Hacker News Reader

Comment by dchuk | Hacker News Reader

(Preface: I'm not particularly interested in the ethical debates around gambling with this comment, though I personally think it's something people should be allowed to do.).

I've been slowly getting into Machine Learning for a while now, and have been looking for projects to accelerate my focus. I also happen to like to gamble (responsibly), so I've recently started researching how to apply Machine/Deep Learning to Baseball betting, and possibly other sports.

It would be mostly a side project I would imagine, I have no dreams of becoming wealthy from this, and since it mixes scraping (which I really like to play with) with machine learning, it will probably be entertaining at least. I'm very curious if I could actually get decent success with it though.

Anyone done this before? Or know of stories/books about this?

View on HN

This discussion could be of interest: https://news.ycombinator.com/item?id=15523186

tldr: keep it fun; if you bet professionally, you'll run into hurdles from bookies.

Thanks!

As with many other areas, the best money might well be in selling the shovels instead of mining the gold claims.

Develop a good system with decent results, sell the recommendations. I've watched this play out in daily fantasy sports over the last few years and I'm convinced it's second only to running the betting sites themselves, only with infinitely less hassle.

I too have considered this [0]. You don't really need to scrape, old data is there via retrosheet and current data is available as XML and JSON via gdx.mlb.com (e.g. for 2017 [1]). Of course, beware the terms of use, but if it's just for yourself, I'd say no prob.

As for predictive analytics themselves, you can dig into all the metrics you want, but training/backcasting is how I'd approach it.

0 - https://github.com/cretz/software-ideas/issues/43

1 - http://gdx.mlb.com/components/game/mlb/year_2017/

Worth noting that this field is reaching diminishing returns due to analyses. You might be 2% better than a stock analysis, but that's usually not a sufficient edge to exploit profitably unless you're talking HUGE numbers.

Finding a problem someone has, solving it with analysis, and offering to license/sell it to them (once it's proven to work) is probably more lucrative.

Unless you're chasing the variable rewards skinner box. In that case, I'd probably just take up a cocaine habit -- it'll be cheaper.

Late response, but figured I'd mention that we have new variables now that we didn't always. Pitch FX, fielder location, etc. Not that it makes a huge different, but playing daily and averaging out over time, it can really affect matchups (again, over a large enough sample size).

I think you would be better off using more well known statistics for betting (ie. multivariate regressions, t-tests, etc..). These stats methods are less of a black box than ML (neural networks, etc..) so you'll be able to understand why they have chosen specific bets and you'll be able to tweak them easier.

ML is often just stats under the hood in most cases -- I think I am targeting neural networks specifically with this comment.

Yeah, I used Machine Learning as a blanket term for both potentially Neural Nets as well as a traditional statistical methods. Good call out though.

You might, but others have a several decades head start on you. Baseball odds started behaving quite differently when the first high-dollar computer-reliant betting syndicates started operating way back when. The rule of thumb back then was that if you could identify a systematic way to extract a decent income from gambling, the size of that income would decline rapidly to equal the minimum wage within 2 or three years.

Baseball is appealing because the game state can be accurately modeled as a state machine. Don't let this fool you -- the impact of variation is much higher for a single baseball game, than for another sport.

I've won 4 of my last 9 baseball "futures" bets, with more than 2-1 payoffs (some were to make the playoffs, or o/u games won, etc.) That's the way to bet on baseball -- use a large sample size. It's easy enough, once the futures lines come out, to find bets that clearly have a positive expected outcome, when using advanced sabermetrics (the baseball term for "statistics" aka Moneyball).

Unless this is a passion, take the time you have allocated to running the analysis, and use it in a healthy manner. Or multiply it by your hourly wage. Then read fangraphs or BP to find the exploitable futures lines.

Daily fantasy baseball is a killer due to variation. Even the best can go 0-for-5 on an off night. Winning at the bottom levels of dollar amounts is easy just by taking into account a few extra factors, but the upper levels are more difficult.

DFS NBA is much more predictable, but then everyone else knows that too so it's also much more competitive. You've got to be seriously dialed in to have any edge at all.

I feel like the PGA is still the softest DFS sport. Lots of unexplored territory there, but not giant amounts of money to win.

Nate Silver: Signal and the noise, Joesph Buchdahl: Squares & Sharps, Suckers & Sharks

I am heavily involved in gambling (trading), posted a simple article on using ML as well recently:

https://news.ycombinator.com/item?id=15739468

Feel free to join our slack group:

https://github.com/liampauling/betfair

You might enjoy this old article about Bob Voulgaris on ESPN. He was/is a very successful NBA gambler that over time started to use some machine learning techniques.

http://www.espn.com/blog/playbook/dollars/post/_/id/2935/mee...

Thanks for the link!

Reply on news.ycombinator.com