat the take. They are the take, obviously. I submit, yet again for you, that some people lose much more than the takeout, so quite obviously, some MUST beat the take.

You still don't answer and try to change the subject. It's almost comical, but in reality, it just means you don't have an answer.

[formula_2002]

. It's almost comical, but in reality, it just means you don't have an answer.

I can only defend the math.. So if you have a question about that, I'm sure I can clear it up for you.

If you have accomplished what I have laid out in the math, that's wonderful.
If you subscribe to some other intellegence, that is also wonderful.

I would advise anyone who thinks they are beating the game to test their results with the math that I have posted here, else they can go with the "I know this guy who owns a million dollar house" logic.

[cj]

Math doesn't get any simpler than this:

If some lose more than the takeout, some MUST lose less than the track takeout. Why is this wrong, in math terms if you must?

[Kreed]

I think Formula reminded us what we DO KNOW, but underplay: the PUBLIC is
VERY GOOD in assessing the Pr of winners. But that "fact" seems more important than what it is. Because the PUBLIC is a FICTIONAL BETTOR; YOU
& ME can assess risk (potentially) better than the Public. It's risk assessment.
It's what companys in the business of making money do everyday ... some more so & more immediate than others. What about reductions in TAKEOUT?
What about REBATES? The first thing cappers need is a SOLID BETTING
SYSTEM ... (that alone takes common sense & some science). Then a BETTING PLAN (this is trickier and requires more math but is easily doable. Then, Record keeping. Think how this game will change if USA racing adopts the Hong Kong system of giving us each horses ACTUAL TIME at many calls in the race. And if there is truth in a physiologist's saying that he developed a hand held device that could measure a horse's heart action (correlated to fitness). Ah, the possibilities ..... So Fabricand only showed what others knew: the math seems to indicate that "THE PUBLIC" makes ODDS that are close to reality, but that finding sorta says nothing.

[formula_2002]

Math doesn't get any simpler than this:

If some lose more than the takeout, some MUST lose less than the track takeout. Why is this wrong, in math terms if you must?


If he keeps on betting, in the long run he will lose to the tune of the track take-out (along the long to short odds biase). Why, because track odds behave as natural odds
where # winners/#expected winners =1
where % expected winners =(1/(odds+1+breakage))*(1-track-out).

However, I'll give you a case where mathematically, someone will always lose more then the track take-out.
(at least prior to computer betting).

He simply bets say, $x on a long shot at 30 seconds to post such that the long shot odds are driven way down, to say the favorite, driving up all the other odds to overlays.
The other bettors, unable to react to the odds change, can pump no more money into the pool.

It you saw your pick go from 3-1 to 10-1, would you bet more money on him? (assuming your betting line was say 5/2)

If you think these situations or situations close to it exist, then you should continue betting.

In the example I posted about 2.45-1 odds, all you have to do is be 28% more efficient then the public.

If you can follow this kind of person around, you do not have to be as efficient.

[formula_2002]

Fabricand only showed what others knew: the math seems to indicate that "THE PUBLIC" makes ODDS that are close to reality, but that finding sorta says nothing.

If it says sorta nothing, do you think Vegas would cover this proposition bet.

"Pay me even money if in the 3-1 odds range
1000 horse will win 21% +-1.5% of the time."

I may have to do a little more work to fine tune that 1.5% :) And the 21% would have to be adjusted to track take-out.

[chickenhead]

If I made $3190 a week and you made $2483 a week, I would be making $707 more then you (28.47%) more then you.

If you went to your boss and said you wanted to make as much as me, you had better ask him for a 28.47% raise, because I'm making 28.47% more then you (I'm doing 28.47% better then you) (1.2847 x 2483 = 3190).

If I made 80 out of 100 free throws, did I make errors at a rate of 20% or 25%? :p

[chickenhead]

If he keeps on betting, in the long run he will lose to the tune of the track take-out (along the long to short odds biase). Why, because track odds behave as natural odds
where # winners/#expected winners =1
where % expected winners =(1/(odds+1+breakage))*(1-track-out).


Isn't it true that ROI is governed by above only if you are betting randomly?

IF you don't believe that statement is true, then you beleive it is just as difficult to show an ROI of -30% as it is to show an ROI of 0% (-/+15% of takeout)? Have you ever tried this? To handicap your database for maximum negative return? How low can you go, and with what confidence level?

It's a pretty good question I think, I'd be interested to see your results...if you can maintain a low ROI, then it proves the opposite, does it not?

[formula_2002]

If I made 80 out of 100 free throws, did I make errors at a rate of 20% or 25%? :p

No, you made no error, but if I wanted to hire some one to hit 100 out of 100, I would make certain that he was 25% better then you.

[GameTheory]

Isn't it true that ROI is governed by above only if you are betting randomly?

IF you don't believe that statement is true, then you beleive it is just as difficult to show an ROI of -30% as it is to show an ROI of 0% (-/+15% of takeout)? Have you ever tried this? To handicap your database for maximum negative return? How low can you go, and with what confidence level?

It's a pretty good question I think, I'd be interested to see your results...if you can maintain a low ROI, then it proves the opposite, does it not?Actually, it is not true that if you bet randomly you will lose the take. A person choosing one horse per race randomly and betting to win the same amount every time will lose more than the take, and they will do it in the long run, probably with an ROI of about 0.76, depending on the take. (1.0 is breakeven ROI.)

If you always bet on the longest shot in the field you'll lose even more than that, and will maintain an ROI of only 0.56 or so.

This is easy to verify for anyone, even Formula.

You will only lose the amount of the take (roughly) if you bet in proportion to the public. So if you were flat betting, you'd have to bet on more favorites than second favorites, more second favorites than third favorites, etc. (If you always bet on the favorite, you'll lose slightly less than the take.)

But losing more than the take, and doing it long-term, is quite easy.

[Exactaman]

I would advise anyone who thinks they are beating the game to test their results with the math that I have posted here, else they can go with the "I know this guy who owns a million dollar house" logic.

I'll just go with the "I know me" logic :):):) it requires no test

[cj]

But losing more than the take, and doing it long-term, is quite easy.

Since Formula keeps bobbing and weaving and won't answer, if it can be proven that people lose more than the track take, isn't it an absolute certainty that some must beat the track take?

Formula tries to imply that some really bad bettors could lose 100% of their money, yet the others would still somehow miraculously lose the same track take amount. I'm not as smart as he is obviously, so tell me where I'm wrong.

[GameTheory]

Since Formula keeps bobbing and weaving and won't answer, if it can be proven that people lose more than the track take, isn't it an absolute certainty that some must beat the track take?

Formula tries to imply that some really bad bettors could lose 100% of their money, yet the others would still somehow miraculously lose the same track take amount. I'm not as smart as he is obviously, so tell me where I'm wrong.There is one point upon which you are wrong on a technical level, but it doesn't negate your point. Since we are talking about ROI, almost no one actually loses 100%. To lose 100% you have to NEVER cash a bet. Even if you eventually lose all the money you have, your ROI is going to be above 0 if you have ever received anything back from a wager. ROI = Amt returned / Amt bet.

Still, it is very easy to lose more than the track take.

BTW, I've been waiting YEARS for Formula to answer questions even simpler than this one.

[formula_2002]

This is for GT.

How much would you lose if you bet every horse in the race at a 15% take-out track with nickle breakage.?

(lets count the words he uses to reply to this simple question.)

If you get the correct numerical answer, we can go on from there with the answer to some other question raised in the more recent post.

If not I'll just go on without you and you need not get involved. By you can watch :cool:

[JustMissed]

Can't believe you guys are still discussing takeout.

Profitabilty for a horseplayer is like a three legged stool.

Ist leg is strike rate/win percentage

2nd leg is average mutuel price

3rd leg is turnover(amout wagered in a given time frame)

Takeout is already factored into the Mutuel Price and is of no further consideration and certainly has no direct bearing on strike rate or turnover.

Some guy, in order to sell his book, some years ago wrote about track takeout and its effect on expected winners or some such nonsense. It means absolutely nothing and is just another one of those gimmicks to sell books.

Track takeout is just another cost of doing business-no more relevant than the price of admission or the cost of a program.

If you can't be profitable at the stated minimum mutuel prices then you will have to change your strike rate or turnover or find another hobby. Track takeout is fixed and you cannot change it-why worry about it.

JM

[GameTheory]

This is for GT.

How much would you lose if you bet every horse in the race at a 15% take-out track with nickle breakage.?

(lets count the words he uses to reply to this simple question.)

If you get the correct numerical answer, we can go on from there with the answer to some other question raised in the more recent post.

If not I'll just go on without you and you need not get involved. By you can watch :cool:There is no analytical answer because it is not a deterministic question, but you would lose approx. 3-4% more than the take if you were to for instance bet $2 on every horse in every race.

I just love how you try to remove anyone from the dicussion that challenges you. It makes you look really confident in yourself.

[JackS]

It seems to me that the ability to beat the track take would also be dependent on the average of favorites that actually win. At tracks where the favorite actually wins at a consistant 33% at all odds, the public accepts payouts that equal this minus the 20% take.
To stay even ,they would need to average $6.00 on each bet cashed, but instead are rewarded $5.60.
If your current track under current conditions happens to be playing below the average 33%, knowledged handicappers should have an advantage that is not normally available since the public will still be betting as though the there were no change.
A track that is hitting with 28% favorites and a public that is playing as though this were 33% and expecting $5.60 on their cashed tickets, are really taking a beating (about 15% in my estimation) and when added to the normal track take of 20% you can see the advantage that a good capper might have in this situation
This is not as knowledgable as most answers here, but to me it seems simple and practical.
Bottom line- the more accurate the public is in estimating the actual ability of horses to win, the harder it is to make any money.

[lsbets]

CJ's point is very simple, and I can't believe formula doesn't get it. If you have 1,000 people who go to the track,bet $200 each per year and lose $100 each per year, they have lost a total of $100,000, or 50%. $15,000 of the loss goes to the track for the take out. The other $85,000 gets distributed among some other group of bettors. Since the first group's return was 35 percentage points lower than the take, the group that gets the $85,000 would have to have an ROI better than minus 15% to balance out the loss of the 85K. Horse betting is a zero sum game - that money did not just disappear and go nowhere, it went into someone's pockets, but not the track's. What is so hard to understand about that? Some people are better at betting the horses than others. While the public generally gives a pretty good indication of who might win, the public is often wrong, and sometimes way wrong. By being patient and carefully choosing the horses to bet, a good handicapper can take the 85K away from the public.

[cj]

Thanks ls, sometimes I don't get my point across as clear as I should, but you hit the nail right on the head.

[JustMissed]

;)

[GameTheory]

We're getting two types of percentages confused. You can lose "only the track take" ROI-wise and still lose 100% of your money.

Imagine I start out with only $2. I bet it on a 4-1 shot and it wins, pays $10. Now I've got $10. In the next race, I bet the entire $10 on a single horse, and it loses.

Total amount bet: $12
Total returned: $10
ROI = 10/12 = .83

So I lost 17%, about equal to the track take.

I also lost 100% of my money.

When we say "lose only the track take", we are referring to the ROI.

[JustMissed]

We're getting two types of percentages confused. You can lose "only the track take" ROI-wise and still lose 100% of your money.

Imagine I start out with only $2. I bet it on a 4-1 shot and it wins, pays $10. Now I've got $10. In the next race, I bet the entire $10 on a single horse, and it loses.

Total amount bet: $12
Total returned: $10
ROI = 10/12 = .83

So I lost 17%, about equal to the track take.

I also lost 100% of my money.

When we say "lose only the track take", we are referring to the ROI.

Since we are imagining, suppose you borrowed $10 from your buddy, bet it to win and lost.

Now you have bet $22 and were returned $10. If your ROI is now 10/22 = .4545, are you say the track take is now about 46%, I don't think so.

The track take in this example is about $3.74.

Come on GT, I thought you were a better numbers guy than to start posting "voodoo math".

JM

[cj]

Let's keep it simple. "Bad Capper Joe" bet $10,000 in 2004, and was returned $5000. Of that $5,000 he lost, 15% of what he bet, or $1,500, went to takeout. Where did the other $3,500 go? Other gamblers pockets.

[JustMissed]

Let's keep it simple. "Bad Capper Joe" bet $10,000 in 2004, and was returned $5000. Of that $5,000 he lost, 15% of what he bet, or $1,500, went to takeout. Where did the other $3,500 go? Other gamblers pockets.

Your right CJ-keep it simply because it is simply.

And another thing CJ-you know that $1,500 that the track kept that did not go to the winning players, that money is just another "cost of doing business" for those winning players.


JM

[GameTheory]

Since we are imagining, suppose you borrowed $10 from your buddy, bet it to win and lost.

Now you have bet $22 and were returned $10. If your ROI is now 10/22 = .4545, are you say the track take is now about 46%, I don't think so.

The track take in this example is about $3.74.

Come on GT, I thought you were a better numbers guy than to start posting "voodoo math".

You must have misunderstood me. I never argued that you couldn't lose more than the track take. I have argued that you can lose quite a bit more, and do it easily. I was just clarifying the use of terms. In my example we would say the bettor lost 17%, not 100%. I wasn't trying to say he could ONLY lose 17%, just that to determine the percentage lost you must look at the total of his wagers, not his initial bankroll. In your example, the bettor lost 65% of his investment of $22, or $12. (His original $2 and the $10 he borrowed.)

My point is that at any ROI under 1.0, you will eventually lose all your cash (if you don't keep renewing it), but we wouldn't then say that you suffered a "100% loss" because that takes all the meaning out of the term. Your percentage loss = 1.0-(amount returned/ amount bet)...

[Kreed]

LETS assume there is ZERO takeout. ALL $$ returned to the bettors. And
let's assume only flat win betting allowed. Now what? The same results,
only more bets made before going broke? One more assumption: Suppose
some capper is able to eliminate 1/2 the field .. that is, declare that the
winner is NOT in that 1/2 field. Lets say, races with 7 horses, 4 could be
tossed; 9 horses, 5 tossed. And let's say that in 1,000 races, the predictiction
was wrong in 30 races. Can we say anything NOW about the odds etc etc.
Just some thoughts please. PS: NEVER were the odds used to determine
if a horse was tossed.

[JustMissed]

If he keeps on betting, in the long run he will lose to the tune of the track take-out (along the long to short odds biase). Why, because track odds behave as natural odds
where # winners/#expected winners =1
where % expected winners =(1/(odds+1+breakage))*(1-track-out).

GT, It is probably better not to even mention track takeout when discussing ROI, profits, losses, etc.

People like Joe post stuff like the above and it seems to get people's thinking all screwed-up as to what it takes to make a profit.

In fact, it is probably better to never respond to any of Joe's post and if he should start TROLLING on any other informative thread, one would be wise to find the Ignore Button.

JM :D

[hurrikane]

I just can't believe this is still going on.

for over a year now Joe has been stuck on stupid. (no offense Joe)

He doesn't get it, I don't think he ever will. After 182 posts...(and that is just this thread) I assure you he is still standing in the same spot.

I have to commend everyone though. Never have I seen such a simple concept explained ad-nauseum- in so many creative ways.

congratulations everyone.

and nice trolling Joe

[formula_2002]

Let's keep it simple. "Bad Capper Joe" bet $10,000 in 2004, and was returned $5000. Of that $5,000 he lost, 15% of what he bet, or $1,500, went to takeout. Where did the other $3,500 go? Other gamblers pockets.


The short answer is this. 2004 was a bad year for bcj but the pervious few years were a real boon for him, such, that when he averaged out all those years he lost but 15%.

I asked GT a question. How much would a bettor lose if he bet every horse in the race at a 15% take-out track
his answer was;

There is no analytical answer because it is not a deterministic question, but you would lose approx. 3-4% more than the take if you were to for instance bet $2 on every horse in every race.

Just go to any 15% take-out track, sum 1/(odds+1). that sum wil be about 1.17, dependng on breakage. which equates to a 15% loss.

The sub sets of any large sample should approximate the sample . So I repeat, a random selection of horses will produce the same loss as the entire sample.

There is of course a short to long odds bias.
Fabricand's table on page 34 was based on play from 1955 to 1962 .

The bias is more extreme then it is today.

he looked at 93011 horses. 25,044 horse in the odds range 20-1 through 100-1 had a dollar loss of 54% .
3-1 lost 11% and
1 to 1.15 lost be 2%.

So bcj could have lost 54% of his money and short price harry would have lost but 2%

In todays market, my data form 2003 to 2004 shows the following;

average 1-1 odds, dollar lost =15%
3-1 , dollar loss= 16%
>=20-1 dollar loss = 34%
that's base on over 100,000 horses.

[Kreed]

I like his posts & they do not deter me or annoy me too much. but he likes
craps which has ripped off more money than Enron. hehe but I agree with
his view that he likes action & hey nothing wrong with that. But it is so
quickly accepted that "public odds & true prob" are so alike ... but what of
it? The safety mechanism is "NO BET." As unthinkable as that sounds, sometimes NOT betting saves you money. But enough of this rational stuff:
WHAT WOULD IT TAKE, JOE, for capping to yield a profit? A very low takeout?
Rebates? If you had your wish list, what would a winning method have to have?

[Jed]

Answering the original question, 2 to 3 overlays per card.

[formula_2002]

WHAT WOULD IT TAKE, JOE, for capping to yield a profit? A very low takeout?
Rebates? If you had your wish list, what would a winning method have to have?

a confidence interval >=2.5 and a profit>=0

[formula_2002]

Answering the original question, 2 to 3 overlays per card.

and what question was that Jed :)

[formula_2002]

I like his posts & they do not deter me or annoy me too much.

Not too much? I have to work a little harder :cool:

[formula_2002]

In fact, it is probably better to never respond to any of Joe's post and if he should start TROLLING on any other informative thread, one would be wise to find the Ignore Button.

JM :D

The reason I'm not ignored, is because there is some doubt in everone's mind.. I may be right!!! :o

And GT hangs around because I never answer his question.
Whatever the hell that may be.

But I think SQ has put me on "ignore".. It would be wonderful if it were so. I'd even vote for jeb bush in 2008, if he were to "ignore" me.

[Kreed]

lol ... i love ideas even when resurrected. Try harder. You can be our
public reality principle .... play the part but don't piss me off, u hear F man.
If you can expand on your ideas, tell us. Would you use the Odds in a
predictive sense? If yeah, then ODDS are just another variable, maybe a
very good one, but yet, just one. Why do you seem to be a one variable
guy?

[GameTheory]

Just go to any 15% take-out track, sum 1/(odds+1). that sum wil be about 1.17, dependng on breakage. which equates to a 15% loss.

Are you saying that if you flat bet on every horse in every race your loss will be approximately equal to the track take? Because it won't be. It will be more than that.

Are you saying that if you flat bet on a single random horse in every race your loss will be approximately equal to the track take? Because it won't be. It will be more than that.

Are you saying if you only bet on extreme longshots (again, flat) that your loss will again be be approximately equal to the track take? Because it won't be. It will be much more than that.

Are you saying if you bet on the favorite in every race that your loss will be approximately equal to the track take? Because it won't be. It will be less than that.

Would you like to see the data? If you can confirm your predictions for me, I'll show you.

[Jed]

and what question was that Jed :)

Just how many overlays do you find on a 9 race card??


new to the thread,
-jed

[formula_2002]

Are you saying that if you flat bet on every horse in every race your loss will be approximately equal to the track take? Because it won't be. It will be more than that.

Are you saying that if you flat bet on a single random horse in every race your loss will be approximately equal to the track take? Because it won't be. It will be more than that.

Are you saying if you only bet on extreme longshots (again, flat) that your loss will again be be approximately equal to the track take? Because it won't be. It will be much more than that.

Are you saying if you bet on the favorite in every race that your loss will be approximately equal to the track take? Because it won't be. It will be less than that.

Would you like to see the data? If you can confirm your predictions for me, I'll show you.

answer to questions
1.. It has to be a dutch bet..
2 .. yes, over a very large sample a flat bet will do,
In a smaller sample, it would be more accurate to use a dutch play. In this case, as the sample grows, the dutch results will approach the flat bet results.
3 and 4 .. I agree. it will have a long to short odds bias as I have mentiond more then once.

I have posted excepts from mine and Fabricand's,and yes I would like to see yours.
A more complete table of mine is posted on my web page.

so lets see your cards...

[GameTheory]

answer to questions
1.. It has to be a dutch bet..
2 .. yes, over a very large sample a flat bet will do,
In a smaller sample, it would be more accurate to use a dutch play. In this case, as the sample grows, the dutch results will approach the flat bet results.
3 and 4 .. I agree. it will have a long to short odds bias as I have mentiond more then once.
Ok then, it sounds like we actually agree (except about #2, random choices will always lose somewhat more than the track take because that slightly favors the picking of longshots).

So the we agree that when flat betting it is very easy to lose from slightly less than the track take (betting on favorites only) to losing much more than the track take (betting on longshots).

Wasn't your whole point that everyone loses equal to the track take (eventually)?

[lsbets]

JM, yes on the 200K bet 30K is the take, but I said of the 100K loss 15K went to the take.

[cj]

The short answer is this. 2004 was a bad year for bcj but the pervious few years were a real boon for him, such, that when he averaged out all those years he lost but 15%.



This is a false assumption on your part, again. Many guys lose much more than the take, year in, and year out, because they are poor bettors, handicappers, money managers, or some combination of these. This is the flaw in your math.

[formula_2002]

Wasn't your whole point that everyone loses equal to the track take (eventually)?

Actually my point was this;

To prove a winning system or a winning player;

SOLVE FOR NH IN THE EXPRESSION

2.5 = (NW-EW)/(SQRT(EW*(1-(EW/NH))))

WHERE EW=((1/(odds+(1+BREAKAGE))/ (1/(1-track take))

EW= EXPECTED WINNERS
NH= NUMBER OF HORSES TO PROVE 99% CONFIDENCE
--------------------------------------------
One of the keys here is the EW, admittedly , I do not use the phrase "along the low to high track bias"

To be even more accurate, the EW is the ACTUAL win % of the group being tested without any handicapping.
-------------------------


It's not a truly random test but the following results are based on winning program numbers where the entire population loss 21% of their dollars.

pgno loss
1 -15%
2 -24%
3 -4%
4 -20%
5 -5%
6 -35%
7 -36%
8 -16%


the total sample had 81,3954 horse.

[formula_2002]

This is a false assumption on your part, again. Many guys lose much more than the take, year in, and year out, because they are poor bettors, handicappers, money managers, or some combination of these. This is the flaw in your math.

Have you heard the story of the West Point Cadet who, when taking his final exam was asked the question, “Hypothetically, if your company was surrounded by 10,000 Indians, what course of action would you take?”

His reply, “I’d call in my hypothetical 500 tanks and blow the hell out of them”

We can all make-up hypothetical stories.

Query your data base for what is possible and not possible by applying at least some attempt of scientific thinking.

[cj]

My situation is not hypothetical, its very real.

[Exactaman]

The short answer is this. 2004 was a bad year for bcj but the pervious few years were a real boon for him, such, that when he averaged out all those years he lost but 15%.



That's one of the most unlikely predictions made on available evidence that I've ever seen :):):) I'm beginning to see why you think this game can't be beaten :):):)


Query your data base for what is possible and not possible by applying at least some attempt of scientific thinking.

you don't have to be a mathemetician to realize that horses' future performance can be predicted based on their recent performance. this is a complex skill, with many different possible approaches, that has to be learned. those who are better at it will do better over time. will all major leaguers will eventually bat to the overall average over the long term? of course not. the better batters will record a higher average. so will the better bettors.

[formula_2002]

That's one of the most unlikely predictions made on available evidence that I've ever seen :):):) I'm beginning to see why you think this game can't be beaten :):):)



you don't have to be a mathemetician to realize that horses' future performance can be predicted based on their recent performance. this is a complex skill, with many different possible approaches, that has to be learned. those who are better at it will do better over time. will all major leaguers will eventually bat to the overall average over the long term? of course not. the better batters will record a higher average. so will the better bettors.

I was just flipping through some old copies of "American Turf Monthly" (1955 through 1978).. There are a lot of people who would agree with you then and now.

[GameTheory]

Mr. Formula who doesn't like hypotheticals:

This is getting ridiculous. Actually, it has been ridiculous for some time. You chide one guy for using a hypothetical and your whole argument doesn't have any data to back it up. You chide other guys for dropping names, but when challenged you shrink into a corner and say, "That's what Fabricand says."

Tell me again what is the mechanism by which you can divine the individual performance of each bettor by knowing the average performance of all bettors?

In fact, why don't you just go ahead and list the horses I've bet on in the last year and show my exact ROI. And do that for everyone else that has posted in this thread. You do have all our individual performances in your database, right? If you don't, I'm sure you can calculate it just by knowing the odds, right? Go ahead and do it then.


%. At that rate, the standard deviation for win percentage would be aboSince you can't answer those questions, let's use some of your own math to see what we can figure out. Let's take that ever-present group of 3-1 horses. According to you, horses with odds that round off to 3-1 win at a rate of 21ut .41, which means all you can say about the performace of the individual bettors is that they average 21% and that approx. 65% of them will have individual win rates of 21% PLUS OR MINUS 41%. So going out only one standard deviation we have some bettors hitting at a rate of 60%+. If we go out two standard deviations (which still only covers 95% of the bettors), that covers ALL possible win rates. So your own math shows that we have LOTS of winners in the 3-1 range.

Of course, your math is wrong. In order for your beloved standard deviation to mean anything, you've got to assume that there is a normal distribution around the mean. According to your other math we were just discussing, we know this is not true because of the "short to long odds bias". There is a heavy skew. What does that mean? It means that the standard deviation that you've been harping about all these years is WORTHLESS because it is based upon a false assumption. An assumption you've just admitted you don't hold (admitted indirectly by recognizing the odds bias), so your math is wrong even according to you if you care to use your brain for a minute.

So we're back to square one. Just how is it that you can tell how each individual bettor bets? YOU CAN'T!!!!!

[formula_2002]

Mr. Formula who doesn't like hypotheticals:.........

..........Tell me again what is the mechanism by which you can divine the individual performance of each bettor by knowing the average performance of all bettors? ............



GT, simply put, I am at a loss for words..You have used them all up.."out of order" but still the same .."used up".

[GameTheory]

GT, simply put, I am at a loss for words..You have used them all up.."out of order" but still the same .."used up".GOOD! See how long you can remain in that "wordless" state...

[formula_2002]

GOOD! See how long you can remain in that "wordless" state...

Don't get all happy just yet
. I think its a passing thing.

It seems to happen when I read some ( see, I'm cutting you some slack) of your post....

I would look into using the "ignore" feature, but it would be against my liberal bent to shut you off...

Shouldn't you be off playing the races now?

[GameTheory]

%. At that rate, the standard deviation for win percentage would be aboSince you can't answer those questions, let's use some of your own math to see what we can figure out. Let's take that ever-present group of 3-1 horses. According to you, horses with odds that round off to 3-1 win at a rate of 21ut .41, which means all you can say about the performace of the individual bettors is that they average 21% and that approx. 65% of them will have individual win rates of 21% PLUS OR MINUS 41%. So going out only one standard deviation we have some bettors hitting at a rate of 60%+. If we go out two standard deviations (which still only covers 95% of the bettors), that covers ALL possible win rates. So your own math shows that we have LOTS of winners in the 3-1 range.This paragraph has suffered a cut-and-paste accident. For those machocists that actually read this stuff, it should have been:

"Since you can't answer those questions, let's use some of your own math to see what we can figure out. Let's take that ever-present group of 3-1 horses. According to you, horses with odds that round off to 3-1 win at a rate of 21%. At that rate, the standard deviation for win percentage would be about .41, which means all you can say about the performace of the individual bettors is that they average 21% and that approx. 65% of them will have individual win rates of 21% PLUS OR MINUS 41%. So going out only one standard deviation we have some bettors hitting at a rate of 60%+. If we go out two standard deviations (which still only covers 95% of the bettors), that covers ALL possible win rates. So your own math shows that we have LOTS of winners in the 3-1 range."

[formula_2002]

This paragraph has suffered a cut-and-paste accident. For those machocists that actually read this stuff, it should have been:

"Since you can't answer those questions, let's use some of your own math to see what we can figure out. Let's take that ever-present group of 3-1 horses. According to you, horses with odds that round off to 3-1 win at a rate of 21%. At that rate, the standard deviation for win percentage would be about .41, which means all you can say about the performace of the individual bettors is that they average 21% and that approx. 65% of them will have individual win rates of 21% PLUS OR MINUS 41%. So going out only one standard deviation we have some bettors hitting at a rate of 60%+. If we go out two standard deviations (which still only covers 95% of the bettors), that covers ALL possible win rates. So your own math shows that we have LOTS of winners in the 3-1 range."



What my math is saying about 3-1 horses is this;
they win 21% of the time at a 15% take-out track.

For 2 1/2 std deviations, you would have to win 74 times in 270 plays for a win % of 27.4% and 10% profit.



I'll look into what your math is saying...

[formula_2002]

This paragraph has suffered a cut-and-paste accident. For those machocists that actually read this stuff, it should have been:

"Since you can't answer those questions, let's use some of your own math to see what we can figure out. Let's take that ever-present group of 3-1 horses. According to you, horses with odds that round off to 3-1 win at a rate of 21%. At that rate, the standard deviation for win percentage would be about .41, which means all you can say about the performace of the individual bettors is that they average 21% and that approx. 65% of them will have individual win rates of 21% PLUS OR MINUS 41%. So going out only one standard deviation we have some bettors hitting at a rate of 60%+. If we go out two standard deviations (which still only covers 95% of the bettors), that covers ALL possible win rates. So your own math shows that we have LOTS of winners in the 3-1 range."


Derive that +-41% for me please.

[Kreed]

I love this thread, sorta. It's cool having a standard of measurement.
Over 270 plays, 3:1s win about 54 times; so, if a player can win 74 times
over the 270, that's great. I find the focus on 3:1s interesting because
i've posted many times that horses paying between 8 to 9 dollars seem to
be one kind of horse; those paying 9+ to ~12 are another. It's the old
discriminant ananysis .... at any rate, as I've asked before, are all 3:1s alike?
I don't think so. Take a field of 10: one 9:5; one 2:1; two, 3:1s etc.
And Frabricand must NOT have thought that as well since he went on (and on)
making up scenarios of Maximum Confusion ... I don't think he was ever too
worried about the 20:1s. Imagine, in my field of 10 example, if the 9:5 was
scratched on the way to the gate. Would there be any 3:1s to bet?

[formula_2002]

I love this thread, sorta. It's cool having a standard of measurement.
Over 270 plays, 3:1s win about 54 times; so, if a player can win 74 times
over the 270, that's great. I find the focus on 3:1s interesting because


I use the 3-1 odds range (>=2.5 through <=3.4) because it has the greatest number of horses then any of other the rounded odds range (1-1,2-1,3-1,4-1 etc.)

[GameTheory]

Derive that +-41% for me please.

If the win percentage is .21, then the variance is .21 * (1 -.21) = .1659

The standard deviation is the square root of the variance sqrt(.1659) = .4073

[formula_2002]

This paragraph has suffered a cut-and-paste accident. For those machocists that actually read this stuff, it should have been:

" At that rate, the standard deviation for win percentage would be about .41, which means all you can say about the performace of the individual bettors is that they average 21% and that approx. 65% of them will have individual win rates of 21% PLUS OR MINUS 41%. So going out only one standard deviation we have some bettors hitting at a rate of 60%+. If we go out two standard deviations (which still only covers 95% of the bettors), that covers ALL possible win rates. So your own math shows that we have LOTS of winners in the 3-1 range."

I'm sure this is not what you mean, but this is what I gather you are saying.

So going out only one standard deviation we have some bettors hitting at a rate of 60%+.
So instead of winning 21% in the 3-1 range, some bettors are winning 60%+ in the 3-1 range.

Or do you mean to imply that 60% of all the bettors are betting 3-1 horses.


I don't know how you can determine how may people are playing 3-1 shots, especially since the odds are determined by dollar amounts wagered, not quanity of people .

[formula_2002]

at any rate, as I've asked before, are all 3:1s alike?


Kreed, to me, they are 21 green jelly beans in a 100 count jar, and all green jelly beans are the same.

I gather you may be a fan of Fabricand's .
A long number of yeas ago I had some correspondence with him (before e-mail).
Tried to buy the rights to his book in the early 70's.
only to have McKay, republish it.

After reading so may articles that sounded like "I know this guy" kind of stuff, his work was a breath of fresh air.

[hurrikane]

that may be the problem joe...

rule #1
in horse racing all the jelly beans are not the same.

[formula_2002]

GT

To clarify in my mind what you are saying;
Base on the standard deviation formula, STD dev =sqrt(win% X losing %);

The standard deviations for a 21% win probability in 100 races for my 3-1 shot would be as follows

1 STD dev = +- 4.07 (min of 25% wins)
2 STD dev = +- 8.14 (min of 29% wins)
2 ½ STD dev =+- 10.17(min of 31% win)
So in 100 races there is a 1% probability of winning 31 races.

That method is also use by Fabricand (see page 176, Horse Sense).

What Fabricand failed to adequately address was sample size.
SO..

Quirin’s use of the formula for determining the min # of plays required to attain a 1% level of confidence, for the above 3-1 shot winning at a 27.5% rate for a 10% profit, would be 270 races.

Back to Fabricand. He indicated that in the 3-1 odds range, his system plays were 25% of all 3-1 horses.

My point is this.

3-1 horse’s represent 9% of all the horses that run (per my data base)
Say 25% of those horses are system plays.

Therefore (270/ (.25 X.09)) = 12,000 horses must be handicapped to prove 1% significance. And that is just for the odds range (>=2.5 through<=3.4).

While some people may be able to back fit data to the above, I think it highly improbable that anyone can support their claim of profitability based on the above method.
I can not prove the sun will rise up every day, but that’s how I would bet it.

[GameTheory]

GT

To clarify in my mind what you are saying;
Base on the standard deviation formula, STD dev =sqrt(win% X losing %);

The standard deviations for a 21% win probability in 100 races for my 3-1 shot would be as follows

1 STD dev = +- 4.07 (min of 25% wins)
2 STD dev = +- 8.14 (min of 29% wins)
2 ½ STD dev =+- 10.17(min of 31% win)
So in 100 races there is a 1% probability of winning 31 races.

That method is also use by Fabricand (see page 176, Horse Sense).

What Fabricand failed to adequately address was sample size.
SO..

Quirin’s use of the formula for determining the min # of plays required to attain a 1% level of confidence, for the above 3-1 shot winning at a 27.5% rate for a 10% profit, would be 270 races.

Back to Fabricand. He indicated that in the 3-1 odds range, his system plays were 25% of all 3-1 horses.

My point is this.

3-1 horse’s represent 9% of all the horses that run (per my data base)
Say 25% of those horses are system plays.

Therefore (270/ (.25 X.09)) = 12,000 horses must be handicapped to prove 1% significance. And that is just for the odds range (>=2.5 through<=3.4).

While some people may be able to back fit data to the above, I think it highly improbable that anyone can support their claim of profitability based on the above method.
I can not prove the sun will rise up every day, but that’s how I would bet it.
I was just going to let this die since we're getting so esoteric and technical and the point of my argument was that my example *DIDN'T* make sense.

So your point is that you have to handicap lots of horses. Got it.

[Exactaman]

i support my claim of profitability on the ability to pick 3-1 shots--for argument's sake, in fact i wouldn't touch one with a ten-foot pole :)--at a higher rate than the crowd average of 21%.

[formula_2002]

E...I'm not certain of your meaning, but it sound's like your'e a winner...

Good luck (It can't hurt!!)

[Exactaman]

thanks, I need it :)

i believe it is mathematically possible for my system picks to score at a higher average than the crowd's for any given odd range, and thus return a long-term profit.

it is my opinion that there are FAR greener pastures than the 3-1 range, or low odds horses in general.

[valueguy]

Agree with Jacks ,find the underlays in the contender ranks(if there are any)
and then bet the remaining contenders if you think they have enough value to bet.Its been my experience that you.will find about 4 races that are playable and maybe 2 that are bettable.
Finding an overlay with high odds is a difficult task A horse going of at 5-1 odds
will only be an overlay about 1 out of 10 times.It gets more difficult as you go higher in the odds.

[Exactaman]

Therefore (270/ (.25 X.09)) = 12,000 horses must be handicapped to prove 1% significance. And that is just for the odds range (>=2.5 through<=3.4).

A horse going of at 5-1 odds
will only be an overlay about 1 out of 10 times.It gets more difficult as you go higher in the odds.

This kind of psychology runs counter to the way I for one play. It seems to envision a random sifting of masses of possible selections.

My method is proactive. I identify horses I wish to play based on observation, with an eye on identifying positive performance the crowd may have missed. Those jellybeans are not all green! I don't have to handicap large numbers of horses to come up with such selections, and I don't think they mathematically must only prove to be overlays at an average rate.

[formula_2002]

thanks, I need it :)

i believe it is mathematically possible for my system picks to score at a higher average than the crowd's for any given odd range, and thus return a long-term profit.

it is my opinion that there are FAR greener pastures than the 3-1 range, or low odds horses in general.

I agree with you that it is "mathematically possible".
My only point is, it requires a vary large number of races to prove the point and I dont think most peolpe quite understand that and, I would estimate that there is a very high probability that no one ever has proven the point in the manner described here and in countless reseach papers .

[valueguy]

This kind of psychology runs counter to the way I for one play. It seems to envision a random sifting of masses of possible selections.

My method is proactive. I identify horses I wish to play based on observation, with an eye on identifying positive performance the crowd may have missed. Those jellybeans are not all green! I don't have to handicap large numbers of horses to come up with such selections, and I don't think they mathematically must only prove to be overlays at an average rate.

I am doing basically the same thing you are doing but only going through the back door .I look at races where i figure there will be one or two underlays(obivious form ,top jockey etc.) with a fairly low M/L,then i look for a horse or horses that can beat the underlay .Usually you can find one ,but the odds the Tote may give you does not fit your risk ,reward ratio.
While you are perusing the PP,s you are also looking for a nugget that will allow you to try for a big score .This may happen only once or twice in 50 races.But once you find a horse with a high probabilty of finishing 1 st or 2 nd
going off at good odds then it becomes a problem of finding the best way to bet it. Which is a good problem to have.
If i remember correctly , one of the old Speed Boys once said that :the secret to winning at the races is having a strong butt. For my style of playing
he is are right.
Regards : have a good one.