« Writing Documentation | Main | Promoting FeedJournal »

Tuesday, August 01, 2006

Bots Soon to High-Card Humans

The online poker rooms are flourishing in what seems to be the Eldorado of our time. But not for long. Around the corner lurk the bots, or computer programs designed for playing poker, currently in hard training to win your hard-earned cash.

Ten years ago I was an avid poker player, constantly on the lookout for new games. Most of my friends that I used to play for nickels and dimes are still playing today. They have taken their chip stacks online, and the limits have increased, but from the look of it they are doing well financially. Online poker ten years back consisted of a yearly e-mail tournament and games played over Internet Relay Chat (IRC), including first-generation bots. Online discussions took place on a spam-filled USENET newsgroup; in short, the situation wasn't very exciting. Today, you can hardly enter any web site without seeing advertising for online poker, with poker sites and fora mushrooming. According to Party Gaming, online poker generated $2.6 billion in gross gaming yield and represented 20% of global online gaming revenue during 2005. To call the development during these ten years an explosion would be an understatement.

But artificial intelligence is about to take over the tables. We have already seen it happen in chess, Deep Blue beat world champion Gary Kasparov in 1997; in checkers; and in backgammon. In these games it is believed that the best software program is superior to, or at least on par with, the world's best human players. Poker is somewhat different than chess and backgammon, but they do share a lot of common ground. In poker the cards decide the outcome of a single hand, and in backgammon the dice decide the outcome of a single game. But in the long run the winner is the one who makes the best choices. The skill level of the computer players are steadily improving, but it is not only smart algorithms that make them a tough opponent for a human. Poker players are notorious for playing long sessions and in the wee hours, the quality of play is decreasing at the same rate as the players' eyelids are closing. Poker bots don't have this problem, nor are they affected by the other major shortcoming of a human poker player: being on tilt, or playing suboptimal out of anger from losing a recent pot.

The game of online poker, with its huge revenue, is an attractive market for bot writers. Sure, online poker rooms prohibit bots; but in reality, there are a lot of poker bots that go undetected. With software dominating the world of chess, checkers and backgammon; poker is the next game. Online poker is not the same lucrative business it was a few years ago, the competition has stiffened and I am convinced that bots already have started to take a piece of the action. I predict that in just a few years a majority of the big money winners in online poker are bots. When that situation occurs it is questionable if online poker can survive at the same level as we see today. With more poker bots getting closer and closer to an optimal game, the playing field will be more even, and the earnings will not be enough to beat the poker rooms' fees, unless some unwary humans will stick around to feed them.

Building a poker bot to play in an online poker room where its participation is banned requires more than writing logic needed at the poker table. The poker servers are fighting tooth and nail in a war for their very existence. They are adopting multiple measures of defense: spyware-like functionality that monitors running processes on your computer, pop-up screens (a.k.a. bot challenges), playing patterns, etc. This war between sites and bots is not fought in the open, both sides prefer to keep a low profile here. The sites do not want to scare away their human clientele, while the bots are fighting detection. Because of their clandestine existence, it is difficult to evaluate the exact state of poker bots today. Some universities are doing research in the field and the University of Alberta seems to lead the way.

It is exciting to know that my fellow finalist Daniel Crenna is writing a framework for hooking up poker bots to play against each other. I hope that his endeavor will help budding poker bot authors to improve their software. It will be very exciting to see the final results of this project! A similar commercial product, Poker Academy, is also available on the market.

© Jonas Martinsson 1995-2006