On review of winners, and probabilities , most were understandable, even allowing for sectionals, nothing in my ratings could get China Win or especiallyRobot Star anywhere near as short as the betting, Purton effect is my only explanation.My HV 1800m standard time I have flagged as potentially tight previously , but China Win was ranked 2 on SPD so probably not held back by a tough standard. Robot Star has an interesting trial but can’t believe he was a 3.0 chance, but he was a 5.2 chance last time so has been found by the market before.
Hong Kong Speed Figures
- Thread starter The Hare
- Start date
Suppose this was the reason for Robot Star Chau off Purton on after this run
ROBOT STAR (K559) | Jumped only fairly. Over the final 400 Metres was badly held up for clear running and could not be ridden out to the finish. Sent for sampling post-race. |
not enough time in the world to factor in everything but I guess if the horse didn’t get to run for 30% of the race and still runs that section faster than the race as a whole then will be on peoples radar , just not mine because I‘m just looking at numbers.
The Hare
Mare
Results and card, top work
O
Outlander
Become more interested in the probabilities side of things , suppose getting the right price is vital , been looking at my attempts so far don't think it is dealing with the lesser races correctly , seemed to be putting too wide a range of odds in races for lesser quality horses, i amended the formula to try to improve and it looks better to me. Ratings haven't changed, previously getting huge odds for lower rated horses in class 5 and lesser extent class 4.
latest attempt attached
will be a work in progress , see the odds telling a story about any failings in the ratings.
latest attempt attached
will be a work in progress , see the odds telling a story about any failings in the ratings.
Saw this on twitter , apparently chasing better ratings is a waste of time, probably right tbh
Even if your ratings get it right like for example Max Que today, what factors make it anywhere near a 1.7 shot, just hard to believe anyone wants to take less than 2.5 but they kept on ‘correcting‘ the odds like it was unbeatable.
Why Hong Kong Betting Syndicates Win
香港極速
@HSpeedking2020
·
14h
·
It’s comforting. It feels logical. And it’s wrong. Not slightly wrong. Structurally wrong.
If betting syndicates in Hong Kong tried to win by predicting races better than the market, they will fail just as reliably as everyone else. They don’t win because they’re better at picking horses. They win because they understand markets, not racing mythology.
· One of the strongest handicapping systems in the world
· Tight class bands that compress ability
· Weight allocation explicitly designed to neutralise dominance
· Enormous betting liquidity Syndicate money embedded into the pools for decades.
This means something most bettors don’t want to accept: Ability is already known. Not perfectly — but well enough that building another “ability rating” mostly redistributes existing information. If your model outputs a ranked list of horses without referencing price, congratulations: you’ve just replicated the handicapper plus noise.
A syndicate edge often exists for seconds, not hours. They are not “right about the race.” They are right at the moment the bet is placed. Liquidity shifts. Late money distorts prices. Pools overreact. If you are betting early without price justification, you are donating to players who understand this better than you.
· Emotion
· Impulse
· Narrative thinking
· Ego
They don’t automate creativity. They automate obedience. Rules matter more than brilliance:
No edge →no bet
Bad price → no bet
Conflicting signals → no bet.
Most retail systems fail not because the model is bad — but because discipline collapses under pressure.
Why? Retail Bettors Copy the Wrong Things Most people try to copy: Ratings Probabilities “Sharp” selections.
They don’t copy:
The brutal reality is that the Hong Kong races are not beatable by clever handicapping alone. If your approach is: “I rank horses and then decide who to back ”You are already playing the wrong game. The syndicate question is not: Who should I back? It is: Which bets must be rejected, no matter how tempting they look? That mindset — not a better rating — is the real edge.
Finally, the syndicates are not magical. They are not “better at racing. ”They are industrial. They take: small fragile uncomfortable edges…and enforce them with scale, discipline, and zero emotion.
If you want to think like a syndicate, stop chasing winners. Start rejecting bad bets. Everything else is noise.
Even if your ratings get it right like for example Max Que today, what factors make it anywhere near a 1.7 shot, just hard to believe anyone wants to take less than 2.5 but they kept on ‘correcting‘ the odds like it was unbeatable.
Why Hong Kong Betting Syndicates Win
香港極速
@HSpeedking2020
·
14h
·
(And Why Chasing “Better Ratings” Is a Dead End)
Every few months, the same idea resurfaces in Hong Kong racing circles: “If I can just build a better rating, I can beat the market.”It’s comforting. It feels logical. And it’s wrong. Not slightly wrong. Structurally wrong.
If betting syndicates in Hong Kong tried to win by predicting races better than the market, they will fail just as reliably as everyone else. They don’t win because they’re better at picking horses. They win because they understand markets, not racing mythology.
The Bill Benter Misunderstanding
Bill Benter is constantly cited as proof that “models beat racing. “What most people miss is what Benter actually did. He didn’t build a magic horse rating. He didn’t out-handicap the handicapper. He didn’t “see things others couldn’t. “He did something far less romantic and far more dangerous: He built a system to price probabilities more accurately than the crowd and then only bet when the price was wrong enough. That distinction matters.Prediction without price is worthless.
Accuracy without execution is irrelevant. Benter’s advantage wasn’t insight — it was pricing discipline at scale. Hong Kong Is hostile to rating-based systems (by design) Hong Kong racing is not a soft environment. It has:· One of the strongest handicapping systems in the world
· Tight class bands that compress ability
· Weight allocation explicitly designed to neutralise dominance
· Enormous betting liquidity Syndicate money embedded into the pools for decades.
This means something most bettors don’t want to accept: Ability is already known. Not perfectly — but well enough that building another “ability rating” mostly redistributes existing information. If your model outputs a ranked list of horses without referencing price, congratulations: you’ve just replicated the handicapper plus noise.
What Syndicates Actually Model
Syndicates do not ask: “Who is the best horse? “They ask: “Given this price, is betting this horse permitted? “This flips the entire problem on its head. They model constraints, not selections:- Which runners are structurally overbeat?
- Which favourites cannot win often enough at this price?
- Where does market behaviour contradict known limitations?
- Where is the crowd expressing certainty without justification?
This is why syndicates obsess over false favourites.
Avoiding a bad favourite is often more profitable than finding a winner.Why strike rate doesn’t matter (and never did)
Retail bettors obsess over strike rate because they think betting is about being right. Syndicates don’t care. They accept: Low hit rates Long losing runs Nasty variance Because they operate at: Massive sample sizes Tight staking discipline Ruthless execution Judging a system off one meeting — or even one season — is meaningless to them. Their edge exists statistically, not emotionally. Timing Beats Prediction Another uncomfortable truth:A syndicate edge often exists for seconds, not hours. They are not “right about the race.” They are right at the moment the bet is placed. Liquidity shifts. Late money distorts prices. Pools overreact. If you are betting early without price justification, you are donating to players who understand this better than you.
Automation Is About Discipline, Not Genius
Syndicates automate to remove:· Emotion
· Impulse
· Narrative thinking
· Ego
They don’t automate creativity. They automate obedience. Rules matter more than brilliance:
No edge →no bet
Bad price → no bet
Conflicting signals → no bet.
Most retail systems fail not because the model is bad — but because discipline collapses under pressure.
Why? Retail Bettors Copy the Wrong Things Most people try to copy: Ratings Probabilities “Sharp” selections.
They don’t copy:
- ·Filters
- Price thresholds
- Execution timing
- Bankroll controls
- Sample-size thinking.
The brutal reality is that the Hong Kong races are not beatable by clever handicapping alone. If your approach is: “I rank horses and then decide who to back ”You are already playing the wrong game. The syndicate question is not: Who should I back? It is: Which bets must be rejected, no matter how tempting they look? That mindset — not a better rating — is the real edge.
Finally, the syndicates are not magical. They are not “better at racing. ”They are industrial. They take: small fragile uncomfortable edges…and enforce them with scale, discipline, and zero emotion.
If you want to think like a syndicate, stop chasing winners. Start rejecting bad bets. Everything else is noise.
I though I'd post a snippet of my early spd monte carlo sim for Race 1 Sha Tin, it's all measured in metres, #8 can lead cleanly, with a 5.5 metre advantage over the #3 and #14 , you can see the #5 should be able to follow the #8 over and the #1 maybe sits outside of the 5. This is basically what we do with Greyhound races , I know its a 2000mtr race but it's for interest sake
Attachments
Thats why i think it's important to be able to adapt the prcing into your ratings , rather then the pricing into your probabilitiesSaw this on twitter , apparently chasing better ratings is a waste of time, probably right tbh
Even if your ratings get it right like for example Max Que today, what factors make it anywhere near a 1.7 shot, just hard to believe anyone wants to take less than 2.5 but they kept on ‘correcting‘ the odds like it was unbeatable.
Why Hong Kong Betting Syndicates Win
香港極速
@HSpeedking2020
·
14h
·
(And Why Chasing “Better Ratings” Is a Dead End)
Every few months, the same idea resurfaces in Hong Kong racing circles: “If I can just build a better rating, I can beat the market.”
It’s comforting. It feels logical. And it’s wrong. Not slightly wrong. Structurally wrong.
If betting syndicates in Hong Kong tried to win by predicting races better than the market, they will fail just as reliably as everyone else. They don’t win because they’re better at picking horses. They win because they understand markets, not racing mythology.
The Bill Benter Misunderstanding
Bill Benter is constantly cited as proof that “models beat racing. “What most people miss is what Benter actually did. He didn’t build a magic horse rating. He didn’t out-handicap the handicapper. He didn’t “see things others couldn’t. “He did something far less romantic and far more dangerous: He built a system to price probabilities more accurately than the crowd and then only bet when the price was wrong enough. That distinction matters.
Prediction without price is worthless.
Accuracy without execution is irrelevant. Benter’s advantage wasn’t insight — it was pricing discipline at scale. Hong Kong Is hostile to rating-based systems (by design) Hong Kong racing is not a soft environment. It has:
· One of the strongest handicapping systems in the world
· Tight class bands that compress ability
· Weight allocation explicitly designed to neutralise dominance
· Enormous betting liquidity Syndicate money embedded into the pools for decades.
This means something most bettors don’t want to accept: Ability is already known. Not perfectly — but well enough that building another “ability rating” mostly redistributes existing information. If your model outputs a ranked list of horses without referencing price, congratulations: you’ve just replicated the handicapper plus noise.
What Syndicates Actually Model
Syndicates do not ask: “Who is the best horse? “They ask: “Given this price, is betting this horse permitted? “This flips the entire problem on its head. They model constraints, not selections:
- Which runners are structurally overbeat?
- Which favourites cannot win often enough at this price?
- Where does market behaviour contradict known limitations?
- Where is the crowd expressing certainty without justification?
This is why syndicates obsess over false favourites.
Avoiding a bad favourite is often more profitable than finding a winner.
Why strike rate doesn’t matter (and never did)
Retail bettors obsess over strike rate because they think betting is about being right. Syndicates don’t care. They accept: Low hit rates Long losing runs Nasty variance Because they operate at: Massive sample sizes Tight staking discipline Ruthless execution Judging a system off one meeting — or even one season — is meaningless to them. Their edge exists statistically, not emotionally. Timing Beats Prediction Another uncomfortable truth:
A syndicate edge often exists for seconds, not hours. They are not “right about the race.” They are right at the moment the bet is placed. Liquidity shifts. Late money distorts prices. Pools overreact. If you are betting early without price justification, you are donating to players who understand this better than you.
Automation Is About Discipline, Not Genius
Syndicates automate to remove:
· Emotion
· Impulse
· Narrative thinking
· Ego
They don’t automate creativity. They automate obedience. Rules matter more than brilliance:
No edge →no bet
Bad price → no bet
Conflicting signals → no bet.
Most retail systems fail not because the model is bad — but because discipline collapses under pressure.
Why? Retail Bettors Copy the Wrong Things Most people try to copy: Ratings Probabilities “Sharp” selections.
They don’t copy:
So the output looks professional, but the process is still amateur. That’s why the same people rebuild models every season and quietly disappear.
- ·Filters
- Price thresholds
- Execution timing
- Bankroll controls
- Sample-size thinking.
The brutal reality is that the Hong Kong races are not beatable by clever handicapping alone. If your approach is: “I rank horses and then decide who to back ”You are already playing the wrong game. The syndicate question is not: Who should I back? It is: Which bets must be rejected, no matter how tempting they look? That mindset — not a better rating — is the real edge.
Finally, the syndicates are not magical. They are not “better at racing. ”They are industrial. They take: small fragile uncomfortable edges…and enforce them with scale, discipline, and zero emotion.
If you want to think like a syndicate, stop chasing winners. Start rejecting bad bets. Everything else is noise.
As I posted the ST2000 race I find that you have to analysis the L400 sectionals a lot more then the shorts. I have attached a jpg, So i'm looking at ST1400m races the first col is the unique date and race number, the 2nd col. is the act vs expected time adjustment and same with the Final, Now the last two cols are my race shape this tells me if a race has been run fast early or slow and how they ran the Final 400=F400 if you look at race number 45410622 about 4th row down, you see its a 1.794 which equates to a very slowly run first 1000mtrs and they absolutely sprinted home, but this is a real red flag not to use that final sectional going forward over longer distances it's a false flag race. I should have stated that the Final column includes the F400 so you need to do a weighted formula adjustment to get the correct Final figure
Attachments
Last edited:
very interesting stuff W woof43 , got to find more time to understand fully, i see these races with false closing sectionals, tried to build sectional ratings that take into account overall time to avoid the false flags, spent a good bit of time trying to evaluate ways of putting numbers on everything without allowing these false flag ratings.
nice start to the meeting,
Finally got all the sectionals and combinations of sectionals i can think of on the same scale as HKJC , been meaning to do it for a while, hopefully will help me a bit with direct comparison to HKJC Ratings, still not sure of merit of individual sections but can analyse them a bit more now.
Sense checked them , don't seem any anomalies.
attached this season so far for anyone interested.
Sense checked them , don't seem any anomalies.
attached this season so far for anyone interested.
Nice work, It is always interesting to see how others put things together, rather then someone posting tips etc, The first thing I looked at was the MGA I then looked back at a runner I have some knowledge of Audacious Spirit, because it's raced recently, raced on varying speed races and over a couple of distances back to back and finished in the top 5 most starts so consistent one could say. I notice your OHR ratings are harsh on slowly run races and give more to Fast run races, which is all pretty much inline with expectation, when I have a slowly run race I uplift the Spd rating and a fast run race is downgraded by my TV adjustment, which is MY goal in dealing with the Finish Time Rating. Now onto the sectionals, I use the first sectional only to create a type of speed map / MCsim, I really don't blend that first section into my overall probabilities. I use the L400m sectional which is where i do all my calculations, so the early portion of the race is race distance minus L400m, I have a track speed algorithm for the L400 and then I measure that against the Overall Track speed figure, I do a straight subtraction and that becomes my Race Shape figure it is not used in adjusting any times, to find the Early Race figures is about Racedist*EarlyTrack speed-400*Last400Trackspeed)/(racedist-400) that gives a figure + or - a negative = Fast run first portion of race, a positive would mean a slowly run first portion of the race. Now I return to the Final 400m same again as before positive and negative, tell me how the race was run, I then use classifications to identify certain race types based on the Fist portion of race and L400metres I have upper and lower bounds. The race classifications tell me more about the ability of the horse and is the best predictor of it's ability. Finsh Time can be contaminated by Race Shape and can be unreliable for all of the first reasons I mentioned in this post. I have attached another jpg, most peoples eyes will glaze over with this stuff, remember this is Race Classification, the top row Fast Early portion Slow Finish, this suited the winner China Win and also this horse the False Run race very slow early so you'd expect this type of runner to have a serious run home rating but not suited by race at all my F400 rat is an internal score which is another flag about the race. My advice Finish Time is dominated by Race Shape and not ability, the Last 400m is the truest expression of a horses ability, this is why the Final 400m ratings often price a field more accurately then finish time
Attachments
interesting as always, i had come to same conclusion that last 400m of race is most predictive, but at the same time using one section can be misleading, looks like you lots of checks and balances in place to account for this in your methods.
Interested also to read you uplift spd rating for slowly run races and downgrade for fast run races, guess i trying to do the same in a different way, the faster late sectionals in a slowly run race uplift the slow race, not the actual spd rating but the sct and efficiency ratings
went well in the middle few back to back races
Wasn't too bad, few probabilities i didn't like , Parents Love and Wonderstar in race 8, clear 1000m turf speed horses, can't take their past performances literally for a 1200m AWT race, horses i wouldn't consider for the race they were in yesterday so not a help to be short on my probabilities, also horse like Gummy Gummy that is a AWT specialist but rated lowly because been running TURF lately, wouldn't be for me but i also know its not an 80/1 shot.
TODAY rank 1 , 3 winners and tiny lsp at hk tote.
Think i'm not a million miles away from a reasonable guide to the card.