For centuries, man has been trying to outsmart games of chance by means both legal and otherwise. It's interesting to note that some of the very same techniques that today's high-tech crooks use don't differ that much from their counterparts from decades--or even centuries--ago.
With the rise of Big Data and sophisticated new technologies, though, I wondered whether gamblers today are using appreciably different tools from their predecessors. To this end, I recently sat down with Adam Kucharski, author of The Perfect Bet: How Science and Math Are Taking the Luck Out of Gambling. The following are excerpts from our conversation.
Note that his publisher sent me a copy of his book gratis sans further obligation.
PS: What was your primary reason for writing the book?
AK: I've always been interested in games and betting, which I think is true of many people who like math. Even at primary school, my friends and I would run a betting stall at school fairs, which proved extremely popular (if somewhat controversial). It wasn't just about the money; betting was also a good way to play around with ideas about risk and chance. And I later discovered that several other scientists have shared this view over the centuries.
The motivation for the book came during my PhD, when I came across more and more highly technical firms that specialized in sports betting. I realized the relationship between science and betting was growing and changing rapidly - from sports prediction to artificial intelligence in poker - and wanted to explore this in the book.
PS: As you write, people have been trying to game systems for centuries. What's different about today?
AK: The big change has been the better availability of data, computational power and predictive methods. Although scientific betting is often thought of in terms of casinos and card counting, the field has moved on enormously. Three examples stand out. Betting teams are now analyzing sports in detail, including games like soccer and hockey which for a long time were seen as too complicated to predict. These scientific approaches are also being picked up by managers, with sports teams increasingly employing analytics teams alongside traditional coaches and scouts.
Meanwhile, artificial intelligence researchers are developing poker bots that are better than any human, and in doing so are challenging popular wisdom about how players should bluff and make decisions. We're also seeing the lines between finance and betting blurring, with high speed trading algorithms permeating both industries, and debates appearing about whether certain activities are predominantly luck (i.e. gambling) or skill (and hence investing).
PS: Why are larger advantages potentially available in more obscure sports such as golf?
AK: In sports betting, you're competing against other gamblers. To win consistently, you therefore need to find situations where the current odds of a particular result - as dictated by public betting activity - don't line up with the actual chances of that result. The problem is that the bigger the market, and the more data available, the more likely it is that scientific teams will have found a way to identify and exploit these mismatched odds. This is why it's now so hard to make money betting at Hong Kong's race tracks, for example.
Because some betting markets have become highly competitive, new teams would do well to turn to other sports. Things like golf and cricket have traditionally received less scientific attention, so stray odds are potentially more common. Throughout the decades, the most successful bettors have always been the one who are looking where others aren't.
PS: Talk to me about the 2006 "music market" study you cite. What does it prove and suggest?
AK: The "music market" was a study run by sociologist Matthew Salganik and colleagues at Columbia University. The aim was to understand how musical fame spreads: do songs become popular because they are better, or is it luck? The team divided 14,000 online participants into nine separate groups, and let them play and download songs. Eight groups could see what songs were most popular in their group, but one - the control group - could not. Participants in the control group could therefore only judge a song on its merits. Yet the most popular songs in this group were not necessarily the most popular elsewhere.
Across the eight groups who could see what other people were listening to, popularity was highly dependent on the randomness of early choices: once a song had gained some attention, fame was amplified as others noticed and picked it too. The study suggested that fame depends a lot on chance events and social behavior: the top ranked song in one group was a relative flop in another.
A similar pattern can occur in finance or business. If we see or do something that has been a success, we often assume it was down to intrinsic quality, rather than other factors. But in life as in gambling, we can sometimes make a bad decision but have the result go our way, or make a good choice but get unlucky. The key is being able to tell the difference.