• Hi Guest Just in case you were not aware I wanted to highlight that you can now get a free 7 day trial of Horseracebase here.
    We have a lot of members who are existing users of Horseracebase so help is always available if needed, as well as dedicated section of the fourm here.
    Best Wishes
    AR

Combining Bowers with AI

MySportsAI

MySportsAI

Filly
Creating Bower ratings has been covered extensively on here so I wont go over the calculations again. What interests me however is the step or steps taken after the basic ratings have been compiled. It is clear that different people are applying different form reading or rules after the basic numbers are hatched. What sparked my interest is could AI or machine Learning provide the key to this final step/s. In other words take the Bower qualifiers ie the upper pruned section of race based on the sorted Bower ratings and then treat these as individual races and feed them to MySportsAI. I did this for 2011 to 2019 I first created the Bower ratings for this period and pruned off the upper tract in each race and of course only those races within a certain spread of mid horse OR ranking to average OR.

OK so what are we hoping to better, well backing all these animals produced a loss of 2.41% ROI but backing the top ranked made +1.76% from 4570 bets. Could AI improve on this 1.76%.

I used 3 input features, they were the first 3 tried but I would not say they were entrirely random. I then ran a Gradient Boosting Machine algorithm on the data using a walkforward train/test. In other words it trained on the 2011 data and tested on 2012 and then trained on the 2011 and 2012 data and tested on 2013 etc etc, each time adding the results to the final total. On completion the top rated horse produced 4199 bets (slightly less because we lose 2011) and made a profit of +5.89% after comm'.

Thoughts and questions welcome ?
 
This is what I am trying to do when I have a month's worth or data results I get the AI to see what can be trimmed off to better the results. So far not too bad but early days yet.
So will be keeping an eye on your posts for this thread MySportsAI MySportsAI
Good luck in your research.
 
Interesting AI is very powerful, apparently AI went rogue for the first time yesterday and hacked into another AI to get a project done faster.
But surely AI could be used more for the betterment of mankind wasting time diagnosing cancer and such when it could be actually solving the VDW missing link , something that has fooled many scholars since the 80s.
Surely the experts on here could feed it the tonnes of literature on the subject and AI could finally crack this mystery.
 
Hi MySportsAI MySportsAI
Are you testing on Balanced Handicaps

Example of a balanced handicap

The Balanced Handicap Check: 5.55 Epsom​

To determine if the JRA Tokyo Trophy fits the structural criteria for the elimination method, the spread of the Official Ratings (OR) must be tested.

The Mathematical Breakdown

First, ranking the field by Official Rating from top to bottom:

  1. Roman Dragon (100)
  2. Invictus Gold (99)
  3. Fine Interview (99)
  4. Kylian (98)
  5. Apollo One (97)
  6. Partisan Hero (96)
  7. Gold Star Hero (96)
  8. Sondad (92)
  9. Topwarrior (92)
  10. Eye Of Dubai (90)
  11. Badri (90)
  12. Twilight Jet (89)
  13. Strike Red (88)
  14. Solar Aclaim (86)
  15. Purest Time (86)
  16. Saint Lawrence (86)
Applying the formula:

  • Top OR: 100
  • Bottom OR: 86
  • Mathematical Midpoint: 93.0 (Calculated as 2100+86)
  • Median OR: 92.0 (In a 16-runner field, the median is the average of the 8th and 9th ranked horses. Sondad at 92 and Topwarrior at 92 average exactly 92.0).

The Verdict: BALANCED (PASSED)​

According to the strict rules of the system, the median OR must sit within a 0-to-2 point variance of the mathematical midpoint.

In this race, the median (92.0) sits exactly 1.0 point below the mathematical midpoint (93.0).

Because it safely falls within the acceptable 2-point maximum variance, this mathematically qualifies as a Balanced Handicap. The handicapper has successfully staggered the weights across this tight 14-pound spread, allowing for a highly logical three-tier class breakdown. This race passes the check and is an excellent candidate for the elimination method.

The Tiers (Structural Breakdown)​

With a 14-pound spread from the Top OR (100) to the Bottom OR (86), the field naturally segments into the following distinct class bands:

🥇 The Top Tier: "Above Class" (OR 96 – 100) These are the most naturally gifted runners in the field according to their ratings. They carry the heaviest weights to equalize their proven historical advantage and must perform to a true Class 2 standard (RPR 98+).

  • Roman Dragon (100) — 16/1
  • Invictus Gold (99) — 16/1
  • Fine Interview (99) — 4/1
  • Kylian (98) — 20/1
  • Apollo One (97) — 10/1
  • Partisan Hero (96) — 8/1
  • Gold Star Hero (96) — 7/1
🥈 The Middle Tier: "In Class" (OR 90 – 95) This is the core of the handicap, sitting squarely on the mathematical median. These runners are facing horses of almost identical current ability, making this the primary battleground.

  • Sondad (92) — 10/1
  • Topwarrior (92) — 8/1
  • Eye Of Dubai (90) — 25/1
  • Badri (90) — 10/1
🥉 The Bottom Tier: "Below Class" (OR 86 – 89) Mathematically the lowest-rated horses in the field, stepping up to face significantly tougher opposition. They receive a handy weight advantage from the top tier to balance the scales.

  • Twilight Jet (89) — 12/1
  • Strike Red (88) — 16/1
  • Solar Aclaim (86) — 33/1
  • Purest Time (86) — 14/1
  • Saint Lawrence (86) — 14/1
 
"first created the Bower ratings for this period and pruned off the upper tract in each race and of course only those races within a certain spread of mid horse OR ranking to average OR."
 

The Balanced Handicap Check: 4.40 Epsom​

To determine if this massive 18-runner field fits the structural criteria for our elimination model, we must test the spread of the Official Ratings (OR).

The Mathematical Breakdown

First, ranking the field by Official Rating from top to bottom:

  1. Ozat (100)
  2. Popmaster (97)
  3. Mister Winston (97)
  4. Thunder Roar (96)
  5. Mirsky (96)
  6. Nizam (95)
  7. Mr Swivell (93)
  8. Sterling Knight (92)
  9. Man Of La Mancha (89)
  10. Principality (89)
  11. Stem (89)
  12. Celeborn (89)
  13. Bragbor (89)
  14. Walsingham (88)
  15. Asmen Warrior (88)
  16. Hot Cash (85)
  17. Jimmy Speaking (84)
  18. Final Night (80)
Applying the formula:

  • Top OR: 100
  • Bottom OR: 80
  • Mathematical Midpoint: 90.0 (Calculated as 2100+80)
  • Median OR: 89.0 (In an 18-runner field, the median is the average of the 9th and 10th ranked horses. Man Of La Mancha at 89 and Principality at 89 average exactly 89.0).

The Verdict: BALANCED (PASSED)​

According to the strict rules of the system, the median OR must sit within a 0-to-2 point variance of the mathematical midpoint.

In this race, the median (89.0) sits exactly 1.0 point below the mathematical midpoint (90.0).

Because it safely falls within the acceptable 2-point maximum variance, this mathematically qualifies as a Balanced Handicap. Managing a 20-pound spread across 18 runners is tricky, but the handicapper has successfully staggered the weights to create a legitimate, structurally sound three-tier race shape. This race easily passes the check and is an excellent candidate for the elimination method.

The Tiers (Structural Breakdown)​

With a 20-pound spread from the Top OR (100) to the Bottom OR (80), the field splits into three distinct class bands:

🥇 The Top Tier: "Above Class" (OR 95 – 100) The proven class elite, carrying the heaviest burdens to equalize their historical advantage.

  • Ozat (100) — 50/1
  • Popmaster (97) — 33/1
  • Mister Winston (97) — 6/1
  • Thunder Roar (96) — 25/1
  • Mirsky (96) — 14/1
  • Nizam (95) — 50/1
🥈 The Middle Tier: "In Class" (OR 88 – 93) The core "pivot" block of the handicap, loaded with horses sitting directly around the mathematical median. This is a highly competitive, congested group.

  • Mr Swivell (93) — 20/1
  • Sterling Knight (92) — 14/1
  • Man Of La Mancha (89) — 7/1
  • Principality (89) — 8/1
  • Stem (89) — 5/1
  • Celeborn (89) — 9/2
  • Bragbor (89) — 16/1
  • Walsingham (88) — 20/1
  • Asmen Warrior (88) — 10/1
🥉 The Bottom Tier: "Below Class" (OR 80 – 85) Mathematically the lowest-rated section of the handicap, stepping up in class but receiving a massive weight concession to balance the scales.

  • Hot Cash (85) — 16/1
  • Jimmy Speaking (84) — 16/1
  • Final Night (80) — 20/1
 
You must log in or register to reply here.
Back
Top