Winner top 4 how often

[Handiman]

I have as many know spent many years researching and writing homegrown computer programs for horseracing. But I have never really attempted to keep good records of how they do...not my strong point. Anyway, I decided to sit down and run a test and keep records. So now I need some input as how close I might be to having something usefull.

My test covered 17 different tracks. Two days at each track at least a month apart and two tracks had only 1 day so that's the reason for 17 tracks intstead of just 15. Those two tracks had one day each.

I checked for the winner in the top 4. I had the winner in the top 4-74% of the time. Minimum payoff was 2.80 and max was 74.00 I did not adjust for anything. Every race handicapped the same. Very low class all the way up to and including breeders cup day at Belmont.
In fact here is the day at Belmont 10-29-05....miss..13.00..6.70..11.00...miss
25.80...miss...63.50....miss...8.60.

What I'd like to know ....is 74% winner in top 4 decent...low...high? After some input I will be able to then try and figure out how to derive some value and maybe tweak things to increase the percent hit depending on class, conditions or some other varible.

Thanks for the input in advance.

Handiman :)

[Dave Schwartz]

About average with public choice... maybe 1-2 points low.

[shanta]

I have as many know spent many years researching and writing homegrown computer programs for horseracing. But I have never really attempted to keep good records of how they do...not my strong point. Anyway, I decided to sit down and run a test and keep records. So now I need some input as how close I might be to having something usefull.

My test covered 17 different tracks. Two days at each track at least a month apart and two tracks had only 1 day so that's the reason for 17 tracks intstead of just 15. Those two tracks had one day each.

I checked for the winner in the top 4. I had the winner in the top 4-74% of the time. Minimum payoff was 2.80 and max was 74.00 I did not adjust for anything. Every race handicapped the same. Very low class all the way up to and including breeders cup day at Belmont.
In fact here is the day at Belmont 10-29-05....miss..13.00..6.70..11.00...miss
25.80...miss...63.50....miss...8.60.

What I'd like to know ....is 74% winner in top 4 decent...low...high? After some input I will be able to then try and figure out how to derive some value and maybe tweak things to increase the percent hit depending on class, conditions or some other varible.

Thanks for the input in advance.

Handiman :)

Handi,
Just seeing a few of the prices you are showing you are doing TREMENDOUS. Keep doing whatever it is cause you have PRICE built into your method.

Maybe like you said spend time and isolate certain:
tracks
classes
distances etc to fine tune it a bit.

This is gonna sound crazy but if you have your top 4 graded in some kind of order go back and check the stats on your 3RD ranked horse in every race. Check these 2 things:
1) ROI and Win % on every horse regardless of odds
2) ROI and Win % on every horse 5/1 odds and up. ( this will NOT produce accurate results if you are going by result charts BUT if you have "live" odds available then you are in good shape)

Good work

My opinion
Richie :)

[traynor]

Handiman wrote: <I checked for the winner in the top 4. I had the winner in the top 4-74% of the time. Minimum payoff was 2.80 and max was 74.00 I did not adjust for anything. Every race handicapped the same. Very low class all the way up to and including breeders cup day at Belmont.
In fact here is the day at Belmont 10-29-05....miss..13.00..6.70..11.00...miss
25.80...miss...63.50....miss...8.60.

What I'd like to know ....is 74% winner in top 4 decent...low...high? >

It is unclear why you selected that specific group of races. It would be more meaningful if you selected 10 or 15 tracks, a week of races at each, then used the overall figure. Something on the order of what Kusyshyn did with harness races some years ago.

The problem is in isolating a small sample as representative of the general population (of all races). If you get consistent results, they should be roughly the same using other samples. I assume that you are using your approach to eliminate non-contenders? I am unsure of how a high percentage in the top four would be used; perhaps you could clarify that.

It would be interesting to see how your approach fares on different sets of races. You may be on to something really good, especially as a qualifier for eliminating non-contenders. The only part that seems shaky is the "top four." That is a really wide area to try to isolate a winner in. Or do you have some further refinements on selecting the winners that you have not mentioned?
Good Luck

[Jeff P]

So far this year in my 2006 database (at the tracks I have followed) the public has managed to pick 81.38 percent winners in their top 4:

By: Odds Rank

Rank Gain Bet Roi Wins Plays Pct Impact
1 -1600.20 10234.00 0.8436 1809 5117 .3535 2.9423
2 -2169.60 9620.00 0.7745 966 4810 .2008 1.6714
3 -1841.20 9696.00 0.8101 739 4848 .1524 1.2687
4 -1954.70 9640.00 0.7972 516 4820 .1071 0.8910
5 -2130.00 9572.00 0.7775 364 4786 .0761 0.6330
6 -2634.20 9108.00 0.7108 225 4554 .0494 0.4112
7 -1726.40 7918.00 0.7820 148 3959 .0374 0.3111
8 -1975.10 6244.00 0.6837 80 3122 .0256 0.2133
9 -1131.10 4456.00 0.7462 46 2228 .0206 0.1718
10 -1595.40 2944.00 0.4581 15 1472 .0102 0.0848
11 -621.20 1540.00 0.5966 7 770 .0091 0.0757
12 -737.20 790.00 0.0668 1 395 .0025 0.0211
13 -46.00 46.00 0.0000 0 23 .0000 0.0000
14 -20.00 20.00 0.0000 0 10 .0000 0.0000
15 0.00 0.00 0.0000 0 0 .0000 0.0000
16 0.00 0.00 0.0000 0 0 .0000 0.0000
17 0.00 0.00 0.0000 0 0 .0000 0.0000
18 0.00 0.00 0.0000 0 0 .0000 0.0000


I've been working on my own power rating. So far in 2006, at those same tracks - at all surfaces/distances/classes (with no further handicapping whatsoever) it has managed to get 76.45 percent winners in the top 4:
By: JPR

Rank Gain Bet Roi Wins Plays Pct Impact
1 -631.50 9910.00 0.9363 1537 4955 .3102 2.5816
2 -2009.30 9898.00 0.7970 985 4949 .1990 1.6565
3 -2226.20 9868.00 0.7744 711 4934 .1441 1.1993
4 -2479.90 9878.00 0.7489 549 4939 .1112 0.9251
5 -1623.50 9782.00 0.8340 446 4891 .0912 0.7589
6 -2326.50 9222.00 0.7477 298 4611 .0646 0.5379
7 -1568.60 7990.00 0.8037 195 3995 .0488 0.4062
8 -2707.40 6196.00 0.5630 104 3098 .0336 0.2794
9 -2022.40 4278.00 0.5273 50 2139 .0234 0.1945
10 -1372.20 2722.00 0.4959 28 1361 .0206 0.1712
11 -879.20 1338.00 0.3429 9 669 .0135 0.1120
12 -283.60 694.00 0.5914 4 347 .0115 0.0959
13 -36.00 36.00 0.0000 0 18 .0000 0.0000
14 -16.00 16.00 0.0000 0 8 .0000 0.0000
15 0.00 0.00 0.0000 0 0 .0000 0.0000
16 0.00 0.00 0.0000 0 0 .0000 0.0000
17 0.00 0.00 0.0000 0 0 .0000 0.0000
18 0.00 0.00 0.0000 0 0 .0000 0.0000

By way of comparison, in that same database, the Bris Prime Power Rating - again, with no further handicapping whatsoever - managed to pick 76.04 percent winners in the top four:
By: Prime Power Rank

Rank Gain Bet Roi Wins Plays Pct Impact
1 -1088.50 9752.00 0.8884 1578 4876 .3236 2.6934
2 -1847.80 9776.00 0.8110 964 4888 .1972 1.6414
3 -2467.80 9870.00 0.7500 662 4935 .1341 1.1164
4 -2013.00 9914.00 0.7970 523 4957 .1055 0.8781
5 -1912.80 9898.00 0.8067 440 4949 .0889 0.7399
6 -2686.00 9440.00 0.7155 290 4720 .0614 0.5113
7 -1780.00 8186.00 0.7826 208 4093 .0508 0.4229
8 -2662.40 6294.00 0.5770 119 3147 .0378 0.3147
9 -1669.00 4248.00 0.6071 77 2124 .0363 0.3017
10 -1044.40 2658.00 0.6071 39 1329 .0293 0.2442
11 -778.60 1200.00 0.3512 12 600 .0200 0.1665
12 -180.00 540.00 0.6667 4 270 .0148 0.1233
13 -40.00 40.00 0.0000 0 20 .0000 0.0000
14 -12.00 12.00 0.0000 0 6 .0000 0.0000
15 0.00 0.00 0.0000 0 0 .0000 0.0000
16 0.00 0.00 0.0000 0 0 .0000 0.0000
17 0.00 0.00 0.0000 0 0 .0000 0.0000
18 0.00 0.00 0.0000 0 0 .0000 0.0000

[Handiman]

Thanks for the input. To clarify a bit. I choose the races I did at random. I wanted a diverse group of tracks, and I wanted races run several weeks apart. I didn't want to cherry pick or appear as though I did.

The horses are ranked based on a power number I developed. It was made to find contenders. I had dreamed of writing a program and selling it to a million people. And I have put a couple of programs out for sale. But I have finally figured out, that I don't have the intestinal fortitude to continue to upgrade and support the software over several years. So I have basically trashed that idea. That being said, I have the utmost admiration for Dave Schwartz and the other members who market their projects and keep servicing them from year to year.

I have actually found it more rewarding taking on projects for other people and writing specific programs for them. I have one project going now, another one lined up and an upgrade to another.

This one I'm asking about right now though is for my personal use.

Thanks again for the comments.

Handi

[Jeff P]

Couple of comments: First, the public is a pretty good opponent. Want evidence of that? Just look at how horses fared when ranked 12th or worse as the public choice in the above sample. Second, the power rating I'm working on represents a LOT of man hours and research - as did (I'm just guessing here) the Bris Prime Power Rating. So if you're already getting 74 percent winners in your top four, and you're getting prices as well - then I'd say you're off to a pretty good start and you are probably doing more than a few things right.


-jp

.

[Murph]

Here's thorostats top 4 win% for the past 365 days for you to compare.

Top 4 Horses

class 44339 29063 65%
pace 44339 30482 68%
power 44338 31753 71%
conn. 44340 30077 67%


Murph

[Dave Schwartz]

ALthough these are just single factors and do not represent the output of a strategy, here are a few hot factors from HSH v3. (Please note that the totals are win percetnage and not percentage of winners.)



rPSR
----
WIN BETS
Field1 Field2 Starts Pays Pct $Net IV PIV
--------------------------------------------------------
1st 25,711 7,184 27.9 $1.72 2.18 1.02
2nd 20,941 4,200 20.1 $1.71 1.56 1.01
3rd 22,749 3,418 15.0 $1.67 1.17 0.97
4th 23,024 2,702 11.7 $1.60 0.92 0.94

Wins: 17,504
Starts: 92,425
Avg Win%: 18.93%
x 4 starters: 75.75%


rcRtg
-----
WIN BETS
Field1 Field2 Starts Pays Pct $Net IV PIV
--------------------------------------------------------
1st 26,808 7,121 26.6 $1.65 2.08 1.00
2nd 25,129 4,715 18.8 $1.67 1.47 1.00
3rd 23,496 3,431 14.6 $1.66 1.14 0.99
4th 23,018 2,602 11.3 $1.59 0.89 0.95

Wins: 17,869
Starts: 98,451
Avg Win%: 18.15%
x 4 starters: 72.60%


rcFT
----
WIN BETS
Field1 Field2 Starts Pays Pct $Net IV PIV
--------------------------------------------------------
1st 25,851 6,587 25.5 $1.66 1.99 1.00
2nd 24,466 4,647 19.0 $1.69 1.48 1.01
3rd 23,671 3,487 14.7 $1.64 1.15 0.98
4th 23,070 2,692 11.7 $1.57 0.91 0.95

Wins: 17,413
Starts: 97,058
Avg Win%: 17.94%
x 4 starters: 71.76%


rcSC
----
WIN BETS
Field1 Field2 Starts Pays Pct $Net IV PIV
--------------------------------------------------------
1st 26,003 6,634 25.5 $1.69 1.99 1.02
2nd 24,773 4,568 18.4 $1.67 1.44 1.00
3rd 23,570 3,511 14.9 $1.67 1.17 1.00
4th 22,966 2,668 11.6 $1.59 0.91 0.94

Wins: 17,381
Starts: 97,312
Avg Win%: 17.86%
x 4 starters: 71.44%



rcEP
----
WIN BETS
Field1 Field2 Starts Pays Pct $Net IV PIV
--------------------------------------------------------
1st 26,127 6,519 25.0 $1.74 1.95 1.03
2nd 24,668 4,500 18.2 $1.70 1.43 1.00
3rd 23,782 3,414 14.4 $1.61 1.12 0.96
4th 23,010 2,823 12.3 $1.68 0.96 0.98

Wins: 17,256
Starts: 97,587
Avg Win%: 17.68%
x 4 starters: 70.73%

[Murph]

So would you say that 70% or better top 4 is good enough to qualify profits from a medium sized sample, Dave? Our method will often show the second or third selection as the highest win% depending on the dataset, usually with better odds.

I think I remember Jeff saying that is true in his studies as well. Sometimes it's not the top selection that is the best %age wager.

Murph

[Dave Schwartz]

I would say that we don't have enough information to make that call.

IMHO, the ability to show profit has more to do with one's ability to exploit the individual race than one's ability to get lots of winners in the top X horses.

I rather liked your 100-race output in the other thread, btw. Very similar to my own approach (i.e. "What wins a race like this one?"). Of course, profit is dependant upon how you apply what you glean from that 100-race sample. If anyone thinks they are simply going to bet the best column and just wait for the money to engulf them... will, they will probably have a long wait.

Dave

[Jeff P]

I think I remember Jeff saying that is true in his studies as well. Sometimes it's not the top selection that is the best %age wager. My own experience tells me the ability to consistently grind out profits has more to do with distinguishing good bets from bad bets than anything else. The fact that a horse is or isn't in the top four of a contender selection schema doesn't mean that horse is a good bet. It really depends on the situation. I have some models that are consistently profitable using any of the top six contenders IF both the odds and the situation are right. Of course it goes without saying that the better the numbers you have to work with the easier it becomes for you to recognize and make good bets.

-jp

.