• 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

Hong Kong Speed Figures

The Hare said:
These are the results and card for Sunday.

As mentioned an excellent 7 winners on O Outlander top rated out of 9. Made the AI look good but only the good information entered for that.

Great work La Cressonnière La Cressonnière with your AI as well. I should factor in the draw more into my ratings, for the result and the race in the future. Especially at Valley the draw is essential. Earlier on in the thread we were talking about % of runners beaten which I have for jockeys and trainers. I should keep win percentages but I don't at the moment, will post in a second. I had them for draw as well but would need to update.

Hello T tacker , I am interested in your observations on Chelmsford. In the summer I messed around with my form ratings based on the AW races being more spread out than turf and also adjusted for going, good to firm and good going in HK are similar but just the softer ground more spread out obviously. But the rating results have not been too good! Much better last season without the changes. I started the ratings again from 2015 so there was no using of the old method. Maybe too much tinkering and should just adjust the aw and turf difference. There are not many races on soft ground in HK though.
Click to expand...
The Hare The Hare, Hi bud, not sure i can offer anything to be honest, i'm always trying to pick O Outlander's brain but with little success to be honest.
 
The Hare said:
These are the results and card for Sunday.

As mentioned an excellent 7 winners on O Outlander top rated out of 9. Made the AI look good but only the good information entered for that.

Great work La Cressonnière La Cressonnière with your AI as well. I should factor in the draw more into my ratings, for the result and the race in the future. Especially at Valley the draw is essential. Earlier on in the thread we were talking about % of runners beaten which I have for jockeys and trainers. I should keep win percentages but I don't at the moment, will post in a second. I had them for draw as well but would need to update.

Hello T tacker , I am interested in your observations on Chelmsford. In the summer I messed around with my form ratings based on the AW races being more spread out than turf and also adjusted for going, good to firm and good going in HK are similar but just the softer ground more spread out obviously. But the rating results have not been too good! Much better last season without the changes. I started the ratings again from 2015 so there was no using of the old method. Maybe too much tinkering and should just adjust the aw and turf difference. There are not many races on soft ground in HK though.
Click to expand...
Every so often i try to get some kind of handle of how this racing works but find it too complicated, every race looks the same though i was impressed by PACKING ANGEL from it's last win but don't really understand the strength of the opposition.
One thing is clear in that there's some excellent contribution on the thread and deserves success.
 
SHA TIN 02/03/2025
R5 SPECIAL HEDGE (Lucky Fionn - The Heir)
R6 LUCKY TOUCH (Viva Chaleur - Escape Route)
R7 " Handle or not handle that's que question ... NOISY BOY => hold and fast finisher
R8 LOVERO (Lady's Choice - Lucy in the Sky)
R10 SAVVY BRILLANT (Masterofmyuniverse - Crimson Flash)
 
Race 7 - THE HONG KONG CLASSIC CUP
I ask IA with my numbers to select TOP 5 horses taking account stamina and speed. Result :

Below is my final analysis of Race 7 at Sha Tin on March 2, 2025, based on the most crucial metrics for evaluating the horses' chances. The key ratings I’ve focused on are HKR (Hong Kong Rating), SR (Speed Rating), and BSR (B Speed Rating, based on sectional speed), as these provide a comprehensive view of each horse’s overall ability, recent speed, and finishing strength, respectively. I’ve also considered recent form, suitability to the 1800m distance at Sha Tin, jockey-trainer combinations, and notable performance comments to refine the assessment. This analysis concludes with a recommended betting strategy.

Horse-by-Horse Analysis

1. MY WISH

Ratings: HKR: 95, SR: 88.79, BSR: 92.44

Analysis: Boasts the highest HKR (95), reflecting top-tier ability. SR and BSR are solid but not exceptional. Proven at 1400m and 1600m at Sha Tin, it’s untested at 1800m but shows tactical versatility ("smooth headway"). Ferraris has a strong record with this horse, and draw 9 is manageable.

Verdict: A top contender due to its high HKR and consistency.

2. PACKING HERMOD

Ratings: HKR: 93, SR: 93.00, BSR: 99.64

Analysis: Slightly lower HKR than MY WISH but has the highest SR and a near-top BSR, signaling elite speed and sectional times. No 1800m runs, but strong Sha Tin form (e.g. 1400m win) and a finishing kick ("good headway from 2f out") stand out. Shinn rides for Lui’s in-form stable, though draw 13 is wide.

Verdict: Major threat with standout speed ratings.

3. RUBYLOT

Ratings: HKR: 93, SR: 93.00, BSR: 90.79

Analysis: Matches PACKING HERMOD’s HKR and SR but has a lower BSR. Won at 1400m and placed at 1600m at Sha Tin; untested at 1800m. Avdulla and Hayes form a solid team, and draw 8 suits its closing style ("stayed on from 1 1/2f out").

Verdict: Decent place chance but may lack the finishing punch.

4. DIVANO

Ratings: HKR: 91, SR: 91.00, BSR: 96.48

Analysis: Lower HKR but a high BSR suggests strong sectional speed. Won at 1200m and placed at 1600m at Sha Tin; no 1800m experience. Marquand rides for Lui, and draw 11 suits its late run ("good headway from 2f out").

Verdict: Place contender if it handles the distance.

5. JOHANNES BRAHMS

Ratings: HKR: 90, SR: 90.00, BSR: 87.68

Analysis: Moderate ratings with the lowest BSR among key runners. Won at 1600m at Sha Tin but inconsistent. Teetan rides from draw 4, ideal for its midfield style.

Verdict: Limited win upside but could place.

6. STUNNING PEACH

Ratings: HKR: 87, SR: 87.00, BSR: 97.57

Analysis: Lower HKR but a very high BSR reflects excellent sectional speed. Proven at 1800m and longer (e.g. third at 2000m at Sha Tin), making it distance-suited. McDonald and Cruz are a top combo, and draw 7 fits its style ("kept on final furlong").

Verdict: Serious contender with stamina and speed.

7. MICKLEY

Ratings: HKR: 85, SR: 85.00, BSR: 93.86

Analysis: Lower HKR but a decent BSR. Placed at 2000m at Sha Tin, suggesting 1800m suitability. Moore and Size are elite, and draw 5 suits its late run ("stayed on well final furlong").

Verdict: Possible placer with upside.

8. NOISY BOY

Ratings: HKR: 84, SR: 84.00, BSR: 100.70

Analysis: Low HKR but the highest BSR (100.70) in the field, indicating exceptional sectional speed. Wins at 2000m and 2200m prove its stamina, and a Sha Tin win adds relevance. Atzeni and Yip are solid, though draw 12 is tough. Its finishing kick ("stayed on well to lead 50yds out") is a weapon.

Verdict: Leading pick with elite BSR and distance form.

9. CAP FERRAT

Ratings: HKR: 83, SR: 83.00, BSR: 98.41

Analysis: Low HKR but a high BSR. Placed at 1600m at Sha Tin; no 1800m runs. Williams rides for Lui from draw 10, suiting its closing style ("stayed on from over 1f out").

Verdict: Outsider with place potential.

10. CALIFORNIATOTALITY

Ratings: HKR: 81, SR: 81.00, BSR: 83.94

Analysis: Lower ratings but proven at 1800m with a Sha Tin win. Badel rides for Cruz from a tough draw 14, yet its versatility ("led 50yds out") helps.

Verdict: Viable win chance due to distance experience.

11. PACKING ANGEL

Ratings: HKR: 81, SR: 81.00, BSR: 96.14

Analysis: Low HKR but a high BSR. Three 1400m wins at Sha Tin show speed, though it’s untested at 1800m. Moreira and Lui are a dream team, and draw 1 suits its style ("quickened to lead").

Verdict: Big threat if it stretches out.

12. MONDIAL

Ratings: HKR: 80, SR: N/A, BSR: N/A

Analysis: Limited data hampers assessment. Leung and Hayes are capable, and draw 3 is good, but unknowns dominate.

Verdict: Wildcard—too risky to back.

13. STEPS AHEAD

Ratings: HKR: 80, SR: 80.00, BSR: 89.84

Analysis: Low HKR but a decent BSR. Recent 1800m win at Sha Tin is a standout, and Hewitson rides for Lui from draw 6 ("led 1 1/2f out").

Verdict: Strong win contender with proven form.

14. SKY TRUST

Ratings: HKR: 79, SR: 79.00, BSR: 99.42

Analysis: Lowest HKR but a near-top BSR (99.42). No 1800m runs, but two 1400m seconds at Sha Tin show promise. Bowman and Fownes excel in big races, and draw 2 suits ("stayed on well").

Verdict: Each-way chance if it adapts.

Top Contenders :

NOISY BOY: Highest BSR (100.70), proven at longer distances, and a lethal finisher. Low HKR is offset by its sectional speed and stamina.

PACKING HERMOD: High SR (93.00) and BSR (99.64), consistent form, and elite speed make it a danger.

STUNNING PEACH: High BSR (97.57), proven at 1800m, and a top jockey add to its appeal.

CALIFORNIATOTALITY: Lower ratings but a recent 1800m Sha Tin win keeps it in the mix.

MY WISH: Highest HKR (95) and strong form make it a reliable option despite no 1800m runs.

Betting Strategy :

Win Bet: NOISY BOY

Its unmatched BSR (100.70), proven stamina over 2000m+, and finishing strength make it the standout choice.

Place Bets:

PACKING HERMOD: High SR and BSR plus consistent form ensure a likely top-three finish.

STUNNING PEACH: High BSR and distance suitability make it a safe place option.

Exotic Bet: Trifecta Box

NOISY BOY, PACKING HERMOD, STUNNING PEACH, CALIFORNIATOTALITY

This covers the top four with a blend of speed, stamina, and form.

Final Thoughts

For Race 7 at Sha Tin over 1800m, NOISY BOY is my top pick, driven by its exceptional BSR (100.70) and proven ability at longer distances. PACKING HERMOD and STUNNING PEACH are close behind with their high ratings and race suitability, while CALIFORNIATOTALITY and MY WISH round out the key threats. This betting strategy leverages the most critical metrics—HKR, SR, and BSR—alongside form and distance fit for a balanced approach. Good luck on March 2!
 
very interesting race, 2nd leg of their triple crown, full of improving horses going up their handicap/class structure.
going into the 1st leg packing hermod and rubylot set the standard of achievement so far, the winner my wish was a stone behind in what he had done previously.
that shows that it’s not all about merit so far, improvement is a big factor.
in that race my wish and divano produced best ratings, divano tops my ratings now with incredible final section coming from last place and nearly winning from draw 14 and if he had the same set up this time would be a big contender.
this time glory elite is not in the field and the race will not set up the same , it will not be as fast early as last time.
this means could be harder to come from behind not ideal for noisy boy he will need that lethal finish La Cressonnière La Cressonnière mentions and even more, also tougher from outside draws, which is a shame for Californiatotality who has shown incredible resilience from prominent positions but is in the car park this time.
my wish maps to be prominent this time with improving packing Angel .
i think this will be an advantage here, my wish could do it again
packing Angel will have no excuses, it’s a matter of if he has the improvement left to bridge the stone he is currently behind my wish in the handicap off level weights, he could easily have that in reserve but it is a step up.
not sure I will bet , I prefer handicaps
 
IA pointed out, when you query DeepSeek, Grok3, or ChatGPT... these tools merely "rearrange" the data you provide them. This is not what we call Deep Learning—these tools do not generate data. If your data is incorrect, they will provide an incorrect answer.

I believe that the only way to move forward on this topic is to develop a classification model using Machine Learning (XGBoost Tree). As for me, I have started working on this with the following classification approach:

  • If POS (The Hare file) is 1, 2, or 3 → 0
  • If POS is greater than 3 → 1
This setup works for me because my focus is on Trio and Quinella bets. The idea is to detect horses that have a probability of finishing in the "top three" based on the race card. I believe that O Outlander 's figures could work very well within this system.

I'll keep you updated on how this progresses, and if you'd like to discuss it further, I'd be happy to do so.

From an initial quick test, race 7 at Happy Valley gives us Jumbo Fortune, Sky Vino, and Turin Warrior as having a high probability of finishing in the top three. However...
 

Attachments

  • xgb.png
    xgb.png
    329.8 KB · Views: 12
La Cressonnière said:
IA pointed out, when you query DeepSeek, Grok3, or ChatGPT... these tools merely "rearrange" the data you provide them. This is not what we call Deep Learning—these tools do not generate data. If your data is incorrect, they will provide an incorrect answer.

I believe that the only way to move forward on this topic is to develop a classification model using Machine Learning (XGBoost Tree). As for me, I have started working on this with the following classification approach:

  • If POS (The Hare file) is 1, 2, or 3 → 0
  • If POS is greater than 3 → 1
This setup works for me because my focus is on Trio and Quinella bets. The idea is to detect horses that have a probability of finishing in the "top three" based on the race card. I believe that O Outlander 's figures could work very well within this system.

I'll keep you updated on how this progresses, and if you'd like to discuss it further, I'd be happy to do so.

From an initial quick test, race 7 at Happy Valley gives us Jumbo Fortune, Sky Vino, and Turin Warrior as having a high probability of finishing in the top three. However...
Click to expand...
Not going to lie La Cressonnière La Cressonnière this is above my levels of intellect, I understand nothing of this machine learning.
Obviously always interested how you get on and progress with any new ideas.
 
AI Race 4- seems keen on Dragon Four Seas


Final Ranking and Rationale
  1. DRAGON FOUR SEAS (101.7% Ability, 94.6% Speed, 3 90%+)
    • Highest peak ability (101.7%) exceeds its required rating (53), with a strong peak speed (94.6%). Three 90%+ ratings reinforce its edge. Secondary factors (recent win, draw 1, H Bowman) only confirm its dominance—no tiebreaker needed.
  2. CHAMPION METHOD (99.0% Ability, 92.2% Speed, 4 90%+)
    • Second-highest peak ability (99.0%) and a solid peak speed (92.2%), with four 90%+ ratings (most in the field). Wide draw (12) and weaker jockey (H T Mo) are irrelevant unless ratings tie, but it ranks below DRAGON FOUR SEAS due to lower peak ability.
  3. ENDUED (97.0% Ability, 96.9% Speed, 6 90%+)
    • Peak ability (97.0%) is slightly below top two, but peak speed (96.9%) is highest, and six 90%+ ratings (most consistent) give it an edge over others. No tiebreaker needed; ratings alone place it third.
  4. RIDING HIGH (97.6% Ability, 95.1% Speed, 4 90%+)
    • Peak ability (97.6%) edges SOMETHING JONAH (97.0%), and peak speed (95.1%) is strong, with four 90%+ ratings. No tie with ENDUED (higher speed, more 90%+), but draw (2) could matter in a closer call.
  5. SOMETHING JONAH (97.0% Ability, 87.2% Speed, 2 90%+)
    • Peak ability (97.0%) matches ENDUED, but lower peak speed (87.2%) and only two 90%+ ratings drop it below RIDING HIGH (better speed, more 90%+). No tiebreaker needed.
  6. SANTORINI (96.1% Ability, 94.7% Speed, 3 90%+)
    • Peak ability (96.1%) and speed (94.7%) are strong, with three 90%+ ratings. Ranks below higher peaks; wide draw (10) irrelevant without a tie.
  7. SWAGGER BRO (96.6% Ability, 90.6% Speed, 4 90%+)
    • Peak ability (96.6%) is solid, but peak speed (90.6%) is lower, with four 90%+ ratings. Ranks below due to lower peaks; draw (3) is a tiebreaker asset.
  8. SUPER CHARIZZARD (96.4% Ability, 90.0% Speed, 2 90%+)
    • Peak ability (96.4%) and speed (90.0%) are competitive, but only two 90%+ ratings lag behind SWAGGER BRO (four 90%+). Strong form/draw (4) irrelevant without a tie.
  9. GLORY CLOUD (95.7% Ability, 89.4% Speed, 3 90%+)
    • Peak ability (95.7%) and speed (89.4%) are decent, with three 90%+ ratings. Lower peaks place it below others; wide draw (11) not a factor here.
  10. MAGIC BEANS (95.2% Ability, 81.3% Speed, 1 90%+)
    • Peak ability (95.2%) is solid, but low peak speed (81.3%) and only one 90%+ rating weaken its standing. Limited data isn’t penalized, but ratings rank it low.
  11. TELECOM POWER (95.0% Ability, 92.1% Speed, 2 90%+)
    • Peak ability (95.0%) and speed (92.1%) are good, but only two 90%+ ratings and lower peaks place it below others. Happy Valley form is strong but not a ranking factor.
  12. LOVEISINTHEAIR (83.1% Ability, 91.2% Speed, 2 90%+)
    • Lowest peak ability (83.1%) and a decent peak speed (91.2%) with two 90%+ ratings (speed only) make it the weakest contender based on ratings.
Recommendations
  • Top Win Contender: DRAGON FOUR SEAS
    • Highest peak ability (101.7%) and strong speed (94.6%) exceed its required rating (53). Recent form (win, places), ideal draw (1), and elite jockey (H Bowman) reinforce its top ranking.
  • Best Place Bet: ENDUED
    • High peak speed (96.9%) and ability (97.0%), with six 90%+ ratings (most consistent), suggest reliability. R Moore’s jockey quality adds place potential despite mixed form.
  • Value Each-Way Bet: RIDING HIGH
    • Peak ability (97.6%) and speed (95.1%) are near the top, with four 90%+ ratings. Favorable draw (2) and solid jockey (M Chadwick) make it undervalued for a place or win.
  • Longshot: CHAMPION METHOD
    • Peak ability (99.0%) is elite, but wide draw (12) and weaker jockey (H T Mo) lower its chances, making it a high-risk, high-reward option if it overcomes tiebreaker flaws.
Rationale Recap
The ranking prioritizes peak ability percentages (DRAGON FOUR SEAS at 101.7% leads), followed by peak speed (ENDUED’s 96.9% stands out), and uses 90%+ rating counts as a secondary metric (ENDUED’s 6 vs. CHAMPION METHOD’s 4). Secondary factors only apply in ties (e.g., DRAGON FOUR SEAS’ form/draw/jockey solidify its lead, but no tie exists). Horses with limited data (e.g., MAGIC BEANS) are judged solely on ratings, ensuring fairness. This data-driven approach aligns with the methodology, focusing on ratings over positions or subjective factors unless explicitly tied.
 
AI Race 5 - Likes Golden Rise despite the draw.

Final Ranking and Rationale
  1. GOLDEN RISE (103.1% Ability, 102.2% Speed, 5 90%+)
    • Highest peak ability (103.1%) and speed (102.2%) exceed its required rating (57), with five 90%+ ratings. No tiebreaker needed; ratings dominate.
  2. PRIME MORTAR (100.9% Ability, 95.6% Speed, 3 90%+)
    • Peak ability (100.9%) exceeds its required rating (56), with a solid peak speed (95.6%) and three 90%+ ratings. Ranks below GOLDEN RISE due to lower peaks.
  3. MEGA BONUS (99.9% Ability, 95.7% Speed, 2 90%+)
    • Peak ability (99.9%) is just shy of 100%, with a strong peak speed (95.7%) and two 90%+ ratings. Draw (1) could matter in a tie, but ratings place it third.
  4. STAR FIGURE (98.7% Ability, 97.6% Speed, 3 90%+)
    • Peak ability (98.7%) and speed (97.6%) are excellent, with three 90%+ ratings. Limited data isn’t penalized; ranks below top three due to lower peaks.
  5. SPICY SPANGLE (97.8% Ability, 95.6% Speed, 6 90%+)
    • Peak ability (97.8%) and speed (95.6%) are strong, with six 90%+ ratings (most consistent). Ranks below STAR FIGURE due to slightly lower peak ability; no tie within 1-2%.
  6. LUCKY BLESSING (97.0% Ability, 87.0% Speed, 2 90%+)
    • Peak ability (97.0%) is solid, but peak speed (87.0%) is lower, with two 90%+ ratings. Ranks below SPICY SPANGLE due to lower speed and fewer 90%+ ratings.
  7. TAIHANG SCENERY (97.2% Ability, 91.8% Speed, 4 90%+)
    • Peak ability (97.2%) edges LUCKY BLESSING (97.0%), with a better peak speed (91.8%) and four 90%+ ratings. No tie needed; ratings place it seventh.
  8. YOUNG HORIZON (96.5% Ability, 84.0% Speed, 2 90%+)
    • Peak ability (96.5%) is strong, but peak speed (84.0%) is moderate, with two 90%+ ratings. Ranks below TAIHANG SCENERY due to lower peaks.
  9. SILVER SPURS (96.2% Ability, 90.6% Speed, 2 90%+)
    • Peak ability (96.2%) and speed (90.6%) are competitive, with two 90%+ ratings. Ranks below YOUNG HORIZON due to slightly lower ability; draw (3) irrelevant without tie.
  10. SMILING EMPEROR (95.3% Ability, 96.6% Speed, 3 90%+)
    • Peak ability (95.3%) is solid, with a strong peak speed (96.6%) and three 90%+ ratings. Ranks below SILVER SPURS due to lower ability.
  11. MACANESE MASTER (93.9% Ability, 82.9% Speed, 2 90%+)
    • Peak ability (93.9%) is decent, but peak speed (82.9%) is low, with two 90%+ ratings. Ranks below SMILING EMPEROR due to lower peaks.
  12. COLOURFUL WINNER (92.0% Ability, 74.0% Speed, 1 90%+)
    • Lowest peak ability (92.0%) and speed (74.0%), with only one 90%+ rating, place it last.
Recommendations
  • Top Win Contender: GOLDEN RISE
    • Highest peak ability (103.1%) and speed (102.2%) exceed its required rating (57). Recent seconds and R Moore’s jockey quality reinforce its top spot, despite draw (10).
  • Best Place Bet: SPICY SPANGLE
    • Peak ability (97.8%) and speed (95.6%), with six 90%+ ratings (most consistent), suggest reliability. Wide draw (11) is a risk, but ratings and M Chadwick make it a strong place contender.
  • Value Each-Way Bet: SILVER SPURS
    • Peak ability (96.2%) and speed (90.6%) are solid, with two 90%+ ratings. Favorable draw (3) and A Badel’s skill make it undervalued for a place or win.
  • Longshot: STAR FIGURE
    • Peak ability (98.7%) and speed (97.6%) are high, with three 90%+ ratings, but lack of race form and limited data make it a speculative pick. H Bowman could elevate it.
Rationale Recap
The ranking prioritizes peak ability percentages (GOLDEN RISE at 103.1% leads), followed by peak speed (GOLDEN RISE’s 102.2%), and uses 90%+ rating counts as a secondary metric (SPICY SPANGLE’s 6 vs. GOLDEN RISE’s 5). Secondary factors only apply in ties (e.g., MEGA BONUS’s draw 1 vs. PRIME MORTAR’s draw 9, but no tie within 1-2%). Horses with limited data (e.g., STAR FIGURE) are judged solely on ratings, ensuring fairness. This data-driven approach aligns with the methodology, emphasizing ratings over positions or subjective factors unless explicitly tied.
 
La Cressonnière said:
IA pointed out, when you query DeepSeek, Grok3, or ChatGPT... these tools merely "rearrange" the data you provide them. This is not what we call Deep Learning—these tools do not generate data. If your data is incorrect, they will provide an incorrect answer.

I believe that the only way to move forward on this topic is to develop a classification model using Machine Learning (XGBoost Tree). As for me, I have started working on this with the following classification approach:

  • If POS (The Hare file) is 1, 2, or 3 → 0
  • If POS is greater than 3 → 1
This setup works for me because my focus is on Trio and Quinella bets. The idea is to detect horses that have a probability of finishing in the "top three" based on the race card. I believe that O Outlander 's figures could work very well within this system.

I'll keep you updated on how this progresses, and if you'd like to discuss it further, I'd be happy to do so.

From an initial quick test, race 7 at Happy Valley gives us Jumbo Fortune, Sky Vino, and Turin Warrior as having a high probability of finishing in the top three. However...
Click to expand...
Excellent La Cressonnière La Cressonnière . Sadly like O Outlander I also have no knowledge of using Machine Learning, but very interested in your methods. Perhaps I will find the time to look into it more, really has to be a better solution than looking one self.

I should point out I have a slight problem for you - I am going on holiday next week for 3 weeks and don't think I will be able to provide the results whilst away. I will return to normal after my return. I should complete up to Sunday before I go. I don't have a scrape programme either mostly just manual with a small macro to sort the times and sectionals so not really much use. Hopefully this won't dampen the excellent ideas though.
 
HAPPY VALLEY 05/03/2025 - My selection taking in account SR Sect Rat. % Draw - % Trainers - % Jockeys
R2 BOWSER (Rewarding Twinkle - Gimme Five)
R3 M UNICORN (Beato - Excellence Value)
R4 DRAFON FOUR SEAS (Super Charizzard - Telecom Power)
R5 SILVER SPURS (Mega Bonus - Colorful Winner)
R7 BEAUTY INFINITY (Mighty Commander - Nordic Dragon)
R8 TURIN WARRIOR (Sky Vino - En Perfecto)
R9 EASON (An Americ Te Specso - California Deeply)
I don't know if the weighting of Draw % Trainers % Jockeys is good. ..
 
The Hare said:
Excellent La Cressonnière La Cressonnière . Sadly like O Outlander I also have no knowledge of using Machine Learning, but very interested in your methods. Perhaps I will find the time to look into it more, really has to be a better solution than looking one self.

I should point out I have a slight problem for you - I am going on holiday next week for 3 weeks and don't think I will be able to provide the results whilst away. I will return to normal after my return. I should complete up to Sunday before I go. I don't have a scrape programme either mostly just manual with a small macro to sort the times and sectionals so not really much use. Hopefully this won't dampen the excellent ideas though.
Click to expand...
Do you mean that the database you update is made almost manually with little web scraping?
Enjoy your vacation!
 
La Cressonnière said:
Do you mean that the database you update is made almost manually with little web scraping?
Enjoy your vacation!
Click to expand...
Well I copy the results from HKJC site and arrange them in order manually, fairly quick. I use Racing Post information to fill some parts and get the times and sectionals of HKJC site and have a macro to sort them out and a few checks to make sure they are correct for each horse.

Then I have some formulae to collect the last 10 runs of each horse for the racecards and formula to show the last 10 ratings for each horse. But collecting the racecard is manual and setting up the second sheet to sort the last 10 ratings is arranged manually. A little lack of knowledge on my part! I really need to learn to scrape but never got round to it.
 
The Hare said:
Well I copy the results from HKJC site and arrange them in order manually, fairly quick. I use Racing Post information to fill some parts and get the times and sectionals of HKJC site and have a macro to sort them out and a few checks to make sure they are correct for each horse.

Then I have some formulae to collect the last 10 runs of each horse for the racecards and formula to show the last 10 ratings for each horse. But collecting the racecard is manual and setting up the second sheet to sort the last 10 ratings is arranged manually. A little lack of knowledge on my part! I really need to learn to scrape but never got round to it.
Click to expand...
I tried to get results from Race Result | 02 Mar 2025 | Sha Tin | Race 1 FU TAI HANDICAP | HK Racing | South China Morning Post website about 4 years ago i remember copying and pasting and trying to rearrange into excel fields and after about 4 hours i still hadn‘t managed to get a results file for a single meeting, i gave up. I wouldn’t bother without your skills at getting the data. That said I wonder if AI could tabulate info like this and not be too time consuming.
 
Outlander said:
Top Win Contender: DRAGON FOUR SEAS
  • Highest peak ability (101.7%) and strong speed (94.6%) exceed its required rating (53). Recent form (win, places), ideal draw (1), and elite jockey (H Bowman) reinforce its top ranking.
Click to expand...

Outlander said:
Top Win Contender: GOLDEN RISE
  • Highest peak ability (103.1%) and speed (102.2%) exceed its required rating (57). Recent seconds and R Moore’s jockey quality reinforce its top spot, despite draw (10).
Click to expand...
Both of these won again so that’s 4 from 4 winners for AI since I started feeding my full data, I suspect though the analysis is leaning heavily on ability rating that have caught these particular 4 winners and that just coincidence i picked out these 4 races. Golden Rise was the only winner I managed to bet from 3 bets myself. Need to find time to run more races through AI and get a better feel of each contest but it can only handle full data for 1 race at a time. It’s difficult to get it to give equal merit to speed and ability ratings without it wanting to average out which can be done without the need for AI , I would need to see it pick a top contender that isn’t ranked 1 for ability before I would believe it is doing anything useful.
 
You must log in or register to reply here.
Back
Top