- Joined
- Jun 26, 2007
- Messages
- 26,190
- Reaction score
- 6
- Points
- 38
- Age
- 62
Poker, with its outlaw image and Midwest jargon, is an oddball sort of game. But it’s also uniquely academic, in that the game remains ‘unsolved’.
While algorithms exist that can play more-or-less unbeatable games of chess, backgammon and draughts, poker – at least, today’s popular ‘No Limit’ version – remains something only humans can fathom.
At least, that was the accepted wisdom until one January afternoon, when an online poker player known as ‘Trebek’ logged onto his usual poker site…
Counting cards
That day, there were hundreds of games in progress, but Trebek wasn’t there to play: he was conducting reconnaissance. He tapped a few usernames into PokerTracker, a data-mining application popular with online poker players.
He had configured PokerTracker to silently observe several weeks’ worth of games, providing him with a rich cache of data he could crunch to reveal weak players – ‘fish’ with potentially profitable weaknesses.
As Trebek prowled for prey, three players – ‘1forthethumb’, ‘full_tilting’ and ‘0_drunkenboxer’ – caught his eye. However, as he pulled up their data, he realised something remarkable – each of the three played an identical strategy.
He had collected dozens of statistics on each – how often each player went to ‘showdown’, how often they made a continuation bet on the turn, how often they raised on the river – and each stat was similar.
Furthermore, these players spent hours playing each day, with no break. Across the whole poker site, among tens of thousands of players, nobody put in more hours than these three players.
Rise of the poker bots
Trebek had been looking for fish. Instead, he’d found bots. “I just sat back in my chair stunned, watching them for a while, trying to absorb the significance of what I was watching,” he posted to his poker newsgroup. “It’s one thing to talk about bots, but it’s quite sobering to actually see them playing in front of you.”
Poker-playing software is not a new development. In 1994, four years before IBM’s Deep Blue defeated Russian Grand Master Garry Kasparov, researchers at the University of Alberta in Western Canada had just finished developing a checkers playing program called Chinook (which gets its name from the warm winds that blow through the Rockies).
Chinook had become the best checkers playing entity on the planet, and researchers were hungry for a new challenge. They found it in poker. Using machine-learning techniques from Chinook, they set up Poki – a server where poker players could come and pit their wits against their script for play money.
The more people that came, the more information Poki had to model its game on. Soon, Poki was playing a solid and profitable game – it became the first ‘poker bot’.
Perfect poker
Poker bots put the creepers on poker players. To understand why, you need to understand that playing good poker is a lot like trying to lose weight: the information is all there, but few people have the discipline to do it.
Poker strategy is a mixture of maths, psychology and self discipline. Countless externalities – like how much sleep you got or whether you had a bad day at the office – can affect whether you win or lose in a session.
So much so that poker legend Doyle Brunson devotes part of his classic book, Super System, to knowing when you shouldn’t play poker. Take to the tables after an argument with your spouse and chances are you’re toast.
As a result, a large amount of today’s poker strategy guides focus on capitalising on other players’ mistakes – errors triggered by tiredness, distraction or a phenomenon called ‘tilt’ – where a player begins to play recklessly after losing a big pot.
The big fear with bots is that they suffer from none of these frailties. They simply know when to hold ’em, and when to fold ’em.
Open source sharks
These days you don’t have to have university funding to build a poker bot. Various freely available software tools exist, such as the open source product OpenHoldem, which can be used to build a bot in minutes.
A community has formed around ‘botting’ – part hobby, part business. “I’ve been botting for going on three-and-a-half years,” says Matr0x, a member of this poker botting community.
“Truthfully, I can’t remember exactly what I was looking for at the time in Google, but I ended up at www.PokerInspector.com, a utility that can be used to build a basic scripted behaviour type bot. Couple it with a screen-scraper like Frogbot, and you have a fully automated, easy to use bot.”
Matr0x says that for many botters, it isn’t about the money: “If it was all play money I’d still be doing it. This very concept is one which defines who will create a winning bot and who will not.
If it’s purely for the money then you are far less likely to have the dedication required to pull it off. If you can’t wait to sit back down and fine tune your code for countless hours, purely out of drive and fun, you’re 10 steps ahead of the other guy before even setting off.”
Botting is a clandestine hobby. Its devotees are constantly avoiding detection. To be discovered would risk having their poker accounts closed and their winnings confiscated.
“I keep to myself and am not involved in the botting community at large,” says Matr0x. “I work with a few other guys (five in total), we design our own bots and are completely non-public. Having been around a while I do chat with the known key botting figures at times.”
But if botting is so secret, why are there Internet forums dedicated to it? Why program open-source software tools to allow anybody to join the gang?
“Designing a fully automated bot playing solution is very difficult,” says Matr0x. “Why should the basic interface be a secret? The vast majority of folk will never fully commit the time required to design a winning bot anyway, let them feed yours. With the openness also comes the benefits of greater testing and bug finding too, which obviously improves your bot.”
Beating the poker bots
However, just as it’s a challenge to build them, Trebek found them a challenge to play against. Trebek reasoned that since poker is a game of information, knowing they were bots must give him some advantage, surely?
He set about trying to capitalise on them: “I decided that before I turned them in, I was going to spend an entire night trying to exploit them on every single hand… no one else knows they’re bots yet, and if I could figure out a good strategy, I might be able to make some decent money first. I mean, if I can’t beat a bot, I must really suck at poker, right?”
First, Trebek noticed that the three bots would join the site at the same time as each other, take seats at separate tables, and play non-stop for hours on end. As he looked deeper, other aspects of their play became apparent.
When they bet, it was always the same amount. The bots would also ‘think’ for a long time before making a decision – even a fairly trivial one. In poker terms, the bots played ‘tight’ – minimising their risk by only playing very strong hands.
Trebek noticed that the bots followed a crude pattern. When they played a hand, they would call to see the flop and then make a ‘continuation bet’. However, if another player responded by raising them, they would almost always fold – surrendering the pot to their opponent.
Trebek had his strategy. He would wait for the bots to enter a pot and then respond to their bet with a raise. “It worked,” he wrote. “Hand after hand… again and again… I’m going to be rich.”
Hours passed, and Trebek took more money from the bots – until they suddenly began playing differently, seemingly adapting to his way of playing. Had they really adapted? No – as the bot began to type in the chat box Trebek realised that its owner had seen what was happening and had stepped in to take over the account to avoid a complete funds blowout.
The revelation sparked one of the longest threads in the history of poker forum www.twoplustwo.com, demonstrating how controversial a topic bots are in the online poker scene. The disgust seemed to boil down to one fact: poker players felt comfortable playing other human beings.
Introduce the possibility that an opponent might be poker’s answer to Deep Blue and the whole game comes falling down like a pack of cards. But in the melee, poker players might have overlooked one thing: are poker players becoming bots themselves?
Are poker players bots anyway?
In a game as old as poker, it’s amazing to think that the online version is only nine years old. In January 1998, Planet Poker – the first poker site – went online.
Suddenly, at a time when casino cardrooms were being replaced with more profitable slot machines – a boom in poker was sparked as a new generation of players suddenly had 24/7 access to the game.
For these guys, poker wasn’t a stop-off on the way to the racetrack or a chance to whistle at a cocktail waitress – it was the chance to make a lot of money fast from the comfort of their own homes.
Online poker players soon found they could play more hands in a week than a Las Vegas dealer might deal all year. Poker success was suddenly about algebra and game theory, instead of keeping a straight face under a ten gallon hat.
Today, with scores of poker sites and tens of millions of players, the online poker landscape is a constantly escalating arms race of information. Players who take the game seriously have at their disposal a dizzying arsenal of software tools to give them a winning edge.
It’s a rare poker player who sits down at an online poker table without first firing up various heads-up displays and popup windows – add-on tools that calculate odds, screenscrape facts and crunch moves made by opponents in previous games to calculate ‘hand ranges’ – a best estimate of what any given player might be holding.
Performance enhancing poker apps
While poker sites forbid software that plays for you (bots, in other words) other tools are generally fair game. And then there are websites – such as SharkScope, ThePokerDB and Official Poker Rankings – that act as repositories, selling screen-scraped data on every online poker player who’s ever played a hand.
Poker players use so many tools and add-ons, that if you could peer over their shoulder you might think for a minute that they were playing the stock market rather than having a game of cards.
“Whenever I play a tournament, I use SharkScope to look up the other players’ history,” says Dan Carter, who has won over £120,000 in live poker tournaments and hundreds of thousands more online.
“There’s no point playing somebody who’s doing really well – I want to play people who are down a lot of money. It’s made me a lot better over the last few years. It’s had a really big impact.”
It seems that in modern online poker, if you can’t beat the bots, the only option is to join them.
London poker player and journalist, Victoria Coren, who won £500,000 in one game in 2007, puts it best: “Poker players are worried about bots. But it’s becoming harder to tell the difference between man and machine.”
More...
While algorithms exist that can play more-or-less unbeatable games of chess, backgammon and draughts, poker – at least, today’s popular ‘No Limit’ version – remains something only humans can fathom.
At least, that was the accepted wisdom until one January afternoon, when an online poker player known as ‘Trebek’ logged onto his usual poker site…
Counting cards
That day, there were hundreds of games in progress, but Trebek wasn’t there to play: he was conducting reconnaissance. He tapped a few usernames into PokerTracker, a data-mining application popular with online poker players.
He had configured PokerTracker to silently observe several weeks’ worth of games, providing him with a rich cache of data he could crunch to reveal weak players – ‘fish’ with potentially profitable weaknesses.
As Trebek prowled for prey, three players – ‘1forthethumb’, ‘full_tilting’ and ‘0_drunkenboxer’ – caught his eye. However, as he pulled up their data, he realised something remarkable – each of the three played an identical strategy.
He had collected dozens of statistics on each – how often each player went to ‘showdown’, how often they made a continuation bet on the turn, how often they raised on the river – and each stat was similar.
Furthermore, these players spent hours playing each day, with no break. Across the whole poker site, among tens of thousands of players, nobody put in more hours than these three players.
Rise of the poker bots
Trebek had been looking for fish. Instead, he’d found bots. “I just sat back in my chair stunned, watching them for a while, trying to absorb the significance of what I was watching,” he posted to his poker newsgroup. “It’s one thing to talk about bots, but it’s quite sobering to actually see them playing in front of you.”
Poker-playing software is not a new development. In 1994, four years before IBM’s Deep Blue defeated Russian Grand Master Garry Kasparov, researchers at the University of Alberta in Western Canada had just finished developing a checkers playing program called Chinook (which gets its name from the warm winds that blow through the Rockies).
Chinook had become the best checkers playing entity on the planet, and researchers were hungry for a new challenge. They found it in poker. Using machine-learning techniques from Chinook, they set up Poki – a server where poker players could come and pit their wits against their script for play money.
The more people that came, the more information Poki had to model its game on. Soon, Poki was playing a solid and profitable game – it became the first ‘poker bot’.
Perfect poker
Poker bots put the creepers on poker players. To understand why, you need to understand that playing good poker is a lot like trying to lose weight: the information is all there, but few people have the discipline to do it.
Poker strategy is a mixture of maths, psychology and self discipline. Countless externalities – like how much sleep you got or whether you had a bad day at the office – can affect whether you win or lose in a session.
So much so that poker legend Doyle Brunson devotes part of his classic book, Super System, to knowing when you shouldn’t play poker. Take to the tables after an argument with your spouse and chances are you’re toast.
As a result, a large amount of today’s poker strategy guides focus on capitalising on other players’ mistakes – errors triggered by tiredness, distraction or a phenomenon called ‘tilt’ – where a player begins to play recklessly after losing a big pot.
The big fear with bots is that they suffer from none of these frailties. They simply know when to hold ’em, and when to fold ’em.
Open source sharks
These days you don’t have to have university funding to build a poker bot. Various freely available software tools exist, such as the open source product OpenHoldem, which can be used to build a bot in minutes.
“Truthfully, I can’t remember exactly what I was looking for at the time in Google, but I ended up at www.PokerInspector.com, a utility that can be used to build a basic scripted behaviour type bot. Couple it with a screen-scraper like Frogbot, and you have a fully automated, easy to use bot.”
Matr0x says that for many botters, it isn’t about the money: “If it was all play money I’d still be doing it. This very concept is one which defines who will create a winning bot and who will not.
If it’s purely for the money then you are far less likely to have the dedication required to pull it off. If you can’t wait to sit back down and fine tune your code for countless hours, purely out of drive and fun, you’re 10 steps ahead of the other guy before even setting off.”
Botting is a clandestine hobby. Its devotees are constantly avoiding detection. To be discovered would risk having their poker accounts closed and their winnings confiscated.
“I keep to myself and am not involved in the botting community at large,” says Matr0x. “I work with a few other guys (five in total), we design our own bots and are completely non-public. Having been around a while I do chat with the known key botting figures at times.”
But if botting is so secret, why are there Internet forums dedicated to it? Why program open-source software tools to allow anybody to join the gang?
“Designing a fully automated bot playing solution is very difficult,” says Matr0x. “Why should the basic interface be a secret? The vast majority of folk will never fully commit the time required to design a winning bot anyway, let them feed yours. With the openness also comes the benefits of greater testing and bug finding too, which obviously improves your bot.”
Beating the poker bots
However, just as it’s a challenge to build them, Trebek found them a challenge to play against. Trebek reasoned that since poker is a game of information, knowing they were bots must give him some advantage, surely?
First, Trebek noticed that the three bots would join the site at the same time as each other, take seats at separate tables, and play non-stop for hours on end. As he looked deeper, other aspects of their play became apparent.
When they bet, it was always the same amount. The bots would also ‘think’ for a long time before making a decision – even a fairly trivial one. In poker terms, the bots played ‘tight’ – minimising their risk by only playing very strong hands.
Trebek noticed that the bots followed a crude pattern. When they played a hand, they would call to see the flop and then make a ‘continuation bet’. However, if another player responded by raising them, they would almost always fold – surrendering the pot to their opponent.
Trebek had his strategy. He would wait for the bots to enter a pot and then respond to their bet with a raise. “It worked,” he wrote. “Hand after hand… again and again… I’m going to be rich.”
Hours passed, and Trebek took more money from the bots – until they suddenly began playing differently, seemingly adapting to his way of playing. Had they really adapted? No – as the bot began to type in the chat box Trebek realised that its owner had seen what was happening and had stepped in to take over the account to avoid a complete funds blowout.
The revelation sparked one of the longest threads in the history of poker forum www.twoplustwo.com, demonstrating how controversial a topic bots are in the online poker scene. The disgust seemed to boil down to one fact: poker players felt comfortable playing other human beings.
Introduce the possibility that an opponent might be poker’s answer to Deep Blue and the whole game comes falling down like a pack of cards. But in the melee, poker players might have overlooked one thing: are poker players becoming bots themselves?
Are poker players bots anyway?
In a game as old as poker, it’s amazing to think that the online version is only nine years old. In January 1998, Planet Poker – the first poker site – went online.
For these guys, poker wasn’t a stop-off on the way to the racetrack or a chance to whistle at a cocktail waitress – it was the chance to make a lot of money fast from the comfort of their own homes.
Online poker players soon found they could play more hands in a week than a Las Vegas dealer might deal all year. Poker success was suddenly about algebra and game theory, instead of keeping a straight face under a ten gallon hat.
Today, with scores of poker sites and tens of millions of players, the online poker landscape is a constantly escalating arms race of information. Players who take the game seriously have at their disposal a dizzying arsenal of software tools to give them a winning edge.
It’s a rare poker player who sits down at an online poker table without first firing up various heads-up displays and popup windows – add-on tools that calculate odds, screenscrape facts and crunch moves made by opponents in previous games to calculate ‘hand ranges’ – a best estimate of what any given player might be holding.
Performance enhancing poker apps
While poker sites forbid software that plays for you (bots, in other words) other tools are generally fair game. And then there are websites – such as SharkScope, ThePokerDB and Official Poker Rankings – that act as repositories, selling screen-scraped data on every online poker player who’s ever played a hand.
Poker players use so many tools and add-ons, that if you could peer over their shoulder you might think for a minute that they were playing the stock market rather than having a game of cards.
“Whenever I play a tournament, I use SharkScope to look up the other players’ history,” says Dan Carter, who has won over £120,000 in live poker tournaments and hundreds of thousands more online.
“There’s no point playing somebody who’s doing really well – I want to play people who are down a lot of money. It’s made me a lot better over the last few years. It’s had a really big impact.”
It seems that in modern online poker, if you can’t beat the bots, the only option is to join them.
London poker player and journalist, Victoria Coren, who won £500,000 in one game in 2007, puts it best: “Poker players are worried about bots. But it’s becoming harder to tell the difference between man and machine.”
More...