Main Menu
News
Current
GGG
 
Top of Page
 
Top of Page
 
Top of Page
  | Home   | Index   | Info   | This Week   | Poker   | News   | Email
News

Welcome to the News desk.

The rise of the bot player 12/02/2009
Robert Blincoe

Online poker is a huge business. And easy, players are told. You don't even have to get dressed to play, just sit there in your front room. And, of course, the poker player's dictum of "if you can't see the mark, it's you" (that is, if you can't see a weaker player, you're the one who'll be cleaned out) doesn't apply; you can't actually see anyone else. They could be dressed, undressed - or even, perhaps, "bots": programs written to make money from slower-witted humans.

And, the Guardian has learned, bots are indeed inhabiting some poker sites, where they are winning money from humans. It turns out that the only thing keeping the human gaming community in with a chance is the sheer difficulty, tedium, and cost-effectiveness of running the bots at a profit. The Guardian spoke to one former poker bot master. (He requested anonymity because naming him could affect his employment prospects.)

Flushed out
"The kind of bots that I've run are playing low-limit [less than £50 a day] games picking money off the weak players. They're not a threat to the good players at all. The good players would kick their arses, which is why I put them in the low-limit games," he says.


This bot man managed to make an annual return of around £40,000 working on gaming sites. That's about the same as an experienced programming job, but tax-free - you don't pay tax on gambling winnings.

"My bots aren't bad, they're good enough to beat the weak player, which is all I was ever trying to do. Getting to that stage is interesting and fun. It's the nasty implementational stuff around it [that isn't]. It's the arranging the sequencing of them to all come online and offline, the 'Oh dear, the screen scraper - the stuff that's interfacing with the poker program - is broken again and needs fixing'."

The latter problem arises because the poker sites, aware of the trouble if it is known that computers are making money off humans, deter bots by constantly tweaking their own software. This means constant tinkering for the bot master to keep his players in business. Furthermore, the bots can't play 24/7 or they'd be caught.

"It becomes a really rather dull programming job," says the bot master.

People also think they can spot bots because they win and don't chat during games. But good players don't chat either, especially if they're playing eight or more games simultaneously. This isn't a problem for the bot master, but it's the reason that the online pro John Tabatabai got kicked off one site by mistake.

Writing good bots isn't easy. There's the AI knowledge (a combination of probabilistic calculations with some stochastic search for our bot man's poker army), the programming experience and the understanding of the strategy of the game at which you're planning to win money.

The University of Alberta, Canada, has an army of PhD students working in its poker research group. It has created great bots. Dr Darse Billings, a consultant with the group, says: "If the average schmo, who knows a bit of programming, is going to write himself a bot, then he's going to be donating to the poker community at least for a number of years, if not indefinitely. If there are programmers who are sharp at both programming and poker, then it is conceivable to write a bot that can win. But someone that smart can take a hell of a lot more money legitimately than by writing a poker bot."

Our bot developer is now back doing regular programming. His response to Billings' comments are: "Your average schmo writes a shit bot. The ones you buy off the shelf are rubbish." But for him, the problem isn't writing the bots. It's the logistics. To avoid investigation, you need your bots to win modest amounts and stay below the radar. Instead of one winning £1,000 a day, you need 100 bots that win £10 a day. "But you have to have 100 accounts, 100 user names, and 100 credit cards to get the money out. That's harder than writing the bots themselves."

There are methods of getting multiple accounts, but the online gaming sites monitor for this. Changing IP addresses can be done with downloaded software, but getting multiple credit cards is hard. Once you've exhausted friends and relatives you still haven't got that many different bits of plastic. "I didn't want to go down the organised crime route. Buying cloned credit cards off someone: it's not somewhere I wanted to go," he says.

Writing bots and playing them online is perfectly legal, it's just a violation of the sites' individual terms and conditions. They can't call the police but they can throw you off the site and seize money from your account.

"Limiting the number of bots you can use [through the difficulty of having multiple accounts] is a major stumbling block," says our bot man. But he hasn't ruled out going back into business.

"It was worth my while, if only it hadn't felt like such a job," he says. "If I wasn't able to get a job as a programmer, I'd go back and make money at it. But it took a bit more ingenuity and hard effort, and a lot of fiddly, annoying stuff."

Playing for the house
It's possible he could get a job on the other side of the fence. Many online gaming sites use bots on their sites, but most won't admit it. "It's for liquidity purposes," said one, off the record. Which means they always have someone online to keep the games going.

Anyone fancying a crack at being a bot master may prefer to tackle backgammon, where the software development has already been done for you. In 1995, an IBM academic, Gerald Tesauro, presented the "temporal difference learning technique", and the resultant neural network backgammon software TD-Gammon. Its successors, Snowie 4 and the open-source GNU Backgammon now trounce the world.

Obviously, all the account problems are the same for this game as for poker. And because the software is so well known, the sites offering it have automated systems that look for similarities to the way computers play. It's almost like fingerprinting, and you're likely to be caught. It seems amateurs don't have that much to fear from the bots - yet.