PDA

View Full Version : Odd's Line Math


MitchS
09-07-2004, 08:44 PM
Hi all,

I am looking for some help in obtaining a mathematical formula in creating an odds line. A 100% line. I used to know the forumla years ago. I just can't seem to remember. The use of easy input odds line programs have made me lazy. If someone has a mathematical formula they would like to share for the use in creating an odd's line from final fig's, i would be much indebted.

Thanks ahead of time,

Mitch

keilan
09-07-2004, 09:28 PM
Mitch -- have you viewed the odds line templates for three and four horse contenders from "Steve Fierro" book titled The Four Quarters of Horse Investing?

Unless you really want to make a line for every horse in the race I think you will find these templates more than satisfactory.

MitchS
09-07-2004, 09:41 PM
Thanks Keilan, but what i am really looking for is something like Gordan Pines Figline. The mathematical Formula that i can input into a database so i do not have to manually input any fig's.

Mitch

keilan
09-07-2004, 09:48 PM
Mitch --try doing a search as this topic has been talked about many times before. Good Luck

Tom
09-07-2004, 10:57 PM
Try this.

http://turfpedia.com/process/make.html

Handle
09-08-2004, 11:17 AM
Mitch,

OK, I think this will work. Feedback appreciated!


You need a formula that will first normalize your weighted "value rating" into a percentage- a number from 0-100, for each horse.

To do this we total up all of the value ratings for each horse. Then we divide each horse's value rating by the total. We multiply this number by 100 to make it into a percentage.

Now we need to convert this percentage into an odds line. We are assuming that this percentage is the percentage of the races that this horse will win based on its value rating.

So we create an algebraic formula where we want to solve for odds. It looks like this (we'll work with 2 dollar bets, though 1 dollar bets could be used to make it cleaner):

Odds = (TotalBet - (winPerc * 2)) / (2 * win%)

TotalBet is important. To Break EVEN you want to set it based on 100 wagers. If you want a cushion in your value line, you make Total Bet larger.

An example:

Odds = (200 - (25 * 2) ) / (2 * 25)

That is, 100 wagers at 2$ per = 200.
25% wins * 2$ returned per win bet (you get your wager back when you win) = 50

Divide this by 2$ wagered times the win percentage (25 in this case) = 50
odds = 150 / 50
Odds = 3/1

And, to show it works:
you start with 2$.
2 bet win = 8$ returned at 3/1. You started with 2$ of your money, so the winnings is now 6$
2 bet loss = 4 total winnings
2 bet loss = 2 total winnings
2 bet loss = 0 total winnings - you've broken even.

Note here that we win 1 out of 4 times at (3/1) to break even. This is a 25% hit rate, which is the win% used in the equation above.

So, in EquiSim's formula view you'd have a column that normalizes your value ratings for each horse and makes it into a percentage. Assuming that the value ratings are in one column, this is simply:

#COLVAL(n) / #COLADD(n) where n is the number of the column that holds the value rating.

Now, the expression that creates (break even) odds is:

(200 - (#COLVAL(x) * 2) ) / (2 * #COLVAL(x) )

where "x" is the number of the column that holds the normalized (and made into a percentage) value rating.

Again, this is the break even odds. Adjust the "total bet" up from 200 to add "cushion" to your value line.


-Nathan

MitchS
09-08-2004, 03:12 PM
Thanks Nathan,

Looks great, but unfortunitly the formula returns a line that is to close together from the top ranked to the bottom ranked horse. More weight need to be applied to the top 4 or 5 horses, the contenders. Not quite sure how to do his in a formula, code.
But were having fun and i guess that is all the matters.

Since i am here i would like to give a plug for ESV5 and the formula view where you can create anything you want in a system or method. With all of features of V4, added to V5 makes it in my opinion one of the top handicapping programs around. Let me tell you when it comes to V5 and the Formula View, I was a little hesitant at first (my math skills are a little short) but i was very surprised at how easy you made it. IF I CAN USE THE FORMULA VIEW ANYONE CAN...


Thanks Nathan for a great product.

Mitch

headhawg
09-09-2004, 12:16 AM
Mitch,

See if you can get your hands on a copy of Dick Mitchell's book Winning Thoroughbred Strategies. He combines a speed and class method to obtain a rating, and then shows how to convert that to an odds rating. The book is a good read even if you don't use his method to the letter.

I also think that CJ created a spreadsheet to calculate a fair odds line off of a rating. I think it was called linemaker.xls. It might be on his site www.pacefigures.com under Free Tools.

HH

hdcper
09-09-2004, 12:33 AM
Mitch,

Headhawg recommendation is an excellent one! Mitchell's method of calculating an odds line works well overall and I used it many years ago.

I would try to summarize it, but am moving in a few weeks and have packed my handicapping books away.

Check it out, might be just what you are looking for.

Bill

JackS
09-09-2004, 12:57 AM
ML Odds total (all horses added) will be over 100%. The total ML can be viewed as actual odds versus odds offered minus track take. Generally the total will be between 120-130%. That leaves 20-30% to subtract from the ML odds offered. This is the easy way but, computer programs may be a little more accurate because of high ML odds usually don't exceed 20-1. To work quickly with a pencil or pen, decrease all odds by 20%. This number should be accurate enough since opinions of true odds vary widely. Your choice.

BillW
09-09-2004, 03:05 AM
MitchS,

Here's another resource:

http://www.handicapping.com/library/carroll/

The righthand column has a few articles by Charles Carroll on making an oddsline.

Bill

plainolebill
09-09-2004, 04:28 AM
After Hdcpr posted I dug out the book. Here's how Mitchell does it:

First he indexes his raw number to fall within the range of 60-100 pts.

Indexed Value = 60 +(40 ((raw score - lowest raw score)/range))

Then converts it to an ability rating. That formula is:

.0245 x (power rating) -1.46
(60pts comes out to be .01, 100pts = .99)

Finally he totals the ability ratings and divides each horse's ability rating by the total. (he actually uses a percentage of the total based on how many contenders he has, and assigns the balance to non contenders)

You can put that into Equsim 5's Formula View and it will give you the probablity percentages and if you like can convert that into a fractional odds line.

Hope I've got this right - It's late and I'm old.

MitchS
09-09-2004, 07:22 AM
A big thank you from everyone who has posted on this thread, especially Plainolebill for digging up the formula. I will feed this into ESV5's formula view and see how this line looks.

Thanks again,
Mitch

Equineer
09-13-2004, 10:22 AM
MitchS,

The formula for a break-even odds is, and always will be, (1/probability)-1.

As far as I can tell, all of the methods suggested to you skip by the major question... why compute odds unless you have tested the accuracy of your underlying probabilities?

Any odds formula/calculation based on "fuzzy math" is an acknowledgement that you are working with handicapped results that fail to show much correlation to probabilities.

cj
09-13-2004, 11:22 AM
Originally posted by Equineer
MitchS,

The formula for a break-even odds is, and always will be, (1/probability)-1.

As far as I can tell, all of the methods suggested to you skip by the major question... why compute odds unless you have tested the accuracy of your underlying probabilities?

Any odds formula/calculation based on "fuzzy math" is an acknowledgement that you are working with handicapped results that fail to show much correlation to probabilities.

I've been through this before with some other guys, but I'll try again anyway.

You could spend a lifetime trying to "prove" that your odds are completely accurate.

Here is what I think. I make a betting line for every race. I follow this betting line when making betting decisions. I have won money for almost 5 years straight now. Therefore, my odds line is good enough. Do I really care if the horses I make 4-1 win 20% of the time? Or if the horses I make 9-1 win 10% of the time? Of course not. Would it help to know this information? Maybe it would, maybe it wouldn't, but here is what I do know.

Horses I make the favorite on my line that don't go off the post time favorite make a substantial profit.

Horses I make prohibitive favorites that also go off as race favorites are best avoided. They win a high percentage of the time, they show a loss if bet flatly, and betting others overlays in the race shows a very big loss.

If I have a race with more than 4 contenders (they are given a morning line by my system), DON'T BET!

There are other rules I use, but you probably get the point. You don't have to scientifically prove your odds line is perfect. Profit is all the proof you need.

Equineer
09-13-2004, 07:22 PM
Cjmilkowski,

I may misunderstand your point or what bone you are trying to pick. Are you saying that natural odds are anything but a representation of probabilities?

For horseracing probabilities, you seem to assert that accurate estimates are a futile pursuit... but then you immediately discuss oddsmaking.

When you say you make a betting line for each race, what are you doing if not expressing estimated probabilities?

If your line is a break-even line, it pretty much has to be a pure expression of estimated probabilities. If your line includes a value premium, you are simply adjusting probabilities to elevate your odds into estimates of what you think constitutes a set of positive-expectation propositions.

Or... are you saying that (1/probability)-1 is bad math? I rather doubt that!

MitchS specifically asked for an odds formula.

Since your odds work for you, why don't you tell him what your formula is if you have one? If your method is purely subjective, that is really not what he asked for.

As for this statement by you,You could spend a lifetime trying to "prove" that your odds are completely accurate.You could also spend a lifetime trying to make a perfect triangle or rectangle, but as futile as this quest for perfection is, we seem to find a lot of utility in millions of reasonably accurate approximations.

And it certainly doesn't require a lifetime effort to simply measure the accuracy of your estimated probabilities.

Speaking of lifetime efforts... the most universal truth is that everyone's mileage will vary! :)

cj
09-14-2004, 03:09 AM
OK, here goes,

I am not saying natural odds are anything but a representation of probabilities. What I am saying is they can be flawed and still be a great tool for betting horses.

I do think completely accurate odds lines are a futile pursuit. I could spend years trying to come up with an accurate line where the percentage of winners matches the percentage I assign each horse. But then what? Maybe I search farther and find that in races for Fillies and Mares sprinting, I'm way better than the average, and in 2yo routes, I'm horrible. And at one track I'm better than others. To make a truly accurate odds line for every class and condition at every track IS a futile pursuit, at least in my opinion.

Accurate, how do you define this? What margin for error? I could prove my 4-1 shots win 20% of the time, but if I did, does it really prove anything? What if they win %25 percent of the time? Sounds great, unless you do more research and find out they only pay like 2-1 shots when they win.

I have no idea why you state the conversion from probability to odds again, I never mentioned it, but I certainly congratulate you on your grasp of Math 101.

I did send Mitch my odds formula. He seemed quite happy with it, though you'd have to ask him yourself. If you search the board long enough, I'm sure you could find it here also. Its also on my site embedded in an Excel spreadsheet, which I'm guessing you could figure out if interested.

I think it boils down to this. You cannot convert a horse race into pure numbers. How are you going to analyze a horse with the best figures in the race, but is dropping 20k in claiming price off of a win? Oh but wait, now you see the trainer wins 33% of the time with this move, but at a slight loss.

How about the horse that 10 races back ran 90 and 95, then ran 6 crappy races in a row, failing to break 70. He then layed off for 8 months, came back, ran a 74 and an 82, and today is entered in his 3d back off of a layoff. How do you "accurately" assess this horse? I would just rather use my general knowledge of situation A and call the horse a noncontender, while betting with both hands on horse B if a 90 will win the race. I'll give him odds based on his number, which probably won't be perfect. But he will win at payoffs large enough to exceed the projected odds enough times to show a profit, and that is all that matters.

So, how would you "simply" measure the accuracy of your probablities? Just lump every race together? That wouldn't prove much. How about breaking the races down by track, distance, class, age, and sex? That would prove something, but in my opinion, it would take damn near a lifetime to have enough races in your sample to prove anything. No thanks, I'll spend my time betting and getting paid.

Equineer
09-14-2004, 08:52 AM
Cjmilkowski,

In your post you scratched the surface by enumerating circumstances that probability handicapping must consider. But all of your examples, plus many additional challenging considerations, are precisely why probability handicapping renders a problem definition that is attractive. Alternatively, if the problem was simple, associated with trivial entry costs and requiring scant knowledge and/or skills that are commonplace, it would not be an attractive proposition as a software project. With respect to futility, futile software endeavors are those that invite widespread mimicry and competition.

You posted examples of handicapping considerations as if to say, "Here is why probability handicapping is futile." However, every one of your examples is circumscribed by data that is available for downloading. Given that, you are most certainly wrong to assume that probability handicapping is futile.

You curiously started your post with "Ok, here goes." In fact, you came across as a bit arrogant in your assessment of what professional developers can accomplish (and have in fact already done). The stuff you described in your post is already in the rear-view mirror. When fast processors and cheap mass storage converged with broadband communications, a new window of opportunity opened, and the symptoms of change are readily apparent... just follow the money streams that have provoked recent wagering controversies back to their sources.

cj
09-14-2004, 10:17 AM
I am well aware of the possibilities of what professional developers can accomplish.

I am stating that it is unnecessary to make a profit at this game, and the time spent trying to work on this puzzle would be better spent on other endeavors.

How about breaking the races down by track, distance, class, age, and sex? That would prove something, but in my opinion, it would take damn near a lifetime to have enough races in your sample to prove anything.

You didn't really address this. All the processing power in the world and brilliant programmers put together can't get a sample size big enough to be meaningful. Let's say you are faced with a 3yo 10,000 claimer at Arlington Park run at 6.5f, for fillies and mares.

How do you account that John Shirreffs was a 30-40% trainer with first time starters for years before his main owner died, but substantially less since?

How about Delta Downs getting slot machines and drawing better horses? How about Maryland racing dropping to near the bottom of the barrel, when it was once the top of the mid tier tracks?

A jockey changes agents and suddenly wins everything, or nothing? The list could go on and on.
By the time you have even moderate sample sizes, so many things in the game will have changed, the data is probably of little value. Different factors also don't change at the same rate. Some pop up overnight, last a few weeks, and disappear. Some gradually occur and stay forever.

The point is that the information required to make a truly accurate odds line is not publicly available, and it never will be. This game can be beaten by focusing on things that are not contained in the downloads you speak of, not on the things that are in the downloads that everyone can see.

Equineer
09-14-2004, 11:49 AM
Cjmilkowski,

With respect to John Shirreffs' dead owner and the other points you raise... everyone is in the same boat, but the players with the most accurate foundation probabilities are best equipped to judge idosyncracies and make wise betting decisions. They don't invest a fortune in software and then ignore the considerations that you cited.

What they want BEFORE the races are more meaningful odds than they would get if they could magically obtain the final track odds BEFORE each race. This naturally gives them an edge when it comes to evaluating the idiosyncracies that you mentioned.

Much of the conventional wisdom about published/downloaded information plays right into the hands of players armed with good probability software. I think this happens because traditional handicappers do not thoroughly address negative factors, are stymied by the sample-size problems inherent in racing data (for horses, pedigrees, jockeys, & trainers), and fail to systematically study the significance of absent/omitted/missing data with respect to probabilities.

Maiden races are a good example of what I mean. Good probability software will substantially outperform the betting public... yet most horseplayers believe that maidens are much more difficult to make-a-line-on than a race for claiming veterans.

ODDS/PROBABILITY COMPARISONS SECTION 40 of 161

MAIDEN-CLAIMING_[12989 Races]

TRACK_ODDS_LEVEL_STATS_____________ HANDICAPPED_ODDS_LEVEL_STATS_______
ODDS Odds Win $2 Odds Win $2
LEVL ____NH ___NW** ___EW Prob _Pct _Net ____NH ___NW** ___EW Prob _Pct _Net
0 4 3 3 72.4 75.0 1.65 52 33 38 72.1 63.5 1.96
1/5 114 83 71 62.6 72.8 1.85 228 128 143 62.9 56.1 1.99
2/5 371 223 204 54.9 60.1 1.75 575 315 316 55.0 54.8 2.08
3/5 834 453* 406 48.7 54.3 1.80 1059 486 516 48.7 45.9 1.95
4/5 1107 539* 483 43.7 48.7 1.80 1223 507 534 43.7 41.5 1.94
Even 1281 531 507 39.6 41.5 1.69 1363 502 539 39.6 36.8 1.87
6/5 1380 528 499 36.1 38.3 1.71 1418 480 513 36.1 33.9 1.89
7/5 756 255 257 33.9 33.7 1.61 775 239 263 33.9 30.8 1.83
3/2 767 263 250 32.6 34.3 1.71 708 237 231 32.6 33.5 1.94
8/5 1615 541 497 30.8 33.5 1.77 1471 430 453 30.8 29.2 1.82
9/5 1733 521 497 28.7 30.1 1.71 1470 398 421 28.6 27.1 1.76
2/1 4663 1266 1193 25.6 27.1 1.73 3472 876 889 25.6 25.2 1.82
5/2 4696 1099 1043 22.2 23.4 1.72 3502 749 777 22.2 21.4 1.73
3/1 4542 908 888 19.6 20.0 1.67 3456 750* 676 19.6 21.7 1.87
7/2 4138 711 725 17.5 17.2 1.60 3308 597 580 17.5 18.0 1.79
4/1 3707 550 588 15.9 14.8 1.54 3336 541 529 15.9 16.2 1.72
9/2 3302 461 478 14.5 14.0 1.58 3279 498 474 14.5 15.2 1.67
5/1 5836 701 749 12.8 12.0 1.54 6252 856 801 12.8 13.7 1.70
6/1 5083 552 565 11.1 10.9 1.61 5775 712* 641 11.1 12.3 1.70
7/1 4384 403 430 9.8 9.2 1.55 5360 496 525 9.8 9.3 1.59
8/1 3904 348 342 8.8 8.9 1.67 4954 429 434 8.8 8.7 1.49
9/1 3484 296 276 7.9 8.5 1.76 4429 360 351 7.9 8.1 1.55
10/1 3265 249 236 7.2 7.6 1.74 4119 320 298 7.2 7.8 1.47
11/1 2867 162 191 6.7 5.7 1.40 3736 265 249 6.7 7.1 1.66
12/1 2599 175 160 6.2 6.7 1.80 3342 235 206 6.2 7.0 1.60
13/1 2359 110 135 5.7 4.7 1.34 3109 192 178 5.7 6.2 1.55
14/1 2179 130 117 5.4 6.0 1.84 2814 151 151 5.4 5.4 1.55
15/1 8405 309** 382 4.5 3.7 1.33 11005 504 502 4.6 4.6 1.32
20/1 6243 195 222 3.6 3.1 1.45 7490 258 268 3.6 3.4 1.26
25/1 18364 336** 430 2.3 1.8 1.29 16539 382 398 2.4 2.3 1.14
50/1 13763 88** 160 1.2 0.6 0.83 8126 63** 96 1.2 0.8 0.45
Avg%Diff/Horse Avg%Diff/Horse
ALL: 117745 12989 0.818% 117745 12989 0.597%

Overlaid_Levels Underlaid_Levels Overlaid_Levels Underlaid_Levels
38538 1.253% 79207 0.606% 58668 0.597% 59077 0.597%
==============================================

ODDS/PROBABILITY COMPARISONS SECTION 41 of 161]

MAIDEN-SPECIAL-WEIGHTS_[8672 Races]

TRACK_ODDS_LEVEL_STATS_____________ HANDICAPPED_ODDS_LEVEL_STATS_______
ODDS Odds Win $2 Odds Win $2
LEVL ____NH ___NW** ___EW Prob _Pct _Net ____NH ___NW** ___EW Prob _Pct _Net
0 10 8 7 72.0 80.0 1.74 22 14 16 71.8 63.6 2.19
1/5 99 78* 62 62.9 78.8 2.00 103 59 65 63.1 57.3 2.03
2/5 359 234* 198 55.2 65.2 1.90 338 175 187 55.2 51.8 1.84
3/5 640 367* 312 48.8 57.3 1.89 569 284 278 48.9 49.9 1.98
4/5 845 404 371 43.9 47.8 1.76 737 322 323 43.8 43.7 1.89
Even 920 406* 366 39.8 44.1 1.80 894 336 355 39.7 37.6 1.82
6/5 929 352 337 36.3 37.9 1.70 900 302 327 36.3 33.6 1.70
7/5 497 192 169 34.0 38.6 1.85 462 132** 157 34.1 28.6 1.55
3/2 498 173 163 32.7 34.7 1.73 435 152 142 32.7 34.9 1.88
8/5 1019 341 315 30.9 33.5 1.77 949 294 293 30.9 31.0 1.82
9/5 1126 338 324 28.8 30.0 1.70 928 275 267 28.8 29.6 1.81
2/1 2928 769 752 25.7 26.3 1.67 2398 606 616 25.7 25.3 1.76
5/2 2903 626 647 22.3 21.6 1.58 2388 545 532 22.3 22.8 1.83
3/1 2763 548 543 19.7 19.8 1.66 2348 490 462 19.7 20.9 1.70
7/2 2571 474 453 17.6 18.4 1.72 2349 416 413 17.6 17.7 1.65
4/1 2412 383 384 15.9 15.9 1.64 2363 404 376 15.9 17.1 1.65
9/2 2076 279 301 14.5 13.4 1.53 2145 343 312 14.5 16.0 1.82
5/1 3665 477 472 12.9 13.0 1.67 4337 579 558 12.9 13.4 1.66
6/1 3428 368 383 11.2 10.7 1.59 4137 498 462 11.2 12.0 1.80
7/1 2941 267 290 9.8 9.1 1.52 3931 370 387 9.8 9.4 1.52
8/1 2639 223 233 8.8 8.5 1.59 3573 324 315 8.8 9.1 1.59
9/1 2368 186 189 8.0 7.9 1.63 3161 265 252 8.0 8.4 1.49
10/1 2220 161 161 7.3 7.3 1.65 2930 206 213 7.3 7.0 1.59
11/1 1912 124 128 6.7 6.5 1.61 2663 176 178 6.7 6.6 1.57
12/1 1772 87 110 6.2 4.9 1.31 2362 142 146 6.2 6.0 1.54
13/1 1606 76 93 5.8 4.7 1.36 2096 109 121 5.8 5.2 1.62
14/1 1479 63 80 5.4 4.3 1.31 2010 95 108 5.4 4.7 1.19
15/1 5717 222** 262 4.6 3.9 1.41 7504 314 344 4.6 4.2 1.28
20/1 4313 145 154 3.6 3.4 1.56 4867 150 174 3.6 3.1 1.10
25/1 12557 241** 296 2.4 1.9 1.32 9706 240 236 2.4 2.5 1.16
50/1 9550 60** 112 1.2 0.6 0.86 5157 55 60 1.2 1.1 0.68
Avg%Diff/Horse Avg%Diff/Horse
ALL: 78762 8672 0.794% 78762 8672 0.541%

Overlaid_Levels Underlaid_Levels Overlaid_Levels Underlaid_Levels
18869 1.671% 59893 0.517% 39388 0.538% 39374 0.544%
==============================================

cj
09-14-2004, 12:01 PM
Your chart confirms what I am talking about. You lump all the maiden special weight and all the maiden claiming races together.

There are many, many totally different types of Maiden Special Weight races. Is the race at Dmr or Beu? Is it a sprint, or a route? Dirt, or turf? 2yos, 3yos, 4yos and up? You would find many differences between these sub-categories. I guess the question is, do you ever know if you have a truly accurate odds line?

To me, the answer lies in the bottom line.

Equineer
09-14-2004, 12:57 PM
Cjmilkowski
To me, the answer lies in the bottom line.Yup, results attract money... you are looking at a substantial investment in data and software development to build a decent probability system... our work-in-progress system is by no means the most comprehensive (from what we know about a couple of others), but it still exceeds 200,000 lines of original code, excluding all the commercial tools and libraries.

It's really fine and good that you are happy and satisfied with your approach. Good luck whenever the occasion is "Ils sont parti!"

Jeff P
09-14-2004, 01:13 PM
posted by Equineer...

As far as I can tell, all of the methods suggested to you skip by the major question... why compute odds unless you have tested the accuracy of your underlying probabilities?


and posted by CJ...

To me, the answer lies in the bottom line. ... I'll spend my time betting and getting paid.



Don't know exactly where, but at some point this thread turned into a good one.

In my mind, both points of view are right. Yes, you need to test the accuracy of your odds line before it really has any meaning. And yes, if your line is good enough, perhaps what you want to do is spend your time betting and getting paid.

In my own case, I've come up a very accurate line based on measuring the results of thousands of historical occurances. I've also created some software that automates the whole process. Guess what? When the software is confronted with fresh data that it hasn't seen yet and it says that a horse has a 30 percent chance of winning todays' race, that horse really does have a 30 percent chance of winning today's race. At least that's what I've seen so far based on measuring results from thousands of new occurances.

But does my line have to be perfect? No. Could I improve upon it? Definitely. In fact, I'm constantly tinkering with and testing out new numbers to see if I can improve upon what I already have. And over time, by keeping improvements and discarding things that have negative effects, my own line gradually becomes stronger than its earlier versions. I'm a freak. I really do enjoy the process of performing my own research and tweaking my own numbers every bit as much as I enjoy grinding out a profit and winning money.

Perhaps not everyone gets enjoyment from such pursuits. Perhaps some just want to win money.

All I or anyone else really needs to make money is this: an oddsline better than what the betting public can produce. Since the majority of those who wager on horses don't keep records of any kind or even work with data driven oddslines, how hard could it be to come up with a useable odds line?

Obviously, the better the oddsline, the more likely it becomes that the player using it can make money.

Obviously, CJ has created a line that works for him. Is is perfect? No. Does it have to be? No. He knows where it's strong and he knows where it's weak. Could he improve upon it? If he wanted to take the time and effort to do so he probably could. But he makes a good argument. At some point, once you have a line that's better than what the betting public has, maybe what you want to do is spend your time betting and getting paid.

keilan
09-14-2004, 02:20 PM
I would like to hear your thoughts on making an odds line only for your contenders vs. making a line on the entire field. What are the advantages or disadvantages associated with both from your perspective.

chickenhead
09-14-2004, 05:22 PM
the problem I have understanding all of this, you can have a dead nuts accurate odds line on average over a very large sample of races without ever once being right about any one specific race.

You'll never know how right you were for any one specific race, all you will know is that on average you were right, which really doesn't tell you much at all.

I know that the average voltage of a sine wave is zero, that doesn't tell me anything as to what the instantaneous voltage is at any point in time.

If I understand CJ right then I agree with him, MONEY is the only measure that tells you you were right more often than not...not the agreement of percentages.

I know the public on average is more or less right...are they a little bit wrong most of the time, are they a lot wrong a little bit of the time, or are they wrong all of the time? You can't tell from the averages, each race in unique.

Tom
09-14-2004, 08:37 PM
Originally posted by keilan
I would like to hear your thoughts on making an odds line only for your contenders vs. making a line on the entire field. What are the advantages or disadvantages associated with both from your perspective.

I have a contender selection methods that has the winner in the top 4 80% of the time, for example. I start out by taking the to 4 contenders and assigning each one 20 points and the rest of the field as a group gets 20, and that is the last time I look at those. The 4 conteders are then mified up or down X points depending on other factors. This focuses you on you 80% probibility plays and saves you a lot of time. Fiero's book with the templates works great for this idea. Just pick one close your opinion of the race. No one will ever know all the relevant factors for any given race. There is just no possible way this can happen. You can only assume you know key ones.
Smarty Jones in the Belmont.....what are his true odds if you know he has bad feet and is hurting more than he did in the Derby? What if you knew he was completely sound and chomping at the but the morning of the race?
If we cannot know this infor for a potential triple crown winner, what chance do we have of knowing how Old Paint feels going to post in the 7th at Fairplex? I look at personal odds lines as just a methological way to look at data consistently so that you can evaluate how effective it is and make corrections.
Like CJ said, the bottom line is all that matters.

Equineer
09-15-2004, 07:09 AM
Originally posted by keilan
I would like to hear your thoughts on making an odds line only for your contenders vs. making a line on the entire field. What are the advantages or disadvantages associated with both from your perspective. From the beginning (with software), I have always handicapped all horses in each race. But I can appreciate why either contender-selection or non-contender-screening methods have value for players who handicap by hand or use programs that require a lot of human interaction for pace-line selections and the like. In these cases, time is money, so reducing the workload enables more races to be handicapped. My software approach is to automatically handicap all downloaded cards, for all races, and all horses.

After thinking about this... maybe some of the issues in previous posts should be re-visited. Here is what I mean. When you buy a racing form, imagine that the final odds for each race were printed as part of the past performances. Good handicappers would really prosper if they knew hours in advance what the final odds would be. Probability handicapping software tries do this (and even surpass the usefulness of final odds), but you don't have to fiddle with it. Even after program changes come out, it only takes seconds to re-handicap any racecard and get new odds/probabilities.

So there really is no conflict between what we do and the stuff posted earlier by CJ and Tom. Since practically no daily work is required, who can deny that they would benefit from a really good set of probabilities and odds when they sit down to handicap the races. Like Jeff said, what more could you ask for than accurate and useful probabilities/odds that are better than what the betting public produces. Notice I said accurate and useful. Jeff and I discussed evaluating the usefulness of odds/probabilities in another thread.

When you achieve a high degree of accuracy, you still have to determine whether your odds/probabilities are useful. This means questioning whether your good "average accuracy" can be reliably used to make money. Of course, real play answers this, but you can also use database/statistical techniques to evaluate this. Good analysis also enables you to fine-tune the value premiums that you use to qualify playable overlays under varying circumstances (e.g., track, field-size, odds-level, race type, etc.).

keilan
09-15-2004, 12:18 PM
Firstly I appreciate the time your taking to discuss your methods.

>From the beginning (with software), I have always handicapped all horses in each race. But I can appreciate why either contender-selection or non-contender- screening methods have value for players who handicap by hand or use programs that require a lot of human interaction for pace-line selections and the like. In these cases, time is money, so reducing the workload enables more races to be handicapped. My software approach is to automatically handicap all downloaded cards, for all races, and all horses.

While I agree software does enable the player to view many races in a short order of time I’ve often wondered how software can for example accurately detect false favourites from legitimate favourites. I whole-heartedly agree “time is money” but can any player wager with conviction if this one aspect is missing from their arsenal? Everyday I see false favourites go off at 8/5 or 2-1 etc, is it the $2 bettor that is making these mistakes or is it someone else?


Ø Good handicappers would really prosper if they knew hours in advance what the final odds would be. Probability handicapping software tries to do this (and even surpass the usefulness of final odds), but you don't have to fiddle with it.

Good handicappers know almost instinctively which races to attack, and that occurs when their opinion is different than the publics. I guess one could call that probability handicapping but whether one has a program or sound judgement I fail to see the advantage.


Ø So there really is no conflict between what we do and the stuff posted earlier by CJ and Tom. Since practically no daily work is required, who can deny that they would benefit from a really good set of probabilities and odds when they sit down to handicap the races.


Here’s the rub for me – is your software as good or better than an odds line generated manually? If the answer is yes than you certainly have a strong advantage over most players, if not then your shelf-life will expire soon enough. In no way do I want you to interpret the following as a challenge because that is not my intent.

Here’s what I propose, you and I select a race each from any track running tomorrow and both of us will post our odds line for that race. Yours will be for the entire field and be produced by your software, my odds line will only include my contenders and will be produced manually. While I appreciate two races is hardly representative of your work I should be able to gain some degree of insight to your software’s usefulness. I ask only that the races we select do not comprise of many FTS, as little can be learnt from that for me.

Equineer
09-15-2004, 06:16 PM
Keilan,
Here’s the rub for me – is your software as good or better than an odds line generated manually? If the answer is yes than you certainly have a strong advantage over most players, if not then your shelf-life will expire soon enough. In no way do I want you to interpret the following as a challenge because that is not my intent.

Here’s what I propose, you and I select a race each from any track running tomorrow and both of us will post our odds line for that race. Yours will be for the entire field and be produced by your software, my odds line will only include my contenders and will be produced manually. While I appreciate two races is hardly representative of your work I should be able to gain some degree of insight to your software’s usefulness. I ask only that the races we select do not comprise of many FTS, as little can be learnt from that for me.Hmmmn... a sample size of one cherry-picked race is hardly a sample at all. Let's figure out the most mutually convenient way to use a sample of at least 1,000 forthcoming races.

Equineer
09-15-2004, 06:38 PM
Keilan,

To clarify what I meant about your proposal in my previous post...

What will one selected race that probably won't be the same for both of us show?

I mean, if anybody handicaps one race, their odds/probabilities compared to actual win percentages are bound to show no correlation except for the one winner, and comparing a 100% win percentage to the predicted odds/probability for that one winner will be meaningless.

We need a decent sample size, consisting of the same races, to compare our predicted odds/probabilities to their actual win percentages.

keilan
09-15-2004, 08:00 PM
I mean, if anybody handicaps one race, their odds/probabilities compared to actual win percentages are bound to show no correlation except for the one winner, and comparing a 100% win percentage to the predicted odds/probability for that one winner will be meaningless.



You are absolutely right and I really did feel the same when I wrote my most recent post. So without the prospect of us establishing an odds line for a hundred races +- I tried to quantify whether your odds-line software has merit.

What I wanted to do is choose a race that would feature a false favourite, running at a distance that would vary from 6 furlongs to 61/2 - 7 furlong sprint with possibility some medication (lasix), equipment (blinkers) adjustments and a barn change for one or more of the horses in the field. I feel that if your software can accurately account for these variations you would have gotten my attention.

To be perfectly honest I read lots of rhetoric associated with handicapping but seldom get an opportunity to see the data or results from the said claim. Having said that I think you may be the exception.

I would agree to any reasonable format that would showcase your software.

Derek2U
09-15-2004, 08:15 PM
You guys don't know ODDS. You're ALL ODD and 121:1 to WIN.
But I still luV u (Xcept ~10 of U) ... and I 4Give Ur idiocies.

keilan
09-15-2004, 08:30 PM
Ahh gees -- now Derek is here, so how is my favourite little perv tonight :) hehe slow nite!

Derek2U
09-15-2004, 08:51 PM
ok let's be 100% true , I am the BEST & I will write a song about this all -- in fact, I am writing it NOW --- & i will perform it at the Mercury Lounge in nyc where I play ~~ so often .... but enuff & I recall how great it was before these freaks joined here & tried so hard to be a GUYS. LOL .... LOL ... BUT KEEP COMING IN THE WR,
cause that's where Predictions Happen & It's Fun.

sjk
09-15-2004, 09:26 PM
The debate about how to compare the efficicacy of computer-generated vs hand-made odds lines points out a huge benefit of the computer generated line: The computer can make lines for 100 races in a matter of minutes which opens up a large number of betting opportunities.

I would agree with Equineer that you would need a significant number of races to test odds lines (although I think a few hundred might be enough). An odds line does not need to be correct every time to generate a useful return.

Odds lines are probabilities and they need to be tested over a long run. You would not test the probability of getting heads in a coin flip by flipping one coin.

keilan
09-15-2004, 10:25 PM
Sjk

So anyone with a computer-generated odds-line now is at a huge advantage. That’s hogwash

That’s like saying because there is now internet/simulcast wagering a losing player will become a winning player because of the number of increased wagering opportunities.

Making an accurate odds line is no easy task but it begins with an acute understanding of handicapping, methodology and programming. Anyone can generate an odds-line electronically or otherwise, but before one begins wagering their bankroll you had better ensure that the line is reliable.

To account for every imaginable variation in each horserace the syntax must be exhaustive. It staggers me to believe the amount of code that would be needed to generate a line that I could rely on. My hats off to those that have reached that plateau.

sjk
09-16-2004, 05:58 AM
keilan,

There are certainly both good and bad computer-generated and hand-generated lines and those who bet the bad ones will go broke.

My point was that if you have a good enough computer-generated line are likely to be able to play lots of races and this can have the effect of multiplying both your fun and your profit always assuming the line is good enough to genereate a profit.

I have bet over 5000 races so far this year. I could not have bet anywhere near as many if I had to do pencil and paper handicapping.

I agree entirely that you need a good understanding of both handicapping and programming to write a good program and that any good program will probably be long and complicated. I found the process interesting and the results very satisfying.

keilan
09-16-2004, 11:55 AM
Okay guys a couple comments before I exit this thread.

Lets assume that most serious players agree that;
1) Different track surfaces favour certain running styles.


Do players need to adapt their handicapping / odds line on the fly if the track is different than what they had handicapped for? The track can be changed by different elements, track maintenance – winds – rain etc. Speaking for myself if I determined that the track to be wet-fast after the 1st race or muddy/heavy I could very well be selecting different horses based on the current track conditions thus changing my odds line. This is a variable IMO that needs to accounted for or else you’re in for a lot of disastrous days unless you decide not to play any track where the track surface is different than what you had handicapped for. IMO if you only play tracks that are unaffected by any elements then you have conceded any advantage.

So my questions to you are;
a) Do you have the ability to accurately detect very early in the card how the track is playing?
b) Does your software have the capacity to recalculate which horses would have the greatest probability of winning based on that current information and issue a new odds line?

It goes without saying that different variables become more or less important when handicapping different class levels. I think this is also true when handicapping “A” tracks vs. “B or C” tracks. When I think of the incredible amount of code that must be written for all situations it blows me away.

keilan
09-16-2004, 09:41 PM
Hey boys, speak up whenever you’re ready.

Just don’t tell me your odds line remains the same regardless of surface conditions. And I thought you guys actually had something, sounded good for awhile. :rolleyes: :rolleyes:

sjk
09-16-2004, 09:59 PM
Keilan,

Years ago I printed reports with three odds lines for speed-biased, neutral and closer biased tracks and used the line that my observation indicated would best apply.

Since I have been using the black box approach and playing 6-12 tracks at a time I have used a single line. There really isn't much time to make the bets, watch the races and still have time to think about biases. The results have been very satisfactory using a single line.

I could very easily have the software adjust the line for bias but as I say I do not.

Occasionally I will see a track that looks so speed-biased that I quit playing that track for the rest of the day, but that is rare.

keilan
09-16-2004, 10:11 PM
This is what I read from your last post. I don’t want to put words in your mouth so correct me if my take is wrong.

You understood there was a need to calculate different odds lines because different track surfaces favour certain runners. What you were unable to do was determine the track weight early enough to take advantage of the current situation, so therefore you simply increased the number of plays to optimize whatever edge you enjoyed.

sjk
09-16-2004, 10:21 PM
Time for just one reply then I need to go.

It has crossed my mind from time to time to make a bias adjustment as you suggest but it never seemed to me to be worth the time and attention that it would require.

I would most likely get shut out a track B while I was thinking about the bias at track A and nothing gets on my nerves more than getting shut out on a winner.

For what it's worth I have done just fine without considering today's bias (consideration is given to the bias at each track for the current week, month and year).

keilan
09-16-2004, 10:32 PM
It has crossed my mind from time to time to make a bias adjustment as you suggest but it never seemed to me to be worth the time and attention that it would require.


Obviously it crossed your mind – you stated in your next to last post “Years ago I printed reports with three odds lines for speed-biased, neutral and closer biased tracks and used the line that my observation indicated would best apply”.

Equineer
09-17-2004, 09:30 AM
Keilan,
"Do players need to adapt their handicapping/odds-line on the fly if the track is different than what they had handicapped for? The track can be changed by different elements, track maintenance, winds, rain etc." Let's add privileged/unpublished/insider information to this question and then examine this issue.

In the various stat charts that I have posted, the Track side reflects what actually transpired, comparing how the betting public handicapped the horses to actual win percentages for the same horses. So to whatever degree the players are tuned into privileged information and also into special real-time circumstances/conditions, this knowledge is reflected in the Track side stats (and the shrewd situational cappers probably played more aggressively than the clueless chumps).

The Handicapped side of the charts compare how the horses were handicapped to actual win percentages for the same horses. Naturally, the handicapping was accomplished without benefit of privileged/insider information and without knowledge of any abnormal real-time circumstances/conditions stemming from track maintenance or weather (just like handicapping by hand on the day before the races). Well, we handicap turf races for on-and-off turf alternatives; but otherwise, the handicapping is based on the expectation that "normal" circumstances will prevail by track, surface, distance, age/sex, and class.

Thus, probability handicappers are challenged to outperform the betting public without benefit of knowledge that is available in varying degrees to handicappers/players during the course of each racecard. So software systems, like our stuff, has to overcome this inherent disadvantage in order to outperform the public.

Then, when it comes to real-time play, the handicapped probabilities become the benchmarks that enable shrewd players to detect and gauge the impact of real-time circumstances/conditions and also spot action driven by privileged information (and decide whether it is relevant or merely steam).

In essence, the good probability systems simply enable the good handicappers/players to gain an extra edge over the competition. In fact, topsy-turvy days are when the good players really thrive.

"When I think of the incredible amount of code that must be written for all situations it blows me away." In my case, I hooked up with two developers who are also experienced players. Our only adult "jobs" have been in development for software/hardware vendors. We were undaunted when we teamed up, but hindsight reveals that we have expended ungodly more hours than anticipated. It's been one of those deals where you can reach 80% of your original objectives in a year, then spend a lifetime expanding your goals and making progress... like grinding out yards in the "red zone." Right now our project code includes over 200,000 compiled statements (not counting 3rd-party toolboxes and libaries). And we have easily discarded/replaced more code than that to get where we're at.

NoDayJob
09-18-2004, 04:23 PM
Originally posted by Derek2U
You guys don't know ODDS. You're ALL ODD and 121:1 to WIN.
But I still luV u (Xcept ~10 of U) ... and I 4Give Ur idiocies.

This is off subject, but have you had a chance to check out Throughbred's new software? Your expertise is valued. Thanks!

NDJ

Overlay
09-25-2004, 06:46 PM
For information on effective methods of developing fair odds for individual horses, full race fields, and selected spot plays without requiring a computer, check out http://www.overlaypublications.com

Lasix1
03-02-2005, 12:05 AM
I agree with the essence of what C.J. says about odds lines. Dick Mitchell, the King of the Value Boys, used to emphasize making an odds line and only betting horses that are a certain percentage above the line. If your odds line is correct and you only bet overlays, there is only one possibility, said Mitchell, and that is winning.

On the other hand, as an astute participant in Mitchell's seminar once asked: "suppose the morning line favorite in a race is laid at 2-1. My odds line also lists the horse at 2-1. I get to the track and discover that the horse is going off at 7-1. Is that a good sign or a bad one? Is that horse a huge overlay or a cold favorite?" Mitchell, always honest, even when caught out in the hyperbole that often characterized his seminars, ruefully admitted that it was probably a bad sign. Ron Ambrose, as skeptical as Dick was hyperbolic, just sat quietly in the corner and smiled.

I always make an odds line and then find when I'm at the track, that I need to disregard certain things in it because of factors I couldn't possibility have foreseen before I got there. Betting overlays may be the key to the mint, as Mitchell emphasized, but figuring out what one is consists of a lot more than looking at your line and betting horses that are over it.

bobbyb
03-02-2005, 07:51 AM
Mitchell and Bris's Power Numbers

I have found that using Bris's Power Number adjusted to 100% and applying Mitchell's formula (Common Sense H) produced pretty good results. Of course each race is unique in it's self, and the usual factors (long layoffs/class moves/surface/FTS) play an important role in whether to use an entry or any entry, or for that matter the race itself.
It's not perfect by any means, but I have found it to be a good guide in setting up my exotic plays.

bobbyb :)

cj
03-02-2005, 08:20 AM
Odds lines are great as a tool. I think generally, most people are afraid to lose on a race and play to many combos. An odds line makes it clear, no value, no bet.

Let's say you have a line like this:
A: 3-1
B: 4-1
C: 4-1
D: 8-1
Now, the real odds are these:
A: 4-5
B: 6-1
C: 4-1
D: 6-1

I now know the A will not be on my tickets, period. No throwing him in exacta savers, no hedging and betting half because I'm afraid he might beat me. Play the race to win some money.

Let's say you are going to bet on the race. Bet B, don't hesitate, no betting $30 to win and a $10 exacta box with the chalk, no tris keying B with the chalk in the other slots. Play the value. I know for me at least, when I used to play those savers, they ate up my profits over time. If you are right, get paid. If you are wrong, turn the page. No one race, or one racing day, is going to kill you in the long run. But whittling away your bank playing undervalued choices will.

bobbyb
03-02-2005, 08:37 AM
well said cj

2/3's of a shot to be right! :) and back to your earlier post(s) on this thread, I too am in it for 1 thing - PROFIT. I really don't care how close my line is to reality - cause I know it will never be - Our biggest advantage is knowing that the public fav. is attempting the impossible - a winner in every race.

bobbyb :) :)

andicap
03-02-2005, 01:45 PM
. On the other hand, as an astute participant in Mitchell's seminar once asked: "suppose the morning line favorite in a race is laid at 2-1. My odds line also lists the horse at 2-1. I get to the track and discover that the horse is going off at 7-1. Is that a good sign or a bad one? Is that horse a huge overlay or a cold favorite?" Mitchell, always honest, even when caught out in the hyperbole that often characterized his seminars, ruefully admitted that it was probably a bad sign. Ron Ambrose, as skeptical as Dick was hyperbolic, just sat quietly in the corner and smiled.


You have to be able to tell a real overlay from a false one. That 2-1 shot could well be a false overlay or the ML maker might really suck. I've seen lots of 2-1 ML horses go off at 6-1 and win just as a ton of 10-1 ML horses bet down to 2-1 lose.

Best way to do this, IMHO, is handicap the public. If the horse on paper, SHOULD be 2-1 and he's 6-1, something's probably fishy. E.G. horse has best Beyer figure and good recent form, you shouldn't get 6-1.

But maybe the horse is 2-1 on the ML because the horse has best recent form and the oddsmaker THINKS the public will bet him down. (Remember the ML is not the linesmaker's opinion on who will win, but what the public will do.) But the public fakes out the oddsmaker and goes for the big class dropper or the 35% winning trainer so the 2-1 ML shot drifts up to 5-1.

Michael Pizzolla suggests and exercise in which before the race you give it a quick glance and decide the top 3 horses the public should like based on what's in the Racing Form. If one of those is a huge longshot be suspicious.
He suggests to prefer ,
a) higher odds horses that have hidden qualitied that you uncover that aren't in the DRF
b) horses who are BET DOWN (but still above YOUR odds line) even though the public should hate them. For example, you make a horse 3-1 based on some obscure trainer angle. The horse is 15-1 ML and looks it on paper, but opens at 6-1. Pizzolla says thats a bet both hands type of bet, the best of both worlds -- an overlay on your handicapping, and a "steam" horse.

Overlay
03-02-2005, 05:42 PM
My hat is off to those who can reliably interpret toteboard trends to detect false overlays. For me, concentrating on the significance of odds shifts gradually erodes my confidence in my own handicapping, and leads to endless second-guessing and a conspiracy frame of mind about whether "today is (or is not) the day" for a particular horse. I try to compensate for the omission of toteboard-reading, and to even out the effects of horses who are "dead on the board", by employing a properly weighted variety of fundamental handicapping factors which have a marked, recurring correlation with a horse's winning chances. This helps me avoid the low mutuels that result from focusing on just one overriding aspect of performance that everyone else is keying on, while at the same time reducing the misgivings that might result when a horse goes off at odds above my fair-value line. By using fundamental factors rather than peripheral aspects of performance, I find that when I come up with a horse that my line rates as an overlay, it will have enough positive aspects to its record that the resulting frequency of winners and the prices I get on them will make up for the losses from horses whose high odds indicate the likelihood of a less-than-optimal effort today.

shanta
03-03-2005, 07:30 AM
Odds lines are great as a tool. I think generally, most people are afraid to lose on a race and play to many combos. An odds line makes it clear, no value, no bet.

Let's say you have a line like this:
A: 3-1
B: 4-1
C: 4-1
D: 8-1
Now, the real odds are these:
A: 4-5
B: 6-1
C: 4-1
D: 6-1

I now know the A will not be on my tickets, period. No throwing him in exacta savers, no hedging and betting half because I'm afraid he might beat me. Play the race to win some money.

Let's say you are going to bet on the race. Bet B, don't hesitate, no betting $30 to win and a $10 exacta box with the chalk, no tris keying B with the chalk in the other slots. Play the value. I know for me at least, when I used to play those savers, they ate up my profits over time. If you are right, get paid. If you are wrong, turn the page. No one race, or one racing day, is going to kill you in the long run. But whittling away your bank playing undervalued choices will.

Now that's money! Pick it up.
Richie :ThmbUp:

plainolebill
03-03-2005, 09:16 PM
Bet B, don't hesitate, no betting $30 to win and a $10 exacta box with the chalk, no tris keying B with the chalk in the other slots. Play the value. I know for me at least, when I used to play those savers, they ate up my profits over time. If you are right, get paid. If you are wrong, turn the page. No one race, or one racing day, is going to kill you in the long run. But whittling away your bank playing undervalued choices will.

Truer words have never been spoken -

LARRY GEORGE
03-03-2005, 09:46 PM
I BELIEVE MARK CRAMER SHOWED HOW HE MADE A 100% LINE IN ONE OF HIS BOOKS I DON'T REMEMBER THE NAME YET :confused:

headhawg
03-03-2005, 10:42 PM
I BELIEVE MARK CRAMER SHOWED HOW HE MADE A 100% LINE IN ONE OF HIS BOOKS I DON'T REMEMBER THE NAME YET :confused:

Both "The Odds On Your Side", and the follow-up, "Value Handicapping".