Welcome to Live View – Take the tour to learn more
Start Tour
There is currently 1 person viewing this thread.
Loved this guy as a racing expert on RUK. Where has he gone? Not appearing on RUK and not writing for R.Post. Has he moved away from racing?
Show More
The racing woman's nookie.
Probably have Ryan Gigs' role if I were to name a team of journos.
Top man. [;)]
Probably have Ryan Gigs' role if I were to name a team of journos.
Top man. [;)]
Naps Champ broke him.
JW was never the same after his "Line in the sand" thread.
JW was never the same after his "Line in the sand" thread.
Think he had more to worry about then getting a slagging on here ffs.
Looking after his very ill son is what I was told so please wind your necks in !!!!
15 years on, Willoughby spoke to the Racing Post's Keith Melrose about...
The future of form study: James Willoughby and Keith Melrose assess how the game will evolve for punters
https://www.racingpost.com/news/features/series/the-future-of-form-study-james-willoughby-and-keith-melrose-assess-how-the-game-will-evolve-for-punters-ayFlm2I0lkQO/
The future of form study: James Willoughby and Keith Melrose assess how the game will evolve for punters
https://www.racingpost.com/news/features/series/the-future-of-form-study-james-willoughby-and-keith-melrose-assess-how-the-game-will-evolve-for-punters-ayFlm2I0lkQO/
It's about time they updated their pic of Willo, I think he was 25 when that one was taken.
Keith Melrose: Because we're talking form study, convention has it that we need to look back before we look forward. James, what would you say has been the biggest change to how we look at form in your racing lifetime?
James Willoughby: In my lifetime, it would be the general acceptance that data and its manipulation are a vital component of the horseracing puzzle. What do you think?
KM: Similar, really. When I came into racing professionally around 20 years ago, time-based analysis was pretty much out of style. Now, it's not accepting time analysis in some form that would be the fringe belief. That's progress rather than fashion, I think.
JW: But basic time analysis is still rarely involved in horseracing broadcasts, even when it conveys dramatic implications about races over the same distance half an hour apart. Do you think that the ITV audience, for example, would appreciate more content like this?
KM: Not all of them, not even a majority, but then ITV does regularly pull in a million viewers. If they could click their fingers and have, say, the Supreme Novices' and Champion Hurdle superimposed on each other – not split screen, more like ghost data in a video game – I think they'd run it. It's visualisation that's the chief hurdle.
JW: To answer the question of how form study might change, we need to know how much factors like times and sectionals genuinely interest people. It seems to me we have more data now than bright ideas.
KM: They definitely interest people. One big change from when I started as a casual punter is that, back then, most people went off what was written in the paper. Now, access to replays has made people more curious. They see more and they want to know what truth lies beneath what they just saw.
JW: Some do, yes, but I think people mostly still want personality. Analysing a horserace is a data analytic test; instinct and experience of riding can only help so much. But there are more ex-riders on the TV than ever. The world has changed a lot since the Post started 40 years ago and I think media companies know better what the public wants. And that's not complexity, is it?
KM: The cult of personality has been the main lesson, across media, from the analytics age. It's water to fish now. Form study is a bit different. We're very lucky in this sport that what is effectively our puzzles page has been given so much real estate down the years.
JW: As time goes on, punters will do less form study and become more disconnected with the process because AI will do the work for them. Form study already sounds an archaic term to me. Restrictive practices by the bookmakers have probably killed the interest of a lot of hobbyist form students.
KM: I wouldn't downplay the effect of account restrictions, but you've hit on something else there. When I think of form study, the mental picture is still those text-laden spreads in the Racing Post between the racecard pages, which I haven't looked at in anger in 15 years. Instinctively, you're right about the AI tools. But as we know from AI generally, unless you understand what the machine is telling you, you're on a hiding to very little, surely?
JW: I Agree. But it is very hard for a newcomer to racing to learn how to make intelligent selections; the media encourages emulation rather than education. Every other sport has gone through a revolution where supposed truths have been replaced by new ways of understanding the game. A newcomer to racing has nothing truly insightful to grab on to.
KM: Sure, everyone starts as a guesser; even Phil Bull or Tony Bloom. But what's your average football fan getting that someone going racing for the first time and getting a tingle does not? I loved how easily xG [expected goals] was accepted, but it's become totemic to some and inadmissible to others – a familiar pattern.
JW: Now you've got me going. xG is a powerful retrodictive metric that predicts how many goals should have been scored with average luck; it is formed by goals regressed on the different components of shots. It also predicts future performance better than actual goals scored. But horseracing just uses data to tell stories. Take Lengths Gained Jumping, for example. Does good jumping cause winning, or does winning cause good jumping?
KM: Winning causes good jumping. Front-running in chases, one of my hobby horses, is similar. If a horse is in front at the end of two miles, it is more likely than its rivals to be in front after each furlong.
JW: Now, another thing we could do for the punters is extract inside information from betting odds. They do market-implied ratings in other sports like the NFL. You can predict ratings from odds and odds from ratings; either way, the difference between the two is a reflection of the strength of inside information. What have your studies of trainer form taught you, Keith?
KM: That the strongest signal from traditional measures comes from their overall strike-rate. Things like percentile scores, which we use now in The Edge, do a far better job than traditional 'hotlists' of telling you how good a recent run of form might be. But they still lack any serious predictive power. The intuitive take that a 'hot' trainer's next two weeks will be at their baseline strike-rate, bumped up a little, is still stronger.
JW: The Edge is a great tool. It takes things forward by a step and that's what we want. So, to make a trainer's strike-rate (SR) predictive, you need to add a prior number of wins and runs to their 14-day or 21-day record based on their longer-term success rate. If you don't do that, you are telling stories about what has happened, when making predictions about what is likely to happen is more useful to the punter. You need to form xSR from SR, in other words.
KM: Right, so you move from simply pointing out how unlikely a string of results would be, to making a reasonable estimate of future results based on that run. It advances our analysis by another step.
JW: Don't get me wrong – what you have done is valid too. We need to use horseracing's basic data much better. Impact value (IV) for example, is a 'gateway' statistic. The punting public could be shown how to make it independent of the strength of past competition with just a minor tweak. This makes it a one-number summary of a horse's merit with serious predictive power which does not even need beaten distances or time.
KM: Gateway is exactly the right word. Impact value is a great tool on samples of horses, because it's so intuitive that, for example, an IV of 1.16 means the sample wins 16 per cent more often than expected by random chance. You're suggesting its use at a horse level, though?
JW: It works reasonably well with trainers and jockeys and sires because they compete in diverse races across the class spectrum. But horses do not. A horse's IV tells you little – until you know the IV of its opponents. David R Hunter's MM algorithm gives us the recipe to make IV a 'Class A' metric! Smart punters would soon be hooked because it gives anyone a way into solving a race with a novel, unambiguous method.
KM: And where's Hunter's algorithm been hiding all this time? I scout around for relevant research papers now and again, because I'm trying to advance mainstream form study even by a tiny amount. This one has escaped me.
JW: Hunter's work would be from before you could legally bet, Keith. You would miss it by looking at newer journals. The bigger question is this: is there an appetite for this kind of thing beyond a few hundred geeks who probably know these things already? For most punters, form study and betting is entertainment at a comfortable level; they don't want to turn it into a maths course. They can already find winners. This is all very dry in comparison.
KM: Never underestimate the capacity for people to want to feel clever. Think of it like the New York Times Games app. I only use it for my Wordle streak and to finish the mini-crossword faster than my colleague Robbie Wilders. I don't play the full-sized crossword, and that's fine. But it's not just 'a few hundred geeks' who will chase the dragon and make it their main hobby.
JW: Fair play to you. So, the future of form study is probably about what the individual wants to make of it to suit their time and appetite for complexity; you can take it anywhere you want and at whatever level suits you. And that's the joy. It is not chess or backgammon where a skilled player who knows the optimal strategy from computer-aided learning will kill you.
KM: Exactly. And chess and backgammon are apt examples. The received wisdom in those games has completely changed in the last 40 or 50 years, because of the access to more powerful computation. They didn't even need sectionals.
JW: Now, it is interesting these things come up. We have touched on what computers can do in other puzzles like the racing one. In backgammon, for instance, world champions like Bill Robertie and Paul Magriel wrote learned books about strategy which turned out to be plain wrong in many situations. In chess, the Soviet School dominated when I was at school until chess computers exposed some of their strategic axioms. That horseracing still uses humans to handicap horses tells you a lot about the type of humans in horseracing: they will be the last to resist the march of the robots; they like to preserve age-old aspects of the sport's practices long after the modern world has changed. This anachronism gives the sport a lot of its charm, I suppose.
KM: People like to think a sport that involves the complexity of 1,000lb animals, travelling at 40mph, with only a faint grasp that they're competing at all, is resistant to such things. In fairness, that position isn't obviously irrational.
JW: Yes, exactly, it will survive analysis a lot longer than many other betting media because there is so much inherent chaos. And, as we touched on when discussing market-implied ratings, there will always be asymmetric information among punters. The key is to find factors that are not in the price and regress or 'blend' your estimate of the true price with the market price, weighting each according to how much predictive power they have, in order to include what you do not know – the inside information, in other words.
KM: This is 'the future of form study' in the popular imagination, I think. People expect every punter to have their own black box model, that they've had an AI build. But asymmetric information will always exist, and no-one knows for sure what the other lot 'knows'.
JW: Yes, and it won't always be data. Everyone has a worthwhile opinion and everyone has a chance to profit from that because the starting price is hard to beat for everyone. I have always said this is the greatest game on Earth because of that. This is an evolving discipline and finding out where it goes is all part of the fun. In this way, at least, it has not changed a bit in 40 years.
KM: I think that's right. Racing has been preserved in its unique chaos through much stronger forces than a new release from OpenAI. It's about solving puzzles and iterating towards better outcomes. Talking of which, I'll definitely be back to pick your brains on those predictive measures you mentioned! James, thank you for having this chat. It's been great fun.
JW: The pleasure's mine. Thanks, Keith.
James Willoughby: In my lifetime, it would be the general acceptance that data and its manipulation are a vital component of the horseracing puzzle. What do you think?
KM: Similar, really. When I came into racing professionally around 20 years ago, time-based analysis was pretty much out of style. Now, it's not accepting time analysis in some form that would be the fringe belief. That's progress rather than fashion, I think.
JW: But basic time analysis is still rarely involved in horseracing broadcasts, even when it conveys dramatic implications about races over the same distance half an hour apart. Do you think that the ITV audience, for example, would appreciate more content like this?
KM: Not all of them, not even a majority, but then ITV does regularly pull in a million viewers. If they could click their fingers and have, say, the Supreme Novices' and Champion Hurdle superimposed on each other – not split screen, more like ghost data in a video game – I think they'd run it. It's visualisation that's the chief hurdle.
JW: To answer the question of how form study might change, we need to know how much factors like times and sectionals genuinely interest people. It seems to me we have more data now than bright ideas.
KM: They definitely interest people. One big change from when I started as a casual punter is that, back then, most people went off what was written in the paper. Now, access to replays has made people more curious. They see more and they want to know what truth lies beneath what they just saw.
JW: Some do, yes, but I think people mostly still want personality. Analysing a horserace is a data analytic test; instinct and experience of riding can only help so much. But there are more ex-riders on the TV than ever. The world has changed a lot since the Post started 40 years ago and I think media companies know better what the public wants. And that's not complexity, is it?
KM: The cult of personality has been the main lesson, across media, from the analytics age. It's water to fish now. Form study is a bit different. We're very lucky in this sport that what is effectively our puzzles page has been given so much real estate down the years.
JW: As time goes on, punters will do less form study and become more disconnected with the process because AI will do the work for them. Form study already sounds an archaic term to me. Restrictive practices by the bookmakers have probably killed the interest of a lot of hobbyist form students.
KM: I wouldn't downplay the effect of account restrictions, but you've hit on something else there. When I think of form study, the mental picture is still those text-laden spreads in the Racing Post between the racecard pages, which I haven't looked at in anger in 15 years. Instinctively, you're right about the AI tools. But as we know from AI generally, unless you understand what the machine is telling you, you're on a hiding to very little, surely?
JW: I Agree. But it is very hard for a newcomer to racing to learn how to make intelligent selections; the media encourages emulation rather than education. Every other sport has gone through a revolution where supposed truths have been replaced by new ways of understanding the game. A newcomer to racing has nothing truly insightful to grab on to.
KM: Sure, everyone starts as a guesser; even Phil Bull or Tony Bloom. But what's your average football fan getting that someone going racing for the first time and getting a tingle does not? I loved how easily xG [expected goals] was accepted, but it's become totemic to some and inadmissible to others – a familiar pattern.
JW: Now you've got me going. xG is a powerful retrodictive metric that predicts how many goals should have been scored with average luck; it is formed by goals regressed on the different components of shots. It also predicts future performance better than actual goals scored. But horseracing just uses data to tell stories. Take Lengths Gained Jumping, for example. Does good jumping cause winning, or does winning cause good jumping?
KM: Winning causes good jumping. Front-running in chases, one of my hobby horses, is similar. If a horse is in front at the end of two miles, it is more likely than its rivals to be in front after each furlong.
JW: Now, another thing we could do for the punters is extract inside information from betting odds. They do market-implied ratings in other sports like the NFL. You can predict ratings from odds and odds from ratings; either way, the difference between the two is a reflection of the strength of inside information. What have your studies of trainer form taught you, Keith?
KM: That the strongest signal from traditional measures comes from their overall strike-rate. Things like percentile scores, which we use now in The Edge, do a far better job than traditional 'hotlists' of telling you how good a recent run of form might be. But they still lack any serious predictive power. The intuitive take that a 'hot' trainer's next two weeks will be at their baseline strike-rate, bumped up a little, is still stronger.
JW: The Edge is a great tool. It takes things forward by a step and that's what we want. So, to make a trainer's strike-rate (SR) predictive, you need to add a prior number of wins and runs to their 14-day or 21-day record based on their longer-term success rate. If you don't do that, you are telling stories about what has happened, when making predictions about what is likely to happen is more useful to the punter. You need to form xSR from SR, in other words.
KM: Right, so you move from simply pointing out how unlikely a string of results would be, to making a reasonable estimate of future results based on that run. It advances our analysis by another step.
JW: Don't get me wrong – what you have done is valid too. We need to use horseracing's basic data much better. Impact value (IV) for example, is a 'gateway' statistic. The punting public could be shown how to make it independent of the strength of past competition with just a minor tweak. This makes it a one-number summary of a horse's merit with serious predictive power which does not even need beaten distances or time.
KM: Gateway is exactly the right word. Impact value is a great tool on samples of horses, because it's so intuitive that, for example, an IV of 1.16 means the sample wins 16 per cent more often than expected by random chance. You're suggesting its use at a horse level, though?
JW: It works reasonably well with trainers and jockeys and sires because they compete in diverse races across the class spectrum. But horses do not. A horse's IV tells you little – until you know the IV of its opponents. David R Hunter's MM algorithm gives us the recipe to make IV a 'Class A' metric! Smart punters would soon be hooked because it gives anyone a way into solving a race with a novel, unambiguous method.
KM: And where's Hunter's algorithm been hiding all this time? I scout around for relevant research papers now and again, because I'm trying to advance mainstream form study even by a tiny amount. This one has escaped me.
JW: Hunter's work would be from before you could legally bet, Keith. You would miss it by looking at newer journals. The bigger question is this: is there an appetite for this kind of thing beyond a few hundred geeks who probably know these things already? For most punters, form study and betting is entertainment at a comfortable level; they don't want to turn it into a maths course. They can already find winners. This is all very dry in comparison.
KM: Never underestimate the capacity for people to want to feel clever. Think of it like the New York Times Games app. I only use it for my Wordle streak and to finish the mini-crossword faster than my colleague Robbie Wilders. I don't play the full-sized crossword, and that's fine. But it's not just 'a few hundred geeks' who will chase the dragon and make it their main hobby.
JW: Fair play to you. So, the future of form study is probably about what the individual wants to make of it to suit their time and appetite for complexity; you can take it anywhere you want and at whatever level suits you. And that's the joy. It is not chess or backgammon where a skilled player who knows the optimal strategy from computer-aided learning will kill you.
KM: Exactly. And chess and backgammon are apt examples. The received wisdom in those games has completely changed in the last 40 or 50 years, because of the access to more powerful computation. They didn't even need sectionals.
JW: Now, it is interesting these things come up. We have touched on what computers can do in other puzzles like the racing one. In backgammon, for instance, world champions like Bill Robertie and Paul Magriel wrote learned books about strategy which turned out to be plain wrong in many situations. In chess, the Soviet School dominated when I was at school until chess computers exposed some of their strategic axioms. That horseracing still uses humans to handicap horses tells you a lot about the type of humans in horseracing: they will be the last to resist the march of the robots; they like to preserve age-old aspects of the sport's practices long after the modern world has changed. This anachronism gives the sport a lot of its charm, I suppose.
KM: People like to think a sport that involves the complexity of 1,000lb animals, travelling at 40mph, with only a faint grasp that they're competing at all, is resistant to such things. In fairness, that position isn't obviously irrational.
JW: Yes, exactly, it will survive analysis a lot longer than many other betting media because there is so much inherent chaos. And, as we touched on when discussing market-implied ratings, there will always be asymmetric information among punters. The key is to find factors that are not in the price and regress or 'blend' your estimate of the true price with the market price, weighting each according to how much predictive power they have, in order to include what you do not know – the inside information, in other words.
KM: This is 'the future of form study' in the popular imagination, I think. People expect every punter to have their own black box model, that they've had an AI build. But asymmetric information will always exist, and no-one knows for sure what the other lot 'knows'.
JW: Yes, and it won't always be data. Everyone has a worthwhile opinion and everyone has a chance to profit from that because the starting price is hard to beat for everyone. I have always said this is the greatest game on Earth because of that. This is an evolving discipline and finding out where it goes is all part of the fun. In this way, at least, it has not changed a bit in 40 years.
KM: I think that's right. Racing has been preserved in its unique chaos through much stronger forces than a new release from OpenAI. It's about solving puzzles and iterating towards better outcomes. Talking of which, I'll definitely be back to pick your brains on those predictive measures you mentioned! James, thank you for having this chat. It's been great fun.
JW: The pleasure's mine. Thanks, Keith.
Willo could make boiling an egg the equivalent of splitting the Atom.
A Marmite character
A Marmite character
Thanks Ram, however got bored halfway through and it’s funny to me that he had to finish by asking questions back and the interviewer often had more to say!
I’ve said it before, my friend that worked in a bookies said, learn to, “Spot the fiddle.”
That’s something not mentioned in that interview but it has often worked for me.
That’s something not mentioned in that interview but it has often worked for me.
Some people wouldn't "Spot A Fiddle" at A Fleadh Cheoil'.
What ever has been spouted above, JW has never used his eyes to understand the relationship between 2 competing horses, he thinks statistics can do it for him at a stroke/press of a button.
That is what makes him hopeless when it comes to betting for a living. His gift is waffle, entertaining waffle it is fair to say, which is why he has to be paid a salary to remain in the industry.
Having said that, it probably applies to ALL salaried tipsters who outnumber pro punters by a 100/1.
That is what makes him hopeless when it comes to betting for a living. His gift is waffle, entertaining waffle it is fair to say, which is why he has to be paid a salary to remain in the industry.
Having said that, it probably applies to ALL salaried tipsters who outnumber pro punters by a 100/1.
uptheirons 01 May 26 17:33
Willo could make boiling an egg the equivalent of splitting the Atom.
A Marmite character
Not really...some people like Marmite
Willo could make boiling an egg the equivalent of splitting the Atom.
A Marmite character
Not really...some people like Marmite
JayTrumpOldTomDubbl • May 1, 2026 8:03 PM BST
Some people wouldn't "Spot A Fiddle" at A Fleadh Cheoil'.
brilliant
Some people wouldn't "Spot A Fiddle" at A Fleadh Cheoil'.
brilliant
They mostly get paid for stating the bleeding obvious.An expert said before the big race at Ascot yesterday that the fav looked great and he thought it had the best chance.How arent we all millionaires with that info at our disposal?
I always found James Willoughby entertaining to watch. Can't say he helped make more money in top races (missing the bleeding obvious by overcomplicating everything). Did help find winners at lesser meetings when he used to show up on track with a presenter such as Nick Luck on Racing TV. I saw them both at Epsom many years ago, James was acting the fool and kept hiding behind a fully opened RP, in the end Nick ripped it away from him.
Willoughby was one of if not the best pundit imo, a great thinker about the game who gave some terrific insight on stuff like pace-scenarios.
And as James said above:
JW: Some do, yes, but I think people mostly still want personality. Analysing a horserace is a data analytic test; instinct and experience of riding can only help so much. But there are more ex-riders on the TV than ever. The world has changed a lot since the Post started 40 years ago and I think media companies know better what the public wants. And that's not complexity, is it?
Which is why you should make a comeback, James!
And as James said above:
JW: Some do, yes, but I think people mostly still want personality. Analysing a horserace is a data analytic test; instinct and experience of riding can only help so much. But there are more ex-riders on the TV than ever. The world has changed a lot since the Post started 40 years ago and I think media companies know better what the public wants. And that's not complexity, is it?
Which is why you should make a comeback, James!
Yes, made for uncomfortable viewing at times when Willo was in the studio and the jumps were on, should have kept his thoughts off camera imo, after all it is a big part of RTV's business and why a large percentage of their viewership subscribe to the channel.
The irony being that when he eventually again wrote the odd article for the RP...It was about jump racing
The irony being that when he eventually again wrote the odd article for the RP...It was about jump racing
Does the US racing for the RP and has tipped up RENEGADE for the Kentucky Derby. I'll probably have a few quid on DANON BOURBON.
only tipsters we follow are the ones who've bought and paid for their own caribbean island, their own apartment in Monte Carlo, their own Mangusta Oceano superyacht and their own 1955 Mercedes-Benz 300 SLR Uhlenhaut Coupe.
all paid for by the bookies.
anyone else is a potless mug, getting paid a pittance to tell you what they fancy.
all paid for by the bookies.
anyone else is a potless mug, getting paid a pittance to tell you what they fancy.
Post Your Reply
Please login to post a reply.
Wonder
Instance ID: 13539