For most poker players, the term ‘bot’ conjures up images of malicious digital creations tirelessly extracting chips from anyone unlucky enough to encounter one. In most cases this paranoia is greatly exaggerated, but recent movements in the field of artificial intelligence (AI) suggest there is reason to be concerned for the future. In fact, experts in the field suggest it’s not long before even the most hardened of Vegas pros will have to yield to superior artificial minds.

AI specialists and poker pros alike have spotlighted limit hold’em as the game most likely to be solved first by computer programs. Limit hold’em is considered ‘solvable’, as the game consists of a series of defined decisions with betting in fixed increments. And so, in August 2007, the latest ‘Man versus Machine’ event was designed to challenge this theory. Two top poker pros were pitted against the very latest in limit hold’em AI to find out exactly how close it is to ‘cracking’ the game.

Over a two-day period consisting of four 500-hand duplicate matches, seasoned pros Phil Laak and Ali Eslami took on ‘Polaris’, a computer program designed by the Computer Poker Research Group based at the University of Alberta in Canada. It comprised two main types of bots, one of which ‘learns’ and adapts to an opponent’s playing style while the other uses optimised game theory to make each decision.

In fact, the only time the learning approach was adopted during the tournament was in the third session, where it unfortunately encountered some problems. The absence of an effective learning presence arguably contributed to the human players’ success. ‘I was able to repeat patterns in my play over and over,’ says Eslami. ‘Against a human, after using a pattern a couple of times, I would probably switch gears to throw my opponent off. Since each hand stood in a vacuum, I was able to find patterns that worked and use them perpetually.’

THE FALL OF MAN

Even with this bug in the machine, the final result was agonisingly close. It was certainly close enough to suggest that when the chinks in its armour are hammered out, there will be little hope of reprieve from Polaris and its ilk. It would seem as though combining the strengths of the learning bot with the efficiency of the optimised game bot could well spell the end for any human attempting to beat the program at its own game.

And there is no shortage of will to win on the bots’ side of the battle. Professor Darse Billings of the Computer Poker Research Group says he is determined to crack the limit problem. ‘Poker is a game of imperfect information,’ says Billings, ‘so developing a form of AI that can effectively think for itself and succeed under these conditions is an extremely challenging and potentially rewarding problem to solve.’

OK COMPUTER

What’s more, Billings believes there could be some great potential benefits for the average poker player too. For a start, a program with the ability to learn and adapt could be an invaluable tool for helping players analyse their own games.

‘When successful AI opponents were created for games like chess, players of all levels found they could learn an awful lot by studying how the computer plays and why it makes various decisions at various times,’ Billings recalls. ‘Indeed, many professionals subsequently incorporated moves they learned from a computer opponent into their regular game. There’s no reason why the same couldn’t eventually be said of poker.’

So, could we soon be learning how to improve our online poker from a robot? Given that bots such as Polaris are theoretically capable of playing the perfect game, there appears to be a very strong argument that we will.

But while this could be a potentially disastrous concept for online limit players, should your average no-limit player be concerned? No-limit’s complex betting structure surely means it’s safe from being cracked by the bots? Billings, however, suggests it wouldn’t be nearly as difficult as it sounds.

‘The differences between limit and no-limit poker are largely mythical, stemming from a lack of fundamental understanding on the part of most poker players. In all likelihood, we could produce a Polaris-calibre no-limit bot within a year if we focused our efforts on that problem.’

JUDGEMENT DAY

So what does the future hold for online poker against this growing robot army? Despite his victory, Ali Eslami offers a rather bleak prediction. ‘Ultimately, once the true equilibrium of the game is discovered, the best any opponent, human or bot, could do over the long-term would be to draw. A true equilibrium strategy cannot be taken advantage of at all.’

However, although poker players are right to be concerned about the impact this sort of research will have on the game, Billings believes much of the anxiety is totally unfounded. ‘However much experience you have, you should be far more concerned about other online players (or indeed lady luck herself) taking your money than your average poker bot,’ he says.

‘It’s not easy to create a bot that can beat the average player, as proven by the dozens of hobbyists who have contributed their own creations to our servers and failed to have any success. Those intelligent enough to create an effective bot will find that there are far more productive and rewarding ways to spend their time.’

Having said that, the lack of a serious threat at present doesn’t mean there won’t be one in the future. With the steadily advancing levels of artificial intelligence, it is reasonable to expect more dangerous creations to appear over the next few years – and these will inevitably be deployed online. Don’t write off the poker robots just yet.

If you think you have what it takes to beat Polaris, you can download the software at www.poker-academy.com