A/I & Computer Handicapping

[classhandicapper]

Here is some data on FTS that showed speed in their debuts broken down by surface and field size. This is from my database covering major tracks for the last 4+ years.

I may decide to change (remove field size) or continue improving the query as I see the data, but it's a start. This is my fist look at it too. I can also easily provide this with an Odds range. I hope this helps.

[formula_2002]

Originally Posted by PaceAdvantage

Another tremendous thread that proves it's worthwhile to keep doing this after almost 20 years...thank you!

AS A DISCUSSION, PROBABLY ONE OF THE BEST THREADS EVER.
I MAY TAKE A YEAR OFF TO READ ALL THE NOTES.
BY THE WAY, ANY NOTE BY JEFF P IS ALWAYS WORTH THE READ.

[VigorsTheGrey]

I’m not a computer person and handicap pen and DRF...but what strikes me as really crucial in all the stats is the relationship of FIELD SIZE to everything else...without first filtering for FIELD SIZE how meaningful can any Metric that isolates another aspect be...?

[Jeff P]

In large samples win rate within incremental odds ranges has a tendency to be consistent across all field sizes.

This is what I have in my database from 01-01-2017 current through Wed 01-23-2019 for thoroughbreds in the 2-1 incremental odds range:

Data Window Settings:
Connected to: C:\JCapper\exe\JCapper2.mdb
999 Divisor  Odds Cap: None
SQL UDM Plays Report: Hide

SQL:  SELECT * FROM STARTERHISTORY
      WHERE ODDS >= 2 
      AND ODDS < 2.5 
      AND [DATE] >= #01-01-2017# 
      AND [DATE] <= #01-23-2019# 
      ORDER BY [DATE], TRACK, RACE


Data Summary          Win         Place          Show
-----------------------------------------------------
Mutuel Totals    49023.90      50256.90      50322.60
Bet             -58512.00     -58512.00     -58512.00
-----------------------------------------------------
P/L              -9488.10      -8255.10      -8189.40

Wins                 7678         14275         18780
Plays               29256         29256         29256
PCT                 .2624         .4879         .6419

ROI                0.8378        0.8589        0.8600
Avg Mut              6.38          3.52          2.68

This is what the above sample looks like with the data broken out by field size:

By: FieldSize

Value      P/L        Bet        Roi    Wins   Plays     Pct     Impact     AvgMut
----------------------------------------------------------------------------------
 1        0.00       0.00     0.0000       0       0   .0000     0.0000       0.00  
 2       -6.00       6.00     0.0000       0       3   .0000     0.0000       0.00  
 3      -15.00      66.00     0.7727       8      33   .2424     0.9237       6.38  
 4     -128.20     908.00     0.8588     122     454   .2687     1.0239       6.39  
 5     -720.90    4824.00     0.8506     644    2412   .2670     1.0174       6.37  
 6    -1996.20   11914.00     0.8324    1555    5957   .2610     0.9946       6.38  
 7    -2530.70   13508.00     0.8127    1719    6754   .2545     0.9698       6.39  
 8    -1807.00   10782.00     0.8324    1404    5391   .2604     0.9923       6.39  
 9     -945.60    7318.00     0.8708     998    3659   .2728     1.0393       6.39  
10     -761.50    5208.00     0.8538     695    2604   .2669     1.0170       6.40  
11     -336.90    2162.00     0.8442     287    1081   .2655     1.0116       6.36  
12     -226.00    1582.00     0.8571     212     791   .2680     1.0212       6.40  
13       32.00     136.00     1.2353      26      68   .3824     1.4569       6.46  
14      -50.50      96.00     0.4740       7      48   .1458     0.5557       6.50  
15        0.00       0.00     0.0000       0       0   .0000     0.0000       0.00  
16        4.40       2.00     3.2000       1       1  1.0000     3.8104       6.40  
17        0.00       0.00     0.0000       0       0   .0000     0.0000       0.00  
18        0.00       0.00     0.0000       0       0   .0000     0.0000       0.00  
19        0.00       0.00     0.0000       0       0   .0000     0.0000       0.00  
20        0.00       0.00     0.0000       0       0   .0000     0.0000       0.00  

Note that the overall win rate for horses in the 2-1 incremental odds range is about 0.2624 (use this as your baseline.)

Note that as field size changes within the incremental odds range there is little deviation from the baseline win rate.

Imo, the deviation from the baseline in rows 2, 3, 13, 14, and 16 is a result of small sample size.




This is what I have in my database from 01-01-2017 current through Wed 01-23-2019 for thoroughbreds in the 8-1 incremental odds range:

Data Window Settings:
Connected to: C:\JCapper\exe\JCapper2.mdb
999 Divisor  Odds Cap: None
SQL UDM Plays Report: Hide

SQL:  SELECT * FROM STARTERHISTORY
      WHERE ODDS >= 8 
      AND ODDS < 9 
      AND [DATE] >= #01-01-2017# 
      AND [DATE] <= #01-23-2019# 
      ORDER BY [DATE], TRACK, RACE


Data Summary          Win         Place          Show
-----------------------------------------------------
Mutuel Totals    34545.70      35213.90      35447.50
Bet             -45406.00     -45406.00     -45406.00
-----------------------------------------------------
P/L             -10860.30     -10192.10      -9958.50

Wins                 1839          4685          7878
Plays               22703         22703         22703
PCT                 .0810         .2064         .3470

ROI                0.7608        0.7755        0.7807
Avg Mut             18.79          7.52          4.50

This is what the above sample looks like with the data broken out by field size:

By: FieldSize

Value      P/L        Bet        Roi    Wins   Plays     Pct     Impact     AvgMut
----------------------------------------------------------------------------------
 1        0.00       0.00     0.0000       0       0   .0000     0.0000       0.00  
 2        0.00       0.00     0.0000       0       0   .0000     0.0000       0.00  
 3        7.20      12.00     1.6000       1       6   .1667     2.0575      19.20  
 4     -145.00     296.00     0.5101       8     148   .0541     0.6673      18.88  
 5     -620.90    2224.00     0.7208      86    1112   .0773     0.9548      18.64  
 6    -1815.00    6864.00     0.7356     268    3432   .0781     0.9640      18.84  
 7    -1801.20    9376.00     0.8079     404    4688   .0862     1.0639      18.75  
 8    -2207.70    8836.00     0.7501     353    4418   .0799     0.9864      18.78  
 9    -1603.80    6906.00     0.7678     283    3453   .0820     1.0118      18.74  
10     -985.00    5720.00     0.8278     251    2860   .0878     1.0835      18.86  
11     -791.50    2714.00     0.7084     102    1357   .0752     0.9279      18.85  
12     -769.10    2142.00     0.6409      73    1071   .0682     0.8415      18.81  
13     -120.20     176.00     0.3170       3      88   .0341     0.4209      18.60  
14       -6.10     138.00     0.9558       7      69   .1014     1.2524      18.84  
15        0.00       0.00     0.0000       0       0   .0000     0.0000       0.00  
16        0.00       0.00     0.0000       0       0   .0000     0.0000       0.00  
17        0.00       0.00     0.0000       0       0   .0000     0.0000       0.00  
18        0.00       0.00     0.0000       0       0   .0000     0.0000       0.00  
19        0.00       0.00     0.0000       0       0   .0000     0.0000       0.00  
20       -2.00       2.00     0.0000       0       1   .0000     0.0000       0.00

Note that the overall win rate for horses in the incremental 8-1 odds range is 0.081 (use this as your baseline.)

Note that as field size changes within the incremental odds range there is little deviation from the baseline win rate.

Imo, the deviation from the baseline in rows 3, 4, 13, 14, and 20 is a result of small sample size.



One observation I've made after years of looking at similar samples for just about every incremental odds range is this:

Win rate for just about every factor I've ever looked at is shaped by incremental odds range far more than field size.

As a general rule:

• The lower the odds the higher the win rate.
  
• The higher the odds the lower the win rate.

Whenever I'm looking at some factor with the data broken out by field size and I see higher win rate at lower field size and lower win rate at higher field size:

Generally if I dig deeper into the data I'll discover the real correlation turns out to be odds based rather than field size based.



But if you are looking at some factor other than the odds and you are breaking the data out by field size:

Why is there often a clear trend in win rate that correlates to changes in field size?

Imo, this is why:

For smaller fields, the money bet gets divided up a smaller number of ways.

In turn, this creates a tendency towards lower odds for most factors in smaller fields. Yet it gives the appearance that field size is causing the higher win rate for most factors in smaller fields. But if you dig deeper, more often than not you'll discover it is the lower odds (not field size) that turns out to be the root cause of the higher win rate.

Whereas in bigger fields, the money bet gets divided up a larger number of ways.

In turn, this creates a tendency towards higher odds for most factors in bigger fields. Yet it gives the appearance that field size is causing the lower win rate for most factors in bigger fields. But if you dig deeper, more often than not you'll discover it is the higher odds (not field size) that turns out to be the root cause of the lower win rate.

Suggestion:

In addition to field size, break the data for whatever you are looking at out by incremental odds ranges.


I hope I managed to type out most of that in a way that makes sense,


-jp

.

[VigorsTheGrey]

Interesting post, Jeff...I would have thought that a 2-1 shot would have a lower hit percentage in a race with more horses in it, amazing stuff indeed...!

One question is that if an 8-1 shot has about an 8% hit rate across all field sizes, would the hit rate for finishing in exactly the 2nd hole be lower or higher than the 8% win hit percentage...? And by how much...?

[headhawg]

Great information again Jeff. Thanks for sharing.

[traveler]

Superb Jeff. Knowing the right questions to ask sure helps and having your superb software.

[VigorsTheGrey]

Originally Posted by VigorsTheGrey

Interesting post, Jeff...I would have thought that a 2-1 shot would have a lower hit percentage in a race with more horses in it, amazing stuff indeed...!

One question is that if an 8-1 shot has about an 8% hit rate across all field sizes, would the hit rate for finishing in exactly the 2nd hole be lower or higher than the 8% win hit percentage...? And by how much...?

I now see that, in the 8-1 example, you showed the win place show pct’s in your post above those being:

0.0810 win
0.2064 place
0.3470 Show

So the rate for EXACTLY 2nd would be 20 minus 8 = 12%
And further, for EXACTLY 3rd would be 34 minus 28 = 6%

So 8% to run EXACTLY first
And 12% to run EXACTLY 2nd
And 6% to run EXACTLY 3rd.

Do I have that, basically correct, for an 8-1 shot...?

[Jeff P]

Originally Posted by VigorsTheGrey

I now see that, in the 8-1 example, you showed the win place show pct’s in your post above those being:

0.0810 win
0.2064 place
0.3470 Show

So the rate for EXACTLY 2nd would be 20 minus 8 = 12%
And further, for EXACTLY 3rd would be 34 minus 28 = 6%

So 8% to run EXACTLY first
And 12% to run EXACTLY 2nd
And 6% to run EXACTLY 3rd.

Do I have that, basically correct, for an 8-1 shot...?

That's close but --

Not all of the races offered show wagering - which skews things a bit.

If a horse finished 3rd but didn't have a show mutuel it doesn't display in the first set of tables I posted.

Same goes for place wagering.





The numbers in the table below should give you exact finish positions for all starters in the 8-1 incremental odds range from 01-01-2017 to 01-23-2019:

*************************************************************************************
FINISH                                IMPACT
POSITION     PLAYS   COUNT      PCT   VALUE     Xth or Better
*************************************************************************************
1ST          22703   1833    0.0807  0.9963     1ST or better  001833/022703 = 0.0807
2ND          22703   2835    0.1249  1.5419     2ND or better  004668/022703 = 0.2056
3RD          22703   3340    0.1471  1.816      3RD or better  008008/022703 = 0.3527
4TH          22703   3750    0.1652  2.0394     4TH or better  011758/022703 = 0.5179
5TH          22703   3525    0.1553  1.9172     5TH or better  015283/022703 = 0.6732
6TH          22703   2922    0.1287  1.5888     6TH or better  018205/022703 = 0.8019
7TH          22703   2033    0.0895  1.1049     7TH or better  020238/022703 = 0.8914
8TH          22703   1249    0.055   0.679      8TH or better  021487/022703 = 0.9464
9TH          22703   663     0.0292  0.3605     9TH or better  022150/022703 = 0.9756
10TH         22703   343     0.0151  0.1864    10TH or better  022493/022703 = 0.9908
11TH         22703   156     0.0069  0.0852    11TH or better  022649/022703 = 0.9976
12TH         22703   46      0.002   0.0247    12TH or better  022695/022703 = 0.9996
13TH         22703   4       0.0002  0.0025    13TH or better  022699/022703 = 0.9998
14TH         22703   3       0.0001  0.0012    14TH or better  022702/022703 = 1.0000
15TH         22703   0       0       0         15TH or better  022702/022703 = 1.0000
16TH         22703   0       0       0         16TH or better  022702/022703 = 1.0000
17TH         22703   0       0       0         17TH or better  022702/022703 = 1.0000
18TH         22703   0       0       0         18TH or better  022702/022703 = 1.0000
19TH         22703   0       0       0         19TH or better  022702/022703 = 1.0000
20TH         22703   0       0       0         20TH or better  022702/022703 = 1.0000
*************************************************************************************

For exact Finish Position use these from the PCT column:

1ST: 0.0807
2ND: 0.1249
3RD: 0.1471
4TH: 0.1652


For Xth or Better Finish Position use the numbers on the far right side of the table.




-jp

.

[VigorsTheGrey]

Now that is a beautiful table of data...thank you so much for this Jeff...the topics addressed here I have asked about many times and often wondered about...these tables show very clearly what really is happening...makes so much sense...!

[VigorsTheGrey]

Jeff, I was wondering if you could post the table for exact finish position for a 2-1 shot just so we could have something to compare to the 8-1 shot. I’ll quote your remark here:

“One observation I've made after years of looking at similar samples for just about every incremental odds range is this:

Win rate for just about every factor I've ever looked at is shaped by incremental odds range far more than field size.

As a general rule:
The lower the odds the higher the win rate.
The higher the odds the lower the win rate.”

I’m thinking you have run a table for every distinct odds to compare them...in the past.

Was wondering if the same basic pattern of percentages for the lower odds categories would more or less resemble the 8-1 table, or if they would be more linear rather than bell shaped...?

[Fox]

Originally Posted by formula_2002

PIC6SIX,
I’m about 9 years your senior.

Wow! 82 if I did my math right. Your stick-to-it-ness is most impressive.

[Jeff P]

Originally Posted by VigorsTheGrey

Jeff, I was wondering if you could post the table for exact finish position for a 2-1 shot just so we could have something to compare to the 8-1 shot.

Here's the top part of the 2-1 incremental odds range again:

query start:         1/27/2019 6:32:52 PM
query end:           1/27/2019 6:33:05 PM
elapsed time:        13 seconds

Data Window Settings:
Connected to: C:\JCapper\exe\JCapper2.mdb
999 Divisor  Odds Cap: None
SQL UDM Plays Report: Hide

SQL:  SELECT * FROM STARTERHISTORY
      WHERE ODDS >= 2 
      AND ODDS < 2.5 
      AND [DATE] >= #01-01-2017# 
      AND [DATE] <= #01-23-2019# 
      ORDER BY [DATE], TRACK, RACE


Data Summary          Win         Place          Show
-----------------------------------------------------
Mutuel Totals    49023.90      50256.90      50322.60
Bet             -58512.00     -58512.00     -58512.00
-----------------------------------------------------
P/L              -9488.10      -8255.10      -8189.40

Wins                 7678         14275         18780
Plays               29256         29256         29256
PCT                 .2624         .4879         .6419

ROI                0.8378        0.8589        0.8600
Avg Mut              6.38          3.52          2.68

And this is what the above sample looks like broken out by finish position:

*************************************************************************************
FINISH                                IMPACT
POSITION     PLAYS   COUNT      PCT   VALUE     Xth or Better
*************************************************************************************
1ST          29256   7611    0.2602  0.9915     1ST or better  007611/029256 = 0.2602
2ND          29256   6570    0.2246  0.8558     2ND or better  014181/029256 = 0.4847
3RD          29256   5022    0.1717  0.6542     3RD or better  019203/029256 = 0.6564
4TH          29256   3699    0.1264  0.4816     4TH or better  022902/029256 = 0.7828
5TH          29256   2685    0.0918  0.3498     5TH or better  025587/029256 = 0.8746
6TH          29256   1686    0.0576  0.2195     6TH or better  027273/029256 = 0.9322
7TH          29256   1045    0.0357  0.136      7TH or better  028318/029256 = 0.9679
8TH          29256   551     0.0188  0.0716     8TH or better  028869/029256 = 0.9868
9TH          29256   219     0.0075  0.0286     9TH or better  029088/029256 = 0.9943
10TH         29256   102     0.0035  0.0133    10TH or better  029190/029256 = 0.9977
11TH         29256   46      0.0016  0.0061    11TH or better  029236/029256 = 0.9993
12TH         29256   13      0.0004  0.0015    12TH or better  029249/029256 = 0.9998
13TH         29256   3       0.0001  0.0004    13TH or better  029252/029256 = 0.9999
14TH         29256   3       0.0001  0.0004    14TH or better  029255/029256 = 1.0000
15TH         29256   0       0       0         15TH or better  029255/029256 = 1.0000
16TH         29256   0       0       0         16TH or better  029255/029256 = 1.0000
17TH         29256   0       0       0         17TH or better  029255/029256 = 1.0000
18TH         29256   0       0       0         18TH or better  029255/029256 = 1.0000
19TH         29256   0       0       0         19TH or better  029255/029256 = 1.0000
20TH         29256   0       0       0         20TH or better  029255/029256 = 1.0000
*************************************************************************************

Originally Posted by VigorsTheGrey

I’m thinking you have run a table for every distinct odds to compare them...in the past.

I have.

Originally Posted by VigorsTheGrey

Was wondering if the same basic pattern of percentages for the lower odds categories would more or less resemble the 8-1 table, or if they would be more linear rather than bell shaped...?

If you're looking at the shape of the data:

There's very little variance within an incremental odds range from its baseline win rate - similar to the field size tables posted above.

In general --

The lower the odds in an incremental odds range: the harder it is to find factors that 'move the needle' away from the baseline win rate of an incremental odds range.

As you look at higher and higher incremental odds ranges: the easier it is to find factors that 'move the needle' away from the baseline win rate of an incremental odds range.



-jp

.

[VigorsTheGrey]

Thank you Jeff, for posting the above tables, they really help...they show that any given horses’ finish position correlates with its own odds in spite of field size...for years I have bet longer odds horses in large fields thinking the opposite, that they will outperform their odds in bigger fields, that if I wanted to bet 2-1 I could bet THAT in a six horse field with a (mistaken) greater chance of winning THERE. Thanks for curing me of that curse, I now see that a 2-1 is a 2-1 irrespective of field size.

One of the maxims I have retained is/ was to “bet GOOD HORSES at GOOD ODDS”...nowadays, we have so many races that are filled with horses with poor records (bad races) that finding those opportunities to follow that maxim is a little more difficult....what I find there sometimes, are BAD HORSES scoring at GOOD ODDS, but horses that are nearly impossible to have, based on their past performances.

...I am drawn to Stakes races and Turf races because I believe that there might be more opportunities to get more GOOD HORSES at GOOD ODDS....horses that clearly show the form to be able to win.

Might the above 2 race categories present more factors to “move the needle”...? Or do these races act like the more common, routine races as well...? I have lost a lot of bets on longer odds horses in stakes races.

And so the hunt for the factors that “move the needle” continues...
But what your tables show for me is that “the public” ...based on odds...is pretty good at handicapping, that the market IS efficient...

...but therein lies the rub...the computers have looked at these patterns and construct their wagers, I assume, based on these tables that you have provided (or other tables similar to them) and so the VALUE is not there due to massive coverage along these lines...

So like other posters have suggested, the value action might lay where the computers “get it wrong”...but this kind of thinking seems to resemble what I have always done, seemingly incorrectly...chasing improbable outcomes for ‘value’....

[Jeff P]

Where's the value? That's always the bigger question isn't it?

Right now as I type this it's post time for R11 at Parx.

On a day that has seen winners distribute their engergy late:

The #6 horse appears to be up against it bias-wise, and is currently the 2-1 favorite.

Which raises the question:

Is that enough of a bet down to create value elsewhere?

Imo, that's the type of question that needs to be asked.



-jp

.