Data mining - the poker dilemma with(out) a solution

Swedish mystery player, Isildur1, burst onto the online poker scene with a true bang a few months ago, and he left with an even bigger bang: he dropped a staggering $4.2 million during one session to Brian Hastings, which didn’t just break the record for the most money ever lost/won in a single online session, but raised a few red flags with experts too. How was it possible that a player who had damn near felted Tom Dwan and who had made money off other Full Tilt nosebleed regulars too, be abused in this absolutely dominant manner by a player who – up until that point, was by no means considered as a member of the true nosebleed elite? Those who suspected something was fishy about the deal were later proved to be correct by Hastings himself, who revealed his secret in an interview.

Apparently, the Cardrunners pro had benefited from hand history analysis provided by Brian Townsend and Cole South. Townsend acquired about 20,000 hands of Isildur1’s play by actually playing the Swede himself and supposedly bought another 30,000 hands of data from another source. The upheaval in the wake of Hasting’s confession was great. Players and experts took a stance either in support of the Hastings-South-Townsend group or against them. According to some, what the three Cardrunners players had done was nothing short of collusion. Others championed the idea that simple data mining has nothing to do with collusion and that therefore the Cardrunners guys should not be the targets of any sort of negative fallout from the Isildur1 incident.

Though Full Tilt Poker did decide to revoke Brian Townsend’s Red player status for a month, Isildur1 would not get any of his money back, so from a financial perspective, the incident didn’t amount to a whole lot. What it did achieve though was that it raised awareness about data mining and its effects on actual play as well as on the whole of the online poker industry. Some poker rooms have already begun to implement measures meant to make data mining impossible. Cake Poker has adopted a policy according to which players are allowed to change their nicknames on a weekly basis. Their account username, based on which the poker room will track their activity, remains unalterable.

What exactly is data mining though and why should one fear it? The problem with data mining is that its definition is so broad, it is extremely difficult to point a finger to something and call it data mining. It is generally accepted though that data mining is the process of extracting patterns from a database (which in the case of online poker, consists of hand histories). Data mining can be done by the user (player) in which case it is considered somewhat acceptable. After all, there’s nothing peculiar about a poker player putting his intelligence to work to secure an advantage at the green felt. The 3rd party data mining operations are the ones drawing the ire of the poker public though. 3rd party data mining is done through online poker websites or external software applications which have the ability to collect data and to interpret it too.

The use of such programs has created several problems for poker players. Some of them like to have their records made available for the public, especially those who are successful and who view this as a way to create a name for themselves. Others resent having their losing records out there for everyone to grab. What I have noticed is that such player records are now being used in table chat to belittle opponents, and that may be a major turn-off for some. Collecting data on opponents and using it to secure an advantage should not be a problem: after all, poker players have been doing that in the B&M poker rooms for years. An advantage acquired through observation and through one’s personal efforts is a fair one. The line has to be drawn somewhere though. It is debatible whether a player should be able to pull detailed statistics on a guy he’s never seen in his life, at the click of a mouse basically. Much like in the case of rakeback, different online poker rooms have adopted different stances on data mining. Given the stance of the poker site in question, it is not explained however why Hastings was allowed to get away with the Isildur1 fleecing.

PokerTracker and Holdem Manager are some of the most popular external tracking programs used by data miners. What is the solution to putting this data mining issue to rest? Cake Poker's above described approach is decent enough to untangle the knots, although I’m pretty certain some are going to be outraged by it. Already there are voices claiming that such nickname-changing shenanigans will provide a perfect cover for cheaters, make it easier for them to hide their tracks, and to continue to prey on the innocent. One example often brought up in this sense is the Absolute Poker/Ultimate Bet cheating scandal. In those instances however, the table nick-names were not instrumental in the cheating, the accounts were. Implementing a Cake-like way to do away with table name relevancy would definitely hurt data miners and it could take online poker back to a purer and possibly more enjoyable form.

- Steve, 2010-02-18