Preliminary Software Results

[tanda]

I have posted several times with results from testing of software that I am designing.

Here is another report:

Rs Ws W% Odds ROI IV
Favorite 415 137 33.0% 184.85 0.78 2.71
230 61 26.5% 110.85 0.75 2.18
43 9 20.9% 17.55 0.62 1.72
2 0 0.0% 0.00 0.00 0.00
690 207 30.0% 313.25 0.75 2.47

Short 208 55 26.4% 157.00 1.02 2.17
Price
466 106 22.7% 313.15 0.90 1.87
196 34 17.3% 101.65 0.69 1.43
36 7 19.4% 27.60 0.96 1.60
906 202 22.3% 599.40 0.88 1.83

Mid 60 6 10.0% 31.40 0.62 0.82
Price
572 82 14.3% 430.45 0.90 1.18
556 69 12.4% 395.70 0.84 1.02
158 12 7.6% 83.40 0.60 0.62
1346 169 12.6% 940.95 0.82 1.03

Longshot 6 0 0.0% 0.00 0.00 0.00
321 20 6.2% 221.10 0.75 0.51
1054 66 6.3% 1058.30 1.07 0.51
1347 28 2.1% 537.40 0.42 0.17
2728 114 4.2% 1816.80 0.71 0.34

Total 5670 692 12.2% 3670.40 0.77 1.00
Value 2336 316 13.5% 1986.50 0.99 1.11
Non-Value3334 376 11.3% 1683.90 0.62 0.93

Quite frankly, I am very excited with these results. I am unaware of any mechanical method that identifies a class of horses composing over 40% of the field who return 0.99 on the dollar.

First, I have grouped the horses by odds range with sub-totals. Within each odds range, the horses are sub-categorized by the handicapping methodology that I have developed and coded.
Second, it is clear that there are identifiable groups of horses who are money burners and out-performers. I have designated those that significantly out-perform the take and breakage as "value" and the others as "non-value". The "non-value" horses are not necessarily non-contenders; they win 0.93 of their fair share of races. But, they return 0.62. More importantly, a flat bet on all "value" horses returned 0.99.

Third, this method is completely objective. The numbers are inputed/imported to the software which designates each horses category. There is no user subjectivity. All races with each horse having one US past performance line were played at all tracks over a period of days .

Fourth, it appears that a handicapper who restricts action to "value" horses will not have to do much to increase the 0.99 to a positive R.O.I.

Fifth, the methodology does not end with the "value"/"non-value" designation. Instead, it categorizes the horses then makes betting decisions from among the "value" horses. I have not tested those betting decisions; that test is commencing now. An oddline is created to make betting decisions. The test will be of: 1) 20% overlays, 2) all overlays, 3) the largest overlay (or smallest underlay) and 4) two horse bets on the largest overlays/smallest underlays.

[ranchwest]

Are you trying to predict value? If so, look out! The tote board watchers will be after you!

[tanda]

Sorry about the formatting. When I was importing it to the post, it appeared orderly.

Ranchwest:

Actually, I was conducting a test of longshots and widened it. I am encouraged because of all the horrendous wagers that can be avoided just by concentrating on the "value" category. By "value", I mean that this category is much, much more likely to contain runners with a positive expectation but obvioulsy not all are value. The "non-value" category cannot contained many positive expectation wagers, so it should be avoided.

[andicap]

so this was multi horse betting? Betting all the "value" horses per race on a mechanical basis?

[ranchwest]

Originally posted by tanda
Sorry about the formatting. When I was importing it to the post, it appeared orderly.

Ranchwest:

Actually, I was conducting a test of longshots and widened it. I am encouraged because of all the horrendous wagers that can be avoided just by concentrating on the "value" category. By "value", I mean that this category is much, much more likely to contain runners with a positive expectation but obvioulsy not all are value. The "non-value" category cannot contained many positive expectation wagers, so it should be avoided.

Ah, so you are predicting value? Imagine that.

[GameTheory]

You say this was for all tracks? Was it for all types of races as well? (You said all races where there was at least one line for each horse, so I assume yes, or were you just looking at certain types/surfaces/distances?)

Are you finding similar results at all tracks? Getting the same performance (with the same method) at both Mountaineer and Aqueduct, for instance, would be impressive indeed. (Or maybe different formulas are used for different tracks?)


In my experience creating computer methods, I have not that it is not that difficult to break even (betting one or several horses a race, depending on the method). I have found in my recent study of claiming sprints that at many tracks, simply betting the last race best (normalized) time is enough to do that. (Other tracks, the last race is virtually meaningless if taken alone.) Making money is a bit tougher. When you are close to the break even point, you need a giant sample to really tell what's going on because just a couple wins or losses here or there move you back and forth over the line easily. However, I have never done an all-encompassing all tracks/all races study as I never had the resources. I am nearly in a position to do that now, so I'm interested in your results.

[tanda]

Game Theory:

All tracks, all races except those with runners with no PP lines. 690 races over a 6-7 day period.

You implied it would not be difficult to break-even, if I read your post correctly. Note that a flat bet on 41% of the runners in all races with the small exception noted yielded 0.99. A flat bet on all runners returned 0.77. I have never seen results of 0.99 for a mechanical system betting over 3 horses a race in almost all races at all tracks. Every race had at least one bet and the average was over 3.

[GameTheory]

The 41% of the field and the all tracks/all races part is the rub, and I can't really comment on how hard that is cause I haven't tried it. But it doesn't surprise me. (I don't mean that as a knock.)

Multiple horse betting often makes it easier to close in on profitability, which many people find surprising. So that makes sense. In other words, I think you'd have a harder time breaking even choosing only one horse per race. (Not that there is any reason to do that.)

Anything I come up with is usually geared to a specific distance structure / surface / race class. Like now I'm looking at dirt claiming sprints, which are very plentiful but are one of the tougher spots (for me, anyway) because the public seems to have the best handle on them. Maidens I find easiest because you can often easily eliminate much of the field even though faves win more.


How about differences between tracks? I find worlds of difference between some tracks...

[PaceAdvantage]

Tanda,

I went in and fixed your post. If you use the CODE tags, then the message board will leave the format as is, and won't try and change it....


==PA


PS In order to use the tags, simply place the word CODE surrounded by [] at the beginning of the text you want to be left alone, and place /CODE surrounded by [] at the end of the text.

[Aussieplayer]

Tanda,

You're doing what I'm doing!!!
Current method (others are being studied of course), being bet is that multiple horse thing I mentioned and recieved wonderful help on from Dick & others. This was originally just a research thing. I noticed that horses not in this initial group were hopeless. Within this (big) group that averaged most of the field, was a subset. This subset had a hit rate of 40% and an ROI of 10%.
After doing a breakdown by number of horses bet (anywhere from 1- 9), I noticed a strange thing. When there were 1-4 horses to bet, this group CONSISTENTLY produced a loss. It was when there were 5 or more horses selected that the ROI was there. (Average of 6 horses per race in this group).

I started betting & things are going great. The benefit, or "lucky" part about this group being profitable as opposed to the 1-4 horses group is that the hit rate is over 50%. In fact, it's (currently) over 70%. The ROI is over 10% due to eliminating the unprofitable 1-4 group.

Maybe you want to do a break down like this???

Being a conservative sort, I was pretty harsh on it. But EACH group: ie. when 1 horse selected, 2 horses, 3 horses or 4 horses were unprofitable.

What happens of course, is that you slowly bleed away, getting back maybe 3/1 after you've outlayed maybe 6 units, then a 20/1 or 30/1 shot hits.
Two points: (a) it's the way to get the longshots. If you started doing traditional things like ranking etc. you can easily miss these.
(b)....sure is NICE for a change hitting MORE races than you lose!!! It's freaky!!!!!!!!!!!!!!!!!!!! :D

As the (small) bank grows I guess I'm gonna have to take Dick out to tea or something
:p

AP