Robots have already famously conquered humankind at chess. In 1997, Deep Blue, a chess-playing computer developed by IBM, beat the very best planet earth had to offer, world champion Gary Kasparov, over six games: two games to one and six draws.
Poker, unlike chess, is a game of imperfect information, and that makes it much harder for artificial intelligence to “solve.” But, in the field of heads-up limit hold’em, at least, they are getting close.
But the answer to the question above is that the robots are already here and they are already taking over. Yep, it’s time to run for the hills.
Computers are Everywhere
We live in a brave new world of computer intelligence and the gambling landscape is no different. Bots use complex algorithms to scour the betting markets for stray odds. They search the financial markets looking for arbitrage opportunities. They exploit the mistakes of bookmakers and other, more human gamblers, like us. These bots can be linked directly into betting exchanges, placing bets fearlessly and consistently, faster and in more volume than we can. They have an advantage over the average gamble, pure and simple.
That gambling syndicates are using computers to gain an edge in sports betting should not be cause for the average bettor to lose much sleep, though. They exist to exploit bookmakers’ mistakes, and if we do encounter them occasionally on the betting exchanges – well, we can still win. Just like we can against bookmakers, at least some of the time.
But the existence of bots in online poker is more alarming. Poker is a zero sum game, which means the bots beat us they take our money, and that really isn’t fair. But just how common are bots in online poker, and just how skillful are they?
Last year, the University of Alberta’s Computer Science Department unveiled “Cepheus,” the latest in its two-decade-long program to develop poker-playing artificial intelligence. Of course, Cepheus is far more sophisticated than anything you’re likely to find plugged into your favorite online poker site, but it does represent just how far AI has come and what the online poker bots of the future might be capable of.
Cepheus uses an algorithm called CFT which its creators claim has created game theoretical strategy for heads-up limit hold’em so close to optimal that “it can’t be beaten with statistical significance within a lifetime of human poker playing.”
Cepheus has “taught” itself to play almost perfectly by playing trillions against itself and learning to “regret” and remember each decision that failed to result in the optimal outcome, rendering it supposedly unexploitable by mere mortals.
As a game of imperfect information, game theory algorithms developed for poker can never quite be truly perfect; instead computer scientists must approximate a “perfect” poker strategy, which is defined as a strategy that cannot be exploited by any other counterstrategy.
Because Cepheus has now played more hands of poker than the entire human race, this is exactly what the creators of Cepheus believe they have achieved.
But this is limit hold’em; no limit hold’em, with vastly more variables, is a different proposition altogether. A recent study placed limit hold’em 14th in a list of strategy game complexity, while no limit was 140th.
The most sophisticated no limit hold’em bot is “Claudico,” developed by Carnegie Mellon University. In April 2015 it was defeated by a group of top online pros led by Doug “WCGRider” Polk over 80,000 heads-up hands. It finished fifth of five, although it could probably still cream this writer for his entire bankroll.
Of course, these bots have been developed not just to whup humankind at poker; they will ultimately have other applications, helping humans negotiate or make decisions in situations where we can’t know all the facts.
How to Spot a Bot
But no-limit-hold’em-playing bots, designed purely to take our money, do exist, and the chances are, if you’re an online player, you may have unwittingly played against one. Just type “pokerbot” into Google and you’ll be greeted with countless sites offering these bots commercially.
Commonly these are “plug and play” bots that you plug into an online poker site and allow it to make some level of game theory-optimized decisions for you while you put your feet up. It’s cheating, of course, and it’s against all online poker sites’ terms and conditions, and discovery will get you banned from the site for life.
Trusted online poker sites invest time and resources to sniff out these bots, of course, using algorithms that flag up “non-human” behavior. A player that plays too perfectly, or fails to take regular breaks, for example, might be considered suspicious. Less trusted online poker sites have been accused of not doing enough to combat bots, who let’s face it, pay their rake just like everyone else.
The prevalence of bots on online poker sites is anyone’s guess, but they are thought to be most commonly found playing in volume at low stakes ring games or tournaments, where they are most likely to have an edge.
Could Bots Kill Online Poker?
There have, however, been instances of more sophisticated bots playing higher. A recent poster to the 2+2 forums confessed to being a “botter” and offered readers a fascinating insight into the world of a high-tech poker cheat.
His was no plug and play bot but a program he had coded himself that required constant tweaks and adjustments. He explained in detail how he fooled the poker network and its players by instructing his bot to randomly sit out and take breaks every couple of hours; to never play for more than six hours at a time; to mis-click every so often; to occasionally type comments into the chatbox, and to frequently change tables. In short, the bot’s behavior was very human indeed.
Could bots kill online poker? Well, yes, of course they could. As they become smarter and more ubiquitous, humans will naturally stay away. But as bot technology develops, so must technology to counter it, and as the cheats play an ever more sophisticated game of cat and mouse with online poker sites, it’s up to the operators to stay vigilant.