|
Here's some bonus content that I didn't include in the main article because it goes off on a little bit of a tangent, but I thought you guys would find it interesting.
Bonus Content
When people do statistical analysis on player stats while trying to find bot-controlled accounts, they rarely look at individual stats. It's really hard to just look for everyone who played something like 15/13 in full-ring games or 21/17 in six-max or whatever. Instead, they compare sets of stats (like the comparisons I showed between VPIP/PFR, PFR/ATS, etc.) They're able to find a tremendous amount of information from comparing sets of stats compared to looking at individual numbers.
To give a cliff notes version of what's been done to find a lot of major NLHE botting rings, hundreds of accounts with thousands of hands each were analyzed based on certain ratios of specific stats. One particular ratio was always in the range of (for example) 5 to 30 for human players. However, four or five accounts stood out as having ratios in the range of 90 to 95. A closer look at these four or five accounts showed that all of their stats were otherwise virtually identical, and they all were found to be using the same computer program to play poker.
This shows that the pairing of certain stats and their respective ratios can be used as very effective indicators when analyzing data to find bot-like behavior.
|