"Bot Ring Discovered On Poker Stars
After being tipped off by this thread on 2+2, PTR has uncovered overwhelming evidence that 10 accounts on PokerStars are being run by computer programs (and not humans). Collectively, they have played at least 8,320,121 hands at the $.25/$.50, $.5/$1 and $1/$2 NL levels, generated $186,572 in rake, and made $57,839 in profit (not including VIP benefits). These accounts are still active and have been observed playing on PokerStars today.
We urge PokerStars to freeze their accounts immediately.
The following 10 accounts stand accused:
In the initial 2+2 thread, Malloc took screenshots from PTR and noted the shocking statistical similarities and consistency in the way 3 accounts moved up and down in stakes. PokerTableRatings has decided to help the effort by analyzing our data. We have found 10 separate accounts that showed remarkable similarities based upon three major factors.
• Extremely similar stats across the board.
• A changing of stakes played on exactly the same day or within a few days of each other.
• A unique and easily identifiable betting pattern from the big blind in unraised pots.
Each of these accounts has essentially the same stats across the board. These stats are much too similar to be explained by mere chance. Here is a table showing the remarkably similar stats for each account. This data can be seen by anyone using PTR Premium.
BOTS.png [ Array Кб | Просмотров: 2 ]
In order to show how similar the stats of these players are compared to the general population we calculated the Euclidian distance for every player at $.25/$.50 NL who has played at least 100,000 hands this year.
The distance was calculated using 7emenov’s stats as a baseline and was based on 6 stats (VP$IP, PFR, AFQ, Flop AFQ, Turn AFQ, River AFQ).
In simple terms, Euclidian distance looks at how far two points are from each other. When you measure the distance between two points with a ruler, that’s an example of Euclidian distance. In this example, however, we are measuring with a ruler in 6 dimensions. The principle is exactly the same as 2 dimensions — the higher the number, the greater difference in their play style. Here is the result from our analysis (671 players fit into our criteria):
bots_vs_50nl_players.png [ Array Кб | Просмотров: 2 ]
As you can see, these 10 players are clustered in the lower left of the chart. They are virtually identical. The biggest distance within the bot group is Daergy at 2.39. The next closest player has a distance of 32.94.
As further evidence, here is a comparison of two of the players using PTR Premium. Again you can see that there is almost no difference in their play style.
BOTS2.png [ Array Кб | Просмотров: 2 ]
BOTS3.png [ Array Кб | Просмотров: 2 ]
Each of these accounts exhibited a remarkably similar pattern in moving down in stakes. As you can see from this chart, they all moved down at nearly the exact same time. This is strong evidence that these accounts were being coordinated by single person / group of people.
BOTS4.png [ Array Кб | Просмотров: 2 ]
Unique Betting Pattern
Another damning piece of evidence is that each of these accounts exhibits a very unique and easily verifiable betting pattern. When they’re in the big blind in an un-raised pot (limpers only) and it’s checked to them on the flop, they will either shove all-in or bluff by making a non all-in bet. If the bot’s bet is called or raised, the bot will always fold to any further aggression.
You can view an example of this betting pattern here https://www.pokertableratings.com/hand-history/stars/33881185259/2914634785
This betting pattern is quite useful in identifying these bots because it means that the bots “fold to flop raise” statistic will always be 100%. If the bot has a good hand, it will always shove. If it decides to bluff, it will take the “bet and shut down” line. We looked at how the 10 accounts in question acted in this situation and in the 1898 times they were raised after betting in the big blind, they folded 1890 times (99.6%). Seven of the Eight outliers occurred in 2008-2009 and may be due to testing their algorithms / manual play.
Bots are playing under these account names. PTR believes it is essential for these types of situations to be pursued actively by the poker community in order to help keep our games safe and fair. We urge PokerStars to immediately freeze these accounts and use the evidence persented here to assist in their investigation. A big nod goes to the folks over at intellipoker who identified a large portion of this group in April and to the 2+2′ers whose thread brought this to our attention."