Has Science Trumped Gambling?

Tue, 03 Jan, 2017

A review of “The Perfect Bet” and other indicators of the possible demise of the gambling industry as we know it

By Michael Franklin and Will Mace

A major milestone was reached last year when AlphaGo - Deep Minds artificial intelligence (AI)  programme – took many by surprise to beat Go’s 9-dan champion Lee Sedol 4 games to 1.

That milestone has just been taken to another level entirely in the Brains vs AI; Upping the Ante poker tournament at the Rivers Casino in the US – where some of the best poker players in the world have just been soundly beaten, over 120,000 hands of Heads-up, No-Limit Texas Hold’em, by Carnegie Mellon’s AI programme - Libratus.   Libratus won pretty comprehensively, taking over $1,700,000 from the human pros and giving rise to lots of ‘fun’ headlines like ‘humans fold, AI comes up Aces’.  #BrainsVsAI 

Poker is considered a much harder challenge for AI than other strategy games like Chess or Go as it has to base its strategic reasoning on imperfect, or partial, information.  That it has now surpassed a humans ability to reason in such circumstances is considered as having set a new benchmark for AI research and development.  Libratus is much more than just a poker programme of course, it has translational AI capability, focusing on arguably bigger issues than poker – such as curing cancer and creating geo-political strategies. But what might this success of AI mean for the gambling industry – operators and customers alike?

With the massive leaps forward in the fields of wearable computers, artificial intelligence, statistical modelling, data analysis and others, Adam Kucharski’s recent book 'The Perfect Bet' questions whether science and maths is taking the luck out of gambling.

A statistician who by day models the spread of infectious diseases, Adam Kucharski has recently turned his expertise to the industry of gambling. Based on both anecdotes and very well researched examples, The Perfect Bet explores the history of applying science to gambling; from flawed early studies of roulette results, to rudimentary analysis of form, to bulk buying lottery strategies when jackpots are high enough. Kucharski also considers the future of applying science to gambling. Machine learning poker programmes that automatically eliminate weaknesses as soon as they are identified. Super sophisticated predictive algorithms - based on unfathomably large data sets - that model future outcomes in sporting events. Measuring a myriad of factors to accurately determine where a roulette ball is likely to land. He goes so far as to suggest that - theoretically at least - there is nothing remotely random about where a roulette ball lands.  His core question is that 'as technological capabilities continue to improve is the luck being taken out of gambling?'

If IBM Watson gave racing tips, or predicted Premier League outcomes – based on crunching thousands if not millions of data points, how much more successful might it be than even the most data driven punter? Predictive models have long existed of course – and many are very effective - but they and the associated betting strategies have always been closely guarded by the syndicates that developed them.  The game changing impact on the industry may come when the output of much improved predictive models becomes universally available.   

If every punter could have a much higher degree of statistical certainty about an outcome than has ever been possible before – how would it change their betting behaviour?  If every punter had free and easy access to Watson's predictions, and placed their bets accordingly, then what would be the impact on prices and margins?  What would be the impact on sports betting as an industry? 

An industry built on probabilities, odds and statistics, there has always been an inherent link between science and gambling, but almost across the board the information inequalities between bookmakers and punters are being eroded by advances in science and technology.  The interesting question for the industry is to consider the effect of information parity. What does it mean for the house advantage? Will margins reduce to unsustainable levels? Will prices reduce to levels uninteresting for punters? If it becomes common practice for punters to look to science, it is not unreasonable to think it could shrink or at least significantly change the shape of the industry altogether.

Kucharski, and indeed common sense, recognises however, that there can be no perfect sporting algorithm that can 'guarantee' its predictions.   The inherent unpredictability of the lucky bounce, the unfortunate deflection, or the rub of the green, almost certainly means there is a future for the sports betting industry although it may be a little different from the future widely assumed today.

As a follow up to these musings, we’ll be conducting a few experiments to test some of these theories – assessing if Artificial Intelligence models are better at picking winners at the races and in the Premier League than well informed punters, and then modeling what this could then mean for the industry if those picks were widely and freely available.