9 ms·

It says he's still winning to this day. Makes me wonder if his current model is using Machine Learning (ML) techniques? Which leads me to wonder how successful

It says he's still winning to this day. Makes me wonder if his current model is using Machine Learning (ML) techniques? Which leads me to wonder how successful ML techniques are with data rich sources of gambling and whether there are any syndicates I can join to invest in and let them do all the heavy lifting... I'm sure there's a whole world out there I am yet to discover.

kratom_sandwich 7y ago
But where would you even start discovery? I guess no one with a remotely useful idea wants to share it ...I did a bit of Googling after the article came out last year and apart from a few research papers and that Wired article about Joan Ginther, not much comes up. I'm surprised he is openly talking about everything, but I suppose many others have caught up and there's not much to lose ...
- yalooze 7y ago
  Yeah, that's the conclusion I am coming to as well. The people with good models aren't going to talk about them. Even if they need investment they'll seek it privately. I guess I was thinking of a market where you can 'invest' your money in someone's model and they take a percentage of your winnings (that's their incentive for adding your money to the pool). Then it's a democratised race for whoever has the best model.
  - lscharen 7y ago
    Here's an interesting article about an NBA sports-better that developed a high-fidelity model to simulate NBA games. https://www.espn.com/blog/playbook/dollars/post/_/id/2935/meet-the-worlds-top-nba-gambler https://www.espn.com/blog/playbook/dollars/post/_/id/2935/me...
  - phillc73 7y ago
    Here's an idea from an Australian website where you can build a very basic model, using past results (which are of course no guarantee of future performance) and then sell it. Or alternatively, you can buy "qualifiers" for these systems for a small fee.[1] [1] http://www.puntingform.com.au/account/systems-dashboard/#toppublicsystems http://www.puntingform.com.au/account/systems-dashboard/#top...
  - foxhound6 7y ago
    In horse racing where the bets are done via parimutuel[1], the challenge is that there's an upper bound on how much you can "invest" in a given wager without changing the odds. This makes it less likely that an expert will seek outside investment unless their model has them betting "wide" (lots of small bets vs. a few large ones). [1] https://en.wikipedia.org/wiki/Parimutuel_betting https://en.wikipedia.org/wiki/Parimutuel_betting
    - PaulRobinson 7y ago
      Not everywhere does Parimutuel - the UK and Ireland for example have basically rejected it despite massive investment into on-course "Tote" systems.
      - phillc73 7y ago
        Is there any hope for Together for the Tote?[1] I like to dream there is sometimes. [1] https://www.togetherforthetote.com https://www.togetherforthetote.com
      - foxhound6 7y ago
        Sorry I should have been more specific. Love your stuff on Twitter by the way.
  - hogFeast 7y ago
    Lol. The people who have the best models are often billionaires, have funds that manage hundreds of millions, and will employ hundreds of people on their own dime...they don't need to make a percentage on your $1000 investment. Btw, these websites already exist but no-one on there makes any money. You don't need a lot of capital to make money in sports betting (because you turn your capital over so many times a year), there is no reason for a good sports bettor to use those sites to publicise themselves.
- Spooky23 7y ago
  If you know the business you can focus on one area. I had a friend who was a fanatic that specialized in workout performance. He was in his late 70s and would use twitter, forums and other sources to identify about a dozen horses that he would track tightly in the spring. Then he'd go to Belmont in the Spring and basically live at Saratoga in the Summer. He'd watch workouts, assess the horses, watch the trainers and jockeys and gather various datapoints. I was skeptical at first, but he at a minimum made enough to live in Saratoga for the meet, which is not cheap, and funded alot of his toy purchases from his winnings. He'd call out of the blue give me picks that were good winners about 50% of the time. Basically, there were two ways he made money -- he followed his dozen horses for the high quality stakes races, and was able to eliminate shitty horses for the lower quality races based on trainer or workout performance. For the low quality races, he would hit a few wins/exactas a week (exacta = bet on 1st and 2nd place), and for the higher quality races he would do more exotic bets (Pick 6, trifecta, exacta) This guy loved horses, the track, and the people around it. He was a widower and had lots of time on his hands. Definitely a labor of love that kept him sharp for a long time.
- hogFeast 7y ago
  There are a ton of papers on sports analytics. The public stuff isn't cutting-edge, most of the results about profitability are totally bogus but you will quickly understand what people at the cutting edge are likely doing. The great thing is though: you can start with a few thousand and bet in small markets, you don't need to do anything sophisticated there. The big syndicates also aren't that secretive anymore. The founders and employees have got too rich (some are billionaires), and they have had to build a profile to hire (if you attended an elite uni in the UK, they recruit there...the largest syndicate in the UK has hundreds of programmers now). And Bill has been talking about this for years. I don't understand why but he was publishing papers and giving talks in the 1990s. The impression I get is that he wants other people to know he is smart, and already has more money than he needs (this was true even in the 1990s). Something that most people here won't get though: the models don't matter. The largest syndicates hire clever people but the real money is made in finding liquidity and people to take the other side. The big UK syndicates got large because they worked with bookies in Asia (the largest syndicate is run by someone who ran an Asian book) who needed people to take risk.
- throwaway40324 7y ago
  I'm from Georgetown, KY / Lexington area (the horse capital of the world) and I have a program where I can input data from upcoming races, including weather predictions, that can simulate a potential race over and over, resulting in some meaningful numbers. It's a bit like a brute force training/ML approach. Data resulting from this approach is FAR more valuable than an actual bet.
kratom_sandwich 7y ago
I think that data availability is the problem, rather than the statistical technique (ML vs. "traditional" statistics). How would you - as an "outsider" - even get hold of the countless variables which determine outcome? If the horse has performed well in a training session? If the jockey had been on vacation prior to a race? Etc. ...
- phillc73 7y ago
  The Hong Kong Jockey Club, in particular is excellent with regards to the free data they provide. They really do try to publish everything, from track work times to horse weights.[1] Other jurisdictions not so much. There are various scrapers available to obtain generic race card information[2] and APIs from some bookmakers/exchanges.[3] A couple of years ago I read an article lamenting the lack of open data in horse racing, which still largely holds true today. Unfortunately, I can't find the article now (it was either on the Paulick Report or Bloodhorse website), but here's a more recent one from May this year calling for more open data.[4] In fact, a question was asked on HN a couple of years ago about an open horse racing database and as far as I know the answers there are still valid.[5] In general, there is data available for US, UK, IRE, Australian racing (the main jurisdictions) but they have to be paid for.[6][7][8][9] [1] https://racing.hkjc.com/racing/english/index.aspx https://racing.hkjc.com/racing/english/index.aspx [2] https://github.com/4A47/rpscrape https://github.com/4A47/rpscrape [3] https://docs.developer.betfair.com/display/1smk3cen4v3lu3yomq5qye0ni/API+Overview https://docs.developer.betfair.com/display/1smk3cen4v3lu3yom... [4] https://www.bloodhorse.com/horse-racing/articles/232487/thoroughbred-idea-foundation-calls-for-affordable-data https://www.bloodhorse.com/horse-racing/articles/232487/thor... [5] https://news.ycombinator.com/item?id=7779117 https://news.ycombinator.com/item?id=7779117 [6] https://www.betwise.co.uk/smartform https://www.betwise.co.uk/smartform [7] https://www.proformracing.com/ https://www.proformracing.com/ [8] https://southcoastdatabase.com.au/ https://southcoastdatabase.com.au/ [9] https://www.equibase.com/ https://www.equibase.com/
  - primeradical 7y ago
    I work in the industry, can confirm, the tactics used to dissuade people for aggregating horse racing data so we can sell $2 PDFs are extremely counter-productive and reflect the age of the industry. Several orgs, including Equibase (US-based, the gate keeper of a good portion of handicapping data) will regularly send cease and desist orders to people who attempt to automate aggregation of data even with free, publicly available content. That's at least half the reason PDFs are used when customers purchase data access, to make aggregation harder (you should see some of the white space, character encoding fuckery they use to throw off aggregators). I suppose some of this often depends the quality of the data as well. Most data entry happens at the track during the race by a human, none of the data collection about races or the horse stats are collected by a computer, it's 95% hand entered. That also goes for pedigree information and other statistics including medications, weights, etc. And 100% of that is usually self-reported. Much of the current handicapping in the industry is everyone trying to protect their personal mountains of data. Tech-minded people would love to provide open, controlled, API services so that people can do what they will with our mountains of data. But "giving it away for free" is a non-starter for the good ole boys at the top..
PaulRobinson 7y ago
The cutting edge for a lot of people right now is using Bayesian approaches over frequentist approaches. Interestingly, a frequentist approach (i.e. just looking at what happened in the past, producing a rough regression, and then assembling ensembles of such models), is likely to get you +EV in markets that don't get frequented by big hitters. It'll help you in your high school football league, for example, but it won't help you make money in the NFL markets. Benter has written about his approaches, and has hinted at various ways of working he used. The frequentist approach he took most likely still stands up in Hong Kong because there are only two tracks, variance is limited, the horse population is limited etc. That doesn't apply anywhere else in the World, or for almost any other sport. There is a whole World out there that you are yet to discover, but it's unlikely the syndicates will welcome you. More money in the market on the other side of the table is always welcome (that's liquidity, always helpful for the market to be bigger), but more money on your/their side makes the job harder. Always remember to only bet what you can afford to lose, and good luck!
- audiometry 7y ago
  Can you describe how/why a bayes approach would work better than frequentist in situation like NFL?
  - PaulRobinson 7y ago
    The frequentist approach states (in simple terms), the observations I have made will help me predict future outcomes, because they are a good measure of the inherent probabilities inherent in what I'm observing. The bayesian approach says the data is fixed, and the probabilities might change. If I look at the stats for the NE Patriots, for example (http://www.nfl.com/teams/newenglandpatriots/statistics?team=NE http://www.nfl.com/teams/newenglandpatriots/statistics?team=...) I see that their first down conversion rate is 30/56 so 53%. Imagine I am watching a game and I am betting in play. I am offered odds of 1.9 (decimal odds) that they will convert the third down they are about to try and convert into a first down. The frequentist approach says given the implied odds of past behaviour is 1.88 (we can convert percentages into decimal odds by dividing 100 by the percentage, so 100/53 = 1.88), and I am being offered 1.9, I should bet! Kelly says I should bet 0.78% of my bankroll, as I have an edge here. Now, does that make sense to you? 30/56 is what happened in the past, and we're using that as an indicator as to what to do next. Would you take that bet? The problem with this approach, I think, is that frequentist approaches whilst practical assume there is an underlying probability we can uncover by measuring it. The Bayesian approach (in simple terms), says we can't be that precise, and the probabilities change over time based on the context. This makes more intuitive sense: the probabilities in poker are fixed and calculable, it seems to me they are much less so in NFL games. In the Bayesian approach, we broadly need to think of a probability distribution and understand our confidence interval, and we use priors and observations to help us calculate both. Doing some maths we might say the chance of the Patriots getting the third down conversion is with 95% confidence the chance of between 51.5% and 54.5%. Well, now the 1.9 on offer isn't quite so sweet - it's within the confidence interval, albeit off to the edge. Getting to that distribution and narrowing your confidence interval (it would be great if we could say it was 52.8% to 52.9%, for example), and then figuring out how to use Kelly accordingly, is relatively state of the art. Doing this in the NFL might be tricky because the data sizes are relatively small - the confidence intervals might be too broad. Also, the frequentist approach is provenly useful in some situations: Bill Benter is richer than either of us, and I don't believe he ever used bayesian statistics. People often think of gamblers as slightly grimy/shady characters with a gold chain and a wad of bills in their hand. That might happen, but all the ones I speak to spend their weekends reading PhD theses from maths and finance departments where people have been trying to figure out this stuff. I hope this answer gives you a flavour.

kratom_sandwich 7y ago

But where would you even start discovery? I guess no one with a remotely useful idea wants to share it ...I did a bit of Googling after the article came out last year and apart from a few research papers and that Wired article about Joan Ginther, not much comes up. I'm surprised he is openly talking about everything, but I suppose many others have caught up and there's not much to lose ...

yalooze 7y ago
Yeah, that's the conclusion I am coming to as well. The people with good models aren't going to talk about them. Even if they need investment they'll seek it privately. I guess I was thinking of a market where you can 'invest' your money in someone's model and they take a percentage of your winnings (that's their incentive for adding your money to the pool). Then it's a democratised race for whoever has the best model.
- lscharen 7y ago
  Here's an interesting article about an NBA sports-better that developed a high-fidelity model to simulate NBA games. https://www.espn.com/blog/playbook/dollars/post/_/id/2935/meet-the-worlds-top-nba-gambler https://www.espn.com/blog/playbook/dollars/post/_/id/2935/me...
- phillc73 7y ago
  Here's an idea from an Australian website where you can build a very basic model, using past results (which are of course no guarantee of future performance) and then sell it. Or alternatively, you can buy "qualifiers" for these systems for a small fee.[1] [1] http://www.puntingform.com.au/account/systems-dashboard/#toppublicsystems http://www.puntingform.com.au/account/systems-dashboard/#top...
- foxhound6 7y ago
  In horse racing where the bets are done via parimutuel[1], the challenge is that there's an upper bound on how much you can "invest" in a given wager without changing the odds. This makes it less likely that an expert will seek outside investment unless their model has them betting "wide" (lots of small bets vs. a few large ones). [1] https://en.wikipedia.org/wiki/Parimutuel_betting https://en.wikipedia.org/wiki/Parimutuel_betting
  - PaulRobinson 7y ago
    Not everywhere does Parimutuel - the UK and Ireland for example have basically rejected it despite massive investment into on-course "Tote" systems.
    - phillc73 7y ago
      Is there any hope for Together for the Tote?[1] I like to dream there is sometimes. [1] https://www.togetherforthetote.com https://www.togetherforthetote.com
    - foxhound6 7y ago
      Sorry I should have been more specific. Love your stuff on Twitter by the way.
- hogFeast 7y ago
  Lol. The people who have the best models are often billionaires, have funds that manage hundreds of millions, and will employ hundreds of people on their own dime...they don't need to make a percentage on your $1000 investment. Btw, these websites already exist but no-one on there makes any money. You don't need a lot of capital to make money in sports betting (because you turn your capital over so many times a year), there is no reason for a good sports bettor to use those sites to publicise themselves.
Spooky23 7y ago
If you know the business you can focus on one area. I had a friend who was a fanatic that specialized in workout performance. He was in his late 70s and would use twitter, forums and other sources to identify about a dozen horses that he would track tightly in the spring. Then he'd go to Belmont in the Spring and basically live at Saratoga in the Summer. He'd watch workouts, assess the horses, watch the trainers and jockeys and gather various datapoints. I was skeptical at first, but he at a minimum made enough to live in Saratoga for the meet, which is not cheap, and funded alot of his toy purchases from his winnings. He'd call out of the blue give me picks that were good winners about 50% of the time. Basically, there were two ways he made money -- he followed his dozen horses for the high quality stakes races, and was able to eliminate shitty horses for the lower quality races based on trainer or workout performance. For the low quality races, he would hit a few wins/exactas a week (exacta = bet on 1st and 2nd place), and for the higher quality races he would do more exotic bets (Pick 6, trifecta, exacta) This guy loved horses, the track, and the people around it. He was a widower and had lots of time on his hands. Definitely a labor of love that kept him sharp for a long time.
hogFeast 7y ago
There are a ton of papers on sports analytics. The public stuff isn't cutting-edge, most of the results about profitability are totally bogus but you will quickly understand what people at the cutting edge are likely doing. The great thing is though: you can start with a few thousand and bet in small markets, you don't need to do anything sophisticated there. The big syndicates also aren't that secretive anymore. The founders and employees have got too rich (some are billionaires), and they have had to build a profile to hire (if you attended an elite uni in the UK, they recruit there...the largest syndicate in the UK has hundreds of programmers now). And Bill has been talking about this for years. I don't understand why but he was publishing papers and giving talks in the 1990s. The impression I get is that he wants other people to know he is smart, and already has more money than he needs (this was true even in the 1990s). Something that most people here won't get though: the models don't matter. The largest syndicates hire clever people but the real money is made in finding liquidity and people to take the other side. The big UK syndicates got large because they worked with bookies in Asia (the largest syndicate is run by someone who ran an Asian book) who needed people to take risk.
throwaway40324 7y ago
I'm from Georgetown, KY / Lexington area (the horse capital of the world) and I have a program where I can input data from upcoming races, including weather predictions, that can simulate a potential race over and over, resulting in some meaningful numbers. It's a bit like a brute force training/ML approach. Data resulting from this approach is FAR more valuable than an actual bet.

kratom_sandwich 7y ago

I think that data availability is the problem, rather than the statistical technique (ML vs. "traditional" statistics). How would you - as an "outsider" - even get hold of the countless variables which determine outcome? If the horse has performed well in a training session? If the jockey had been on vacation prior to a race? Etc. ...

phillc73 7y ago
The Hong Kong Jockey Club, in particular is excellent with regards to the free data they provide. They really do try to publish everything, from track work times to horse weights.[1] Other jurisdictions not so much. There are various scrapers available to obtain generic race card information[2] and APIs from some bookmakers/exchanges.[3] A couple of years ago I read an article lamenting the lack of open data in horse racing, which still largely holds true today. Unfortunately, I can't find the article now (it was either on the Paulick Report or Bloodhorse website), but here's a more recent one from May this year calling for more open data.[4] In fact, a question was asked on HN a couple of years ago about an open horse racing database and as far as I know the answers there are still valid.[5] In general, there is data available for US, UK, IRE, Australian racing (the main jurisdictions) but they have to be paid for.[6][7][8][9] [1] https://racing.hkjc.com/racing/english/index.aspx https://racing.hkjc.com/racing/english/index.aspx [2] https://github.com/4A47/rpscrape https://github.com/4A47/rpscrape [3] https://docs.developer.betfair.com/display/1smk3cen4v3lu3yomq5qye0ni/API+Overview https://docs.developer.betfair.com/display/1smk3cen4v3lu3yom... [4] https://www.bloodhorse.com/horse-racing/articles/232487/thoroughbred-idea-foundation-calls-for-affordable-data https://www.bloodhorse.com/horse-racing/articles/232487/thor... [5] https://news.ycombinator.com/item?id=7779117 https://news.ycombinator.com/item?id=7779117 [6] https://www.betwise.co.uk/smartform https://www.betwise.co.uk/smartform [7] https://www.proformracing.com/ https://www.proformracing.com/ [8] https://southcoastdatabase.com.au/ https://southcoastdatabase.com.au/ [9] https://www.equibase.com/ https://www.equibase.com/
- primeradical 7y ago
  I work in the industry, can confirm, the tactics used to dissuade people for aggregating horse racing data so we can sell $2 PDFs are extremely counter-productive and reflect the age of the industry. Several orgs, including Equibase (US-based, the gate keeper of a good portion of handicapping data) will regularly send cease and desist orders to people who attempt to automate aggregation of data even with free, publicly available content. That's at least half the reason PDFs are used when customers purchase data access, to make aggregation harder (you should see some of the white space, character encoding fuckery they use to throw off aggregators). I suppose some of this often depends the quality of the data as well. Most data entry happens at the track during the race by a human, none of the data collection about races or the horse stats are collected by a computer, it's 95% hand entered. That also goes for pedigree information and other statistics including medications, weights, etc. And 100% of that is usually self-reported. Much of the current handicapping in the industry is everyone trying to protect their personal mountains of data. Tech-minded people would love to provide open, controlled, API services so that people can do what they will with our mountains of data. But "giving it away for free" is a non-starter for the good ole boys at the top..

PaulRobinson 7y ago

The cutting edge for a lot of people right now is using Bayesian approaches over frequentist approaches. Interestingly, a frequentist approach (i.e. just looking at what happened in the past, producing a rough regression, and then assembling ensembles of such models), is likely to get you +EV in markets that don't get frequented by big hitters. It'll help you in your high school football league, for example, but it won't help you make money in the NFL markets. Benter has written about his approaches, and has hinted at various ways of working he used. The frequentist approach he took most likely still stands up in Hong Kong because there are only two tracks, variance is limited, the horse population is limited etc. That doesn't apply anywhere else in the World, or for almost any other sport. There is a whole World out there that you are yet to discover, but it's unlikely the syndicates will welcome you. More money in the market on the other side of the table is always welcome (that's liquidity, always helpful for the market to be bigger), but more money on your/their side makes the job harder. Always remember to only bet what you can afford to lose, and good luck!

audiometry 7y ago
Can you describe how/why a bayes approach would work better than frequentist in situation like NFL?
- PaulRobinson 7y ago
  The frequentist approach states (in simple terms), the observations I have made will help me predict future outcomes, because they are a good measure of the inherent probabilities inherent in what I'm observing. The bayesian approach says the data is fixed, and the probabilities might change. If I look at the stats for the NE Patriots, for example (http://www.nfl.com/teams/newenglandpatriots/statistics?team=NE http://www.nfl.com/teams/newenglandpatriots/statistics?team=...) I see that their first down conversion rate is 30/56 so 53%. Imagine I am watching a game and I am betting in play. I am offered odds of 1.9 (decimal odds) that they will convert the third down they are about to try and convert into a first down. The frequentist approach says given the implied odds of past behaviour is 1.88 (we can convert percentages into decimal odds by dividing 100 by the percentage, so 100/53 = 1.88), and I am being offered 1.9, I should bet! Kelly says I should bet 0.78% of my bankroll, as I have an edge here. Now, does that make sense to you? 30/56 is what happened in the past, and we're using that as an indicator as to what to do next. Would you take that bet? The problem with this approach, I think, is that frequentist approaches whilst practical assume there is an underlying probability we can uncover by measuring it. The Bayesian approach (in simple terms), says we can't be that precise, and the probabilities change over time based on the context. This makes more intuitive sense: the probabilities in poker are fixed and calculable, it seems to me they are much less so in NFL games. In the Bayesian approach, we broadly need to think of a probability distribution and understand our confidence interval, and we use priors and observations to help us calculate both. Doing some maths we might say the chance of the Patriots getting the third down conversion is with 95% confidence the chance of between 51.5% and 54.5%. Well, now the 1.9 on offer isn't quite so sweet - it's within the confidence interval, albeit off to the edge. Getting to that distribution and narrowing your confidence interval (it would be great if we could say it was 52.8% to 52.9%, for example), and then figuring out how to use Kelly accordingly, is relatively state of the art. Doing this in the NFL might be tricky because the data sizes are relatively small - the confidence intervals might be too broad. Also, the frequentist approach is provenly useful in some situations: Bill Benter is richer than either of us, and I don't believe he ever used bayesian statistics. People often think of gamblers as slightly grimy/shady characters with a gold chain and a wad of bills in their hand. That might happen, but all the ones I speak to spend their weekends reading PhD theses from maths and finance departments where people have been trying to figure out this stuff. I hope this answer gives you a flavour.