bris_power_figure beats 'fair" odds

[formula_2002]

This is a look at over 24,000 races.

If 'fair" odds, is the track odds normalized,
(that is (1/(odds+1))/(sum (1/odds+)) then the top ranked bris power figure won more races than expected.

The top ranked figure won 7,100 races.
The fair odds sum is 6,485.


Unfortunately, this is not a fair game.
You would have to win 7990 races just to break even using the top bris power fig.
(sum 1/(odds+1) w/o normalizing = 7990)


This is a dutch bet analysis where 1/(odds+1)= expected wins.

[BetHorses!]

What about 2nd and 3rd choices?

[Wizard of Odds]

This is a look at over 24,000 races.

If 'fair" odds, is the track odds normalized,
(that is (1/(odds+1))/(sum (1/odds+)) then the top ranked bris power figure won more races than expected.

The top ranked figure won 7,100 races.
The fair odds sum is 6,485.


Unfortunately, this is not a fair game.
You would have to win 7990 races just to break even using the top bris power fig.
(sum 1/(odds+1) w/o normalizing = 7990)

This is a dutch bet analysis where 1/(odds+1)= expected wins.


Mindlessly choosing all 4/5 and lower odds horses will also beat their "fair odds" and will still not break even. At about 1/5 odds and less, we begin to break even.

All this assumes normal track odds and no rebate.


Regards,
Wizard

[formula_2002]

What about 2nd and 3rd choices?


As with almost every Bris or Always generated figure, it's all down hill from the top.
Off hand, I don't recall any that improves as a stand alone factor.

[formula_2002]

Mindlessly choosing all 4/5 and lower odds horses will also beat their "fair odds" and will still not break even. At about 1/5 odds and less, we begin to break even.

All this assumes normal track odds and no rebate.


Regards,
Wizard


that is true, and the top Bris Power figure still out performs the random pick of <=4/5

[Wizard of Odds]

that is true, and the top Bris Power figure still out performs the random pick of <=4/5

I believe it is probably true that the M/L favorite (at major tracks) will also beat their 'fair odds' and still not break even because of the takeout. BRIS might still outperform the M/L favorite, but probably not by much.


Regards,
Wizard

[classhandicapper]

Has anyone studied the win percentage and ROI for their top rated horse based on their class rating?

How does it compare to some commonly used speed figures including their own.

[Overlay]

Overall results from Nunamaker's Modern Impact Values for rank based on average of BRIS class ratings for last three races versus average of BRIS speed ratings for last two races:

Class

1 2.14 impact value; 26.08 W%; $1.77 ROI
2 1.55; 18.86%; $1.77
3 1.16; 14.20%; $1.64
Front Half (but not in Top Three) 1.01; 9.82%; $1.69
Rear Half .56; 6.45%; $1.33

Speed

1 2.15; 26.20%; $1.78
2 1.52; 18.60%; $1.75
3 1.15; 14.01%; $1.68
Front Half .94; 9.18%; $1.61
Rear Half .58; 6.69%; $1.34

[midnight]

When I was using the trial of JCapper, I did a lot of work with the database re the BRIS numbers. They all produced pedestrian results, and the best I could do was 0.88 ROI on any of it. The further down in rank the factor was, the lower the ROI.

[Speed Figure]

These were posted by Overlay. I'm Just trying to line it up.



Class
impact value W% ROI
1 2.14 26.08% $1.77
2 1.55 18.86% $1.77
3 1.16 14.20% $1.64
Front Half 1.01 9.82% $1.69
Rear Half .56 6.45% $1.33

Speed

1 2.15 26.20% $1.78
2 1.52 18.60% $1.75
3 1.15 14.01% $1.68
Front Half .94 9.18% $1.61
Rear Half .58 6.69% $1.34

[Jeff P]

posted by midnight - When I was using the trial of JCapper, I did a lot of work with the database re the BRIS numbers. They all produced pedestrian results, and the best I could do was 0.88 ROI on any of it. The further down in rank the factor was, the lower the ROI. Bris generated numbers (Prime Power/Bris Speed Figs, etc) are seen by far too many pairs of eyes each day to be effective. You need to do something different than the crowd if you want to have any chance of success.


posted by - classhandicapper Has anyone studied the win percentage and ROI for their top rated horse based on their class rating?

How does it compare to some commonly used speed figures including their own.

First 3 Quarters 2005, every race at all tracks that I've downloaded, here's what I have based on a flat $2.00 win bet for all starters broken out by rank for JRating:

Data Window Settings:
Divisor = 999
Surface: (ALL*) Distance: (All*) (From Index File: D:\2005\Q3_2005\pl_Complete_History.txt)


Data Summary Win Place Show
Mutuel Totals 234017.70 232646.90 227901.90
Bet -307492.00-307492.00-307492.00
Gain -73474.30 -74845.10 -79590.10

Wins 19409 38596 56968
Plays 153746 153746 153746
PCT .1262 .2510 .3705

ROI 0.7611 0.7566 0.7412
Avg Mut 12.06 6.03 4.00



By: JRating Rank

Rank Gain Bet Roi Wins Plays Pct Impact
1 -1637.20 38386.00 0.9573 4919 19193 .2563 2.0302
2 -5849.30 38376.00 0.8476 3598 19188 .1875 1.4854
3 -7262.60 38394.00 0.8108 2839 19197 .1479 1.1715
4 -7887.60 38304.00 0.7941 2411 19152 .1259 0.9972
5 -9392.40 37818.00 0.7516 1949 18909 .1031 0.8165
6 -10489.50 35138.00 0.7015 1471 17569 .0837 0.6632
7 -11043.10 28892.00 0.6178 945 14446 .0654 0.5182
8 -7216.40 21046.00 0.6571 624 10523 .0593 0.4697
9 -4935.50 13750.00 0.6411 338 6875 .0492 0.3894
10 -3728.80 8596.00 0.5662 175 4298 .0407 0.3225
11 -1895.00 4380.00 0.5674 77 2190 .0352 0.2785
12 -1271.20 2648.00 0.5199 33 1324 .0249 0.1974
13 -362.80 794.00 0.5431 13 397 .0327 0.2594
14 -306.90 724.00 0.5761 13 362 .0359 0.2845
15 -103.00 128.00 0.1953 1 64 .0156 0.1238
16 -98.20 106.00 0.0736 2 53 .0377 0.2989
17 15.20 2.00 8.6000 1 1 1.0000 7.9214
18 -6.00 6.00 0.0000 0 3 .0000 0.0000
19 -4.00 4.00 0.0000 0 2 .0000 0.0000




I'm also currently working on a JCapper Power Rating (JPR.) First 3 Quarters this year, based on a flat $2.00 win bet on every starter, I have the following:

By: JPR Rank

Rank Gain Bet Roi Wins Plays Pct Impact
1 -2780.70 38356.00 0.9275 5712 19178 .2978 2.3593
2 -4279.80 38364.00 0.8884 3951 19182 .2060 1.6316
3 -5556.60 38398.00 0.8553 2976 19199 .1550 1.2279
4 -8811.80 38352.00 0.7702 2180 19176 .1137 0.9005
5 -10084.30 38024.00 0.7348 1703 19012 .0896 0.7096
6 -10112.10 35694.00 0.7167 1193 17847 .0668 0.5295
7 -11759.60 29796.00 0.6053 768 14898 .0516 0.4084
8 -9134.80 22038.00 0.5855 455 11019 .0413 0.3271
9 -5324.70 14534.00 0.6336 284 7267 .0391 0.3096
10 -3325.10 8586.00 0.6127 131 4293 .0305 0.2417
11 -1895.90 3684.00 0.4854 39 1842 .0212 0.1677
12 -376.60 1430.00 0.7366 14 715 .0196 0.1551
13 57.70 146.00 1.3952 3 73 .0411 0.3255
14 -76.00 76.00 0.0000 0 38 .0000 0.0000
15 -8.00 8.00 0.0000 0 4 .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 -6.00 6.00 0.0000 0 3 .0000 0.0000
19 0.00 0.00 0.0000 0 0 .0000 0.0000

-jp

[Tom]

So, what if you eliminate thowe in the rear half of the filed in terms of either factor - then re-do the race with 50% less horses in it. You should stilll be dealing with about 94% of the winners, but what do the numbers look like with all thoses not-bloody likely horses out of it?

I love elimination factors that get the race down to a vital few.

[Jeff P]

First 3 Quarters this year, based on a flat $2.00 win bet on every starter, here's what I have broken out by Bris Prime Power:

By: Prime Power Rank

Rank Gain Bet Roi Wins Plays Pct Impact
1 -5808.20 39074.00 0.8514 5974 19537 .3058 2.4222
2 -4800.90 39192.00 0.8775 4093 19596 .2089 1.6545
3 -5957.00 39600.00 0.8496 2978 19800 .1504 1.1914
4 -7809.50 39346.00 0.8015 2218 19673 .1127 0.8931
5 -9947.90 38980.00 0.7448 1571 19490 .0806 0.6385
6 -10796.10 35578.00 0.6966 1078 17789 .0606 0.4800
7 -9689.30 29082.00 0.6668 697 14541 .0479 0.3797
8 -8619.50 20274.00 0.5748 417 10137 .0411 0.3259
9 -3812.70 13108.00 0.7091 226 6554 .0345 0.2732
10 -3698.60 7638.00 0.5158 91 3819 .0238 0.1888
11 -1784.00 3686.00 0.5160 43 1843 .0233 0.1848
12 -635.30 1628.00 0.6098 18 814 .0221 0.1752
13 -135.70 216.00 0.3718 3 108 .0278 0.2200
14 -70.20 78.00 0.1000 1 39 .0256 0.2031
15 -2.00 2.00 0.0000 0 1 .0000 0.0000
16 100.60 2.00 51.3000 1 1 1.0000 7.9214
17 -2.00 2.00 0.0000 0 1 .0000 0.0000
18 -2.00 2.00 0.0000 0 1 .0000 0.0000
19 -4.00 4.00 0.0000 0 2 .0000 0.0000

-jp

.

[formula_2002]

with a few simple handicapping rules and the top ranked Bris_power fig;

back fitted results showed a 1% win pool profit in 390 races.

Out of sample profit was 1% for 892 races!!

In an additional 116 races, from nov 1 to date,
there was a 45 cent profit.

All the above covers races from 2002 to date

Joe M

[Overlay]

These were posted by Overlay. I'm Just trying to line it up.

Thanks for the formatting help. :)

[Jeff P]

poted by Tom - So, what if you eliminate thowe in the rear half of the filed in terms of either factor - then re-do the race with 50% less horses in it. You should stilll be dealing with about 94% of the winners, but what do the numbers look like with all thoses not-bloody likely horses out of it?

I love elimination factors that get the race down to a vital few.
What? You mean I have to start handicapping now? :D :D

Here's what I have after removing the bottom half of the field for each rating. Interestingly enough, this raised the roi for all the horses remaining to something higher than what I have for the top ranked Bris Prime Power horse. I'm fairly certain there's more than a few paths I could go down from here to find some new solidly profitable models.

Data Window Settings:
999 Divisor
Filters Applied: 2HLFJPR-2HLFJ
Surface: (ALL*) Distance: (All*) (From Index File: D:\2005\Q4_2005\pl_Complete_History.txt)
Data Summary Win Place Show
Mutuel Totals 98030.30 98083.60 96684.80
Bet -110416.00-110416.00-110416.00
Gain -12385.70 -12332.40 -13731.20

Wins 11412 20896 28471
Plays 55208 55208 55208
PCT .2067 .3785 .5157

ROI 0.8878 0.8883 0.8756
Avg Mut 8.59 4.69 3.40


By: JRating Rank

Rank Gain Bet Roi Wins Plays Pct Impact
1 -1186.50 34100.00 0.9652 4677 17050 .2743 1.3270
2 -3174.10 30394.00 0.8956 3228 15197 .2124 1.0276
3 -3758.40 24868.00 0.8489 2154 12434 .1732 0.8381
4 -2761.20 13996.00 0.8027 977 6998 .1396 0.6754
5 -1161.10 5678.00 0.7955 309 2839 .1088 0.5265
6 -299.40 1318.00 0.7728 66 659 .1002 0.4845
7 -43.00 60.00 0.2833 1 30 .0333 0.1613
8 -2.00 2.00 0.0000 0 1 .0000 0.0000
9 0.00 0.00 0.0000 0 0 .0000 0.0000
10 0.00 0.00 0.0000 0 0 .0000 0.0000
11 0.00 0.00 0.0000 0 0 .0000 0.0000
12 0.00 0.00 0.0000 0 0 .0000 0.0000
13 0.00 0.00 0.0000 0 0 .0000 0.0000
14 0.00 0.00 0.0000 0 0 .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
19 0.00 0.00 0.0000 0 0 .0000 0.0000




By: JPR Rank

Rank Gain Bet Roi Wins Plays Pct Impact
1 -2330.60 34544.00 0.9325 5257 17272 .3044 1.4724
2 -3313.50 30842.00 0.8926 3188 15421 .2067 1.0001
3 -2765.10 24540.00 0.8873 1934 12270 .1576 0.7625
4 -2842.80 13772.00 0.7936 741 6886 .1076 0.5206
5 -820.20 5316.00 0.8457 245 2658 .0922 0.4459
6 -227.50 1316.00 0.8271 47 658 .0714 0.3456
7 -84.00 84.00 0.0000 0 42 .0000 0.0000
8 -2.00 2.00 0.0000 0 1 .0000 0.0000
9 0.00 0.00 0.0000 0 0 .0000 0.0000
10 0.00 0.00 0.0000 0 0 .0000 0.0000
11 0.00 0.00 0.0000 0 0 .0000 0.0000
12 0.00 0.00 0.0000 0 0 .0000 0.0000
13 0.00 0.00 0.0000 0 0 .0000 0.0000
14 0.00 0.00 0.0000 0 0 .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
19 0.00 0.00 0.0000 0 0 .0000 0.0000

-jp

.

[formula_2002]

Jeff P

I have found similar results.

see the "Bris Report" on my home page for over 100,000 horses

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

[formula_2002]

This is a look at over 24,000 races.

If 'fair" odds, is the track odds normalized,
(that is (1/(odds+1))/(sum (1/odds+)) then the top ranked bris power figure won more races than expected.

The top ranked figure won 7,100 races.
The fair odds sum is 6,485.


Unfortunately, this is not a fair game.
You would have to win 7990 races just to break even using the top bris power fig.
(sum 1/(odds+1) w/o normalizing = 7990)


This is a dutch bet analysis where 1/(odds+1)= expected wins.

using cj's top fig offers similar results;

This is a look at over 14,000 races.

If 'fair" odds, is the track odds normalized,
(that is (1/(odds+1))/(sum (1/odds+)) then the top ranked cj figure won more races than expected.

The top ranked figure won 3,797 races.
The fair odds sum is 3,501.



You would have to win 4325 races just to break even using the top cj fig.
(sum 1/(odds+1) w/o normalizing = 4325)


bris fair odds advantage 7100/6485 = 1.095
cj fair odds advantage 3779/3501=1.08

[bettheoverlay]

I have been using the Bris PP #s in real life betting situations for over 5 years. I have never been able to find a way to use the top # for profit. And believe me I tried!

The 2nd and 3rd numbers are a different story. I use the 2nd at 6/1 up and the third at 8/1 and up unless its within 5 of the top, then 6/1. I throw out races in which the top # is 10 pts higher than the second. If the top # is 5 pts higher I only consider the second #. Of the horses that qualify odds wise, I probably throw out almost half, as I look for horses with good recent form.

Im really just following the simple adage taught me by my neighbor 30 years ago when he told me to bet horses with good recent form going off at high odds. I just have much better success when they are also the 2nd and 3rd PP#. I tried this approach with the e-ponies computer rankings for awhile, but they presented far fewer high odds horses in the 2nd and 3rd slot.

[cj]

My ratings measure each race as one performance, the Prime Power ratings are measuring a whole bunch of stuff over many races. Comparing them is apples to oranges.

[formula_2002]

My ratings measure each race as one performance, the Prime Power ratings are measuring a whole bunch of stuff over many races. Comparing them is apples to oranges.

Hey, apples and oranges are generated from the same stuff also.
Basic stuff like sun, earth etc… :cool:
The figs are really all based on the same stuff, just that some come out like oranges and some come out like apples..

[midnight]

Generally a welll-cosntructed factor (such as the BRIS rating, HTR's K rating, etc.) that is more-or-less based on conventional handicapping factors, e.g. things that John Q. Public looks at, will have roughly the following:

Top 1 30%
Top 2 50%
Top 3 63%
Top 4 72%
Top 5 78%

The average field size being about 9 horses, that means that the bottom four horses will win 22% of the time. In that 22% will be most of the longshots. As Jeff said, if you want to be ahead of the crowd, you have to do something that they don't do.

One of the things I judge programs/methods based on how the horses do, relative to their odds and their rankings in the various factors. In other words, if a lot of horses which aren't listed in the program's top four selections are getting off at 4-1 or lower and winning or running close, then I know the program is missing something. On the other hand, if the program is picking a lot of 6-1 or higher horses which are winning or running well, then I have reason to believe the program is onto something. In that way, I can get a feel for how well the program is doing without using an extremely large sample. I may get the wrong impression, but as a rule, I have a fairly good idea.

[classhandicapper]

Thanks. Amazingly similar!


Overall results from Nunamaker's Modern Impact Values for rank based on average of BRIS class ratings for last three races versus average of BRIS speed ratings for last two races:

Class

1 2.14 impact value; 26.08 W%; $1.77 ROI
2 1.55; 18.86%; $1.77
3 1.16; 14.20%; $1.64
Front Half (but not in Top Three) 1.01; 9.82%; $1.69
Rear Half .56; 6.45%; $1.33

Speed

1 2.15; 26.20%; $1.78
2 1.52; 18.60%; $1.75
3 1.15; 14.01%; $1.68
Front Half .94; 9.18%; $1.61
Rear Half .58; 6.69%; $1.34

[formula_2002]

Thanks. Amazingly similar!

These results are also simliar for the other 80+ factors in the All-ways/ Bris data files

[classhandicapper]

One of the things I judge programs/methods based on how the horses do, relative to their odds and their rankings in the various factors. In other words, if a lot of horses which aren't listed in the program's top four selections are getting off at 4-1 or lower and winning or running close, then I know the program is missing something. On the other hand, if the program is picking a lot of 6-1 or higher horses which are winning or running well, then I have reason to believe the program is onto something. In that way, I can get a feel for how well the program is doing without using an extremely large sample. I may get the wrong impression, but as a rule, I have a fairly good idea.

I think that's a good technique.

There's an extention of that IMO helps in the learning process. Some people just look at their top rated horse when analyzing results. If it wins, they assume the figures accurately picked the most likely winner.

I like to look at the results of races and make subjective judgments about what actually happened. I want to know if I picked the right horse for the right reason or if my figures were wrong and I just got lucky.

For example:

Assume my top number was the 7-1 3rd choice. Assume the favorite and 2nd choice (who I rated lower) got into a savage duel, my horse got a perfect trip sitting off them, and won by a neck.

I may have picked the winner, but there could be something wrong with my ratings! Perhaps I should have rated the favorite and 2nd choice higher because they were obviously tons the best, the betting public understood that, and it seems as though I just got lucky.

Now if I predicted that duel, that would be a different story.

This is not perfect either because their are subjective issues and horses form changes etc... But when I pick a high priced horse that wins, I like to know that I actually beat the favorites when they ran well and not because they didn't fire a good shot or some other trip related issue. My grandmother picks longshot winners here or there.

You want to have the right horse for the right reason.

[mcikey01]

Jeff P.

I am unclear why the results for your new Jcappper Power Rating (JPR) show a decrease in the the number of winners in both Rank 1 and Rank 2 when the bottom half of the filed in JPR Ran is excluded from the anlaysis....Aren't the top two ranks always in the top half of the field in all races? Or did you exclude races where there were ties in Rank 1 and/or Rank 2 or other ranks?

I am very interested in the general methodology utilized by you to develop the JPR, including the weighting scheme for the key factors (without you revealing the ID of proprietary handicapping elements comprising the rating). I currently utilize a set of power- type ratings available from another online vendor for which the component elements of the overall rating and the weighting mechanism are available for user review and customization, if desired.... I'd be interested in sharing ideas in this area...

Do you have plans to offer the JPR as a stand alone rating for daily races at one or more tracks, available for purchase independent of your JCapper software?

[Jeff P]

Jeff P.

I am unclear why the results for your new Jcappper Power Rating (JPR) show a decrease in the the number of winners in both Rank 1 and Rank 2 when the bottom half of the filed in JPR Ran is excluded from the anlaysis....Aren't the top two ranks always in the top half of the field in all races? Or did you exclude races where there were ties in Rank 1 and/or Rank 2 or other ranks? My original post on the previous page shows JPR for all starters in the database as:

JPR Win
Rank Percent
1 .2978
2 .2060
3 .1550
4 .1137
TOTAL .7725


The next post (on this page) for JPR shows what happens when horses ranked in the bottom half of the field for either JRating or JPR are filtered out. After doing this elimination, I'm left with:

JPR Win
Rank Percent
1 .3044
2 .2067
3 .1576
4 .1076
TOTAL .7763

JRating and JPR are different enough from each other that in some instances the top JPR horse might be ranked in the bottom half of the field for JRating and vice versa. That's why there are fewer plays in the second sample for both JPR and JRating than in the first post.

I am very interested in the general methodology utilized by you to develop the JPR, including the weighting scheme for the key factors (without you revealing the ID of proprietary handicapping elements comprising the rating). I currently utilize a set of power- type ratings available from another online vendor for which the component elements of the overall rating and the weighting mechanism are available for user review and customization, if desired.... I'd be interested in sharing ideas in this area... I'm basically evaluating hundreds of different factors one factor at a time. For each factor I'm calculating something I call a "relative strength" which actually is a combined impact value for both win percent and flat win bet roi. Let's say I take a factor, call it FactorA, and run it through the database. I see that the top ranked horse for FactorA has a win rate of .29 and the flat bet win roi to one dollar is .88. I use each to calculate an impact value using all horses in the database as a benchmark. In the above example let's say that win rate for all starters in the database is .1265 and flat bet win roi for all horses in the database is .7625. I would calculate relative strength for FactorA as follows:

RSa = (.29/.1265) x (.88/.7625)

or

RSa = 2.6458

I then multiply the relative strengths for each of the individual factors used in the mix together to arrive at the power rating.

JPR = RSa x RSb x RSc x RSd x RSe...

Right now each separate factor is weighted equally. But that could change in time as I get more and more into developing it. I'm still in the early stages...

Some keys to doing this successfully, at least for me, are to use factors that have relatively strong roi components to them - those that the public consistently overlooks. For example, the ML Favorite has a very high impact value for win percent. But I don't use it because it has a relatively weak roi component. A lot of the factors I use are ranks and gaps for numbers that I created myself. Typically these have a relatively strong roi component simply because outsiders can't get a handle on them. I think it's also important to stick to causal factors - those I can conceptualize as shaping the outcome of the race in a logical way that makes sense to me. Doing this, even with a database and solid programmnig skills, takes an incredible amount of research time. For every one factor that finds its way into the mix there are hundreds that I dropped because they tested poorly.


Do you have plans to offer the JPR as a stand alone rating for daily races at one or more tracks, available for purchase independent of your JCapper software?No.

JCapper software users will be the only ones with access to it. I do not want this or any other JCapper number available outside the software. Right now I'm working to have JPR integrated into JCapper (hopefully) by mid year 2006. Even though results I've posted here appear very promising, I'm still in the development stage. If I put JPR into JCapper right now it's very possible that quite a few people would be happy with it as is. Because I can see what's possible as I develop this I'm going to keep plugging away, working on improving it a little bit at a time until development time spent on it starts producing only miniscule gains.

Earlier this year I gave another vendor permission to use a handful of JCapper numbers on a report he publishes. I believe I made a mistake in doing so. Earlier this year I also began selling my own daily JCapper report on a subscription basis. I did this at the request of a handful of individuals who approached me wanting my information but not my software. At the time it seemed like a good idea. I was even a little flattered that some pretty heavy hitters had sought me out. Even though the subscription money is nice I believe this was a mistake. A few paying subscribers began emailing copies of my report all over the place. This realization recently hit home when someone I met at DMR (who isn't a paying customer) had a laptop with a copy of my HTML Report on the screen. We exchanged email addresses and he began forwarding copies of my own report to me as a favor. Obviously he had no idea who I was. I was sick to my stomach when I saw over 20 names on this guy's distribution list. Of course, none of them are paying customers. At the time some of my better customers were calling me and asking which ones I liked for myself. Not knowing any better I cheerfully gave them out. It soon became painfully obvious to me that during the Saratoga meet there was some pretty serious money landing on many of my own horses. One horse that I gave out (12-1 ML in a 10 horse field) won as the 8-5 favorite. I've since stopped naming specific horses as those I'm going to play for myself. I'm also trying to devise a better security layer (maybe tying the report to each paying customer's ip address) so that the report will only display on a real customer's computer. That's a lot of work. I'm somewhat more inclined to stop renewing subscriptions and simply eliminate the report altogether one customer at a time. It's certainly a headache - one that I don't like dealing with.

More and more, lately all I really want to do is bet.

-jp

[mcikey01]

I'm basically evaluating hundreds of different factors one factor at a time. For each factor I'm calculating something I call a "relative strength" which actually is a combined impact value for both win percent and flat win bet roi. Let's say I take a factor, call it FactorA, and run it through the database. I see that the top ranked horse for FactorA has a win rate of .29 and the flat bet win roi to one dollar is .88. I use each to calculate an impact value using all horses in the database as a benchmark. In the above example let's say that win rate for all starters in the database is .1265 and flat bet win roi for all horses in the database is .7625. I would calculate relative strength for FactorA as follows:

RSa = (.29/.1265) x (.88/.7625)

or

RSa = 2.6458

I then multiply the relative strengths for each of the individual factors used in the mix together to arrive at the power rating.

JPR = RSa x RSb x RSc x RSd x RSe...

Right now each separate factor is weighted equally. But that could change in time as I get more and more into developing it. I'm still in the early stages...

Some keys to doing this successfully, at least for me, are to use factors that have relatively strong roi components to them - those that the public consistently overlooks. For example, the ML Favorite has a very high impact value for win percent. But I don't use it because it has a relatively weak roi component. A lot of the factors I use are ranks and gaps for numbers that I created myself. Typically these have a relatively strong roi component simply because outsiders can't get a handle on them. I think it's also important to stick to causal factors - those I can conceptualize as shaping the outcome of the race in a logical way that makes sense to me. Doing this, even with a database and solid programmnig skills, takes an incredible amount of research time. For every one factor that finds its way into the mix there are hundreds that I dropped because they tested poorly.

-jp


A couple of points:

You're dead on concerning the causality issue...so much statistical correlation in handicapping factors- glibly inferred as causality- is bandied about that dazzles and misleads the handicapper/investor.
My experience in anaylzing Power numbers from a variety of sources is that the final single figure, while valuable, is not as critical to profitable wagering as the weighted scores of the component factors for each of the competitors. Three reasons: 1) everyone can see and utilize the final total Power number, 2)the meaning of the relative difference between the Power numbers of each competitor is not explicit or is misunderstood, 3)when component figures are indexed or benchmarked a revised final Power Number can result in which lower ranked "original" Power numbers are now ranked much higher, resulting in hidden overlays.
At this stage, your ratings appear to be "race condition-neutral". As your research evolves and you delve into potential class, distance,age variables for each type of race analyzed , the weightings can take on a whole new dimension, dramatically impacting winning percentages and profitability. And, of course, the uncovering of hertofore unseen overlays.
ML odds accuracy and profitability fluctuates across racing circuits, racing seasons, individual tracks and race conditions. Therefore, I don't think the ML factor can be uniformly dismissed as non-impactive.
The display of the critical intermediate component values of the Final Power number within the software can evolve into a handicapping process in itself, centering upon the analysis of trends, relative point score differences, postive and negative angles, unplayable races etc.
I personally utilize the availability of weighted and indexed intermedaite Power Number components to develop a fair odds line, assist in determing playable and unplayable race scenarios and finalizing wagering choices
Your caution in offering independent JPR ratings for sale independent of the JCapper software is quite appropriate. Selling or freely offering potentially profitable (from your personal wagering perspective) race and horse specific info online is "dissing" the old maxim that "The value of investment information is inversely proportional to its use"
Best of Luck in your endeavor...You're definitely among the select few who are on the leading edge of handicapping research and application.

[formula_2002]

In a study of 814 favorite plays (most of which were back fitted ) the top bris_power, combined with several handicapping "factors" produced a 2% profit.

In an out of sample set of 892 races, the same conditionds produced a 1% loss.

Additionally, in Nov 2005, 154 out of sample plays produced a loss of 5.5%. Compared to the 12% dollar loss of all favorites in my Nov data base, that 's not too bad a showing.

I'll spend a few hundred buck's for Decembers's Bris data files.
If I can keep the loss between 1 and 4% I just may have some fun with the rebate shops.
Of course, showing a profit, "it wouldn't hurt"!

[formula_2002]

Just noticed that the top Bris_power fig with the same "x13" factors produced a 2% loss not only the win pool, but also in the place pool.

that was 649 winning place bets in 1075 races for a 60% placing rate.
So where do I find these 7% rebate shops!!

[mcikey01]

In a study of 814 favorite plays (most of which were back fitted ) the top bris_power, combined with several handicapping "factors" produced a 2% profit.

In an out of sample set of 892 races, the same conditionds produced a 1% loss.

Additionally, in Nov 2005, 154 out of sample plays produced a loss of 5.5%. Compared to the 12% dollar loss of all favorites in my NovJust noticed that the top Bris_power fig with the same "x13" factors produced a 2% loss not only the win pool, but also in the place pool.

that was 649 winning place bets in 1075 races for a 60% placing rate.
So where do I find these 7% rebate shops!!

Have you analyzed your BRIS power figure data by different race conditions: track, surface, class,age,distance etc.? if so, how does the Win% and ROI vary by the above factors?

[formula_2002]

Have you analyzed your BRIS power figure data by different race conditions: track, surface, class,age,distance etc.? if so, how does the Win% and ROI vary by the above factors?

All the above plus.. (including odds lines and standard deviations ),
Although I have tested the 80+- bris/all-ways factors in the past, I have only tested the Bris_power fig and CJ fig with the x13 model

I'll be testing all the factors with the model.

[spitthebit]

I'm joining this thread a little late, but am interested in the analysis of Bris Prime Power numbers. Anyone try filtering by looking at the point differential between the #1 choice and #2? All-ways apparently lets you play with this, but you seem to need really big databases, and it seems pretty hit or miss anyway.

[hdcper]

Hope this helps,

Hdcper






Data Summary Win Place Show
Mutuel Totals 266588.20 264031.80 260068.10
Bet -351232.00-351232.00-351232.00
Gain -84643.80 -87200.20 -91163.90

Wins 22014 43796 64629
Plays 175616 175616 175616
PCT .1254 .2494 .3680

ROI 0.7590 0.7517 0.7404
Avg Mut 12.11 6.03 4.02


By: Prime Power Gap

>=Min <Max Gain Bet Roi Wins Plays Pct Impact
-999.00 0.00 -77885.00 306760.00 0.7461 15199 153380 .0991 0.7905
0.00 1.00 -1817.20 10006.00 0.8184 1157 5003 .2313 1.8449
1.00 2.00 -1305.30 7580.00 0.8278 976 3790 .2575 2.0544
2.00 3.00 -966.70 6170.00 0.8433 879 3085 .2849 2.2730
3.00 4.00 -674.60 4844.00 0.8607 755 2422 .3117 2.4868
4.00 5.00 -709.80 3784.00 0.8124 590 1892 .3118 2.4877
5.00 6.00 -199.20 2906.00 0.9315 546 1453 .3758 2.9977
6.00 7.00 -196.00 2230.00 0.9121 435 1115 .3901 3.1123
7.00 8.00 -217.90 1742.00 0.8749 355 871 .4076 3.2514
8.00 9.00 -163.50 1298.00 0.8740 269 649 .4145 3.3065
9.00 10.00 -110.80 1014.00 0.8907 213 507 .4201 3.3515
10.00 11.00 -69.90 686.00 0.8981 153 343 .4461 3.5585
11.00 12.00 -58.80 462.00 0.8727 96 231 .4156 3.3153
12.00 13.00 -65.70 414.00 0.8413 89 207 .4300 3.4299
13.00 14.00 -13.70 260.00 0.9473 61 130 .4692 3.7433
14.00 15.00 -19.70 202.00 0.9025 47 101 .4653 3.7123
15.00 16.00 -47.90 174.00 0.7247 36 87 .4138 3.3010
16.00 17.00 -9.80 108.00 0.9093 28 54 .5185 4.1365
17.00 18.00 -6.60 104.00 0.9365 27 52 .5192 4.1421
18.00 999999.00 -105.70 488.00 0.7834 103 244 .4221 3.3675

[formula_2002]

Here is how the x13 model/bris_power did yesterday.
(some picks were posted on the "selection" board)

6 wins in 11 plays.......... net $2.90
7 places in 11 plays........ net $0.10

21 plays x .07 rebate= $1.47
(1 place play did not qualify for the rebate)

I can live with (or on) that.