Based on my previous investigations to identify bots online One and Two, I performed the same procedure for the field in 888 cash games.

This time I dug a little deeper and found that not only weird strategy in multiway-pots unites them all. Moreover, in many of them the strategy in multiway-pots is generally brought up in a normal way, which tells us that they are progressing and are working on eliminating inefficiencies and disguising themselves.

Therefore, I first of all paid attention to their basic strategy and such lines that are incredibly difficult to normalize for the robot. For example, every time I had a question: is it a robot or not, I looked at the sizing distribution in the limped BBvsSB bank, where the BB on the flop is facing with the SB flop check. All the bots that I have discovered use increased aggression in this line and only two exact sizings 1BB and 2BB (no others sizings!).



Rake is taken from the bank on the flop, so the bank on the flop turns out to be a little less than 2 blinds and because of this, the distribution of sizings on the chart looks like this: Either 50% of the pot, or an overbet and there are no intermediate sisings. For a million hands, none of them ever bet 1.3BB 1.5BB 1.7BB Bets. This can happen if you setup calculation in the solver where you allowed BB to use only 1BB and 2BB sizings, and then you tried to reproduce the result at the table.

Also interesting is the distribution of CBET Flop IP in a raised pot. Unlike humans, these bots never use a CBET of 1/3POT on the flop in position against BB.

BTNvsBB Flop Cbet IP Distribution (Bots)


COvsBB Flop Flop Cbet IP Distribution (Bots)


Flop Flop Cbet IP Distribution (EP-MP) vsBB (Bots)


At the moment, it is possible to calculate the strategy on the flop for all possible spots and boards, but this requires a large amount of resources and in order to reduce this amount of resources, you have to sacrifice the tree filling, for example, cut the number of possible sizings. That is why such features occur when some sizing is cut out to optimize the cost of calculation.

Another feature is that they do not use open limps on SB. Someone will say that this happens because the rake taken from the flop and it is disadvantageous to play hands with limps, but the same tendency is present in bots in MTT, where they do not take rake from the flop. Most likely this means that it is difficult for them to calculate limped pots on the flop with an SPR about 50.

The typical profile that these bots use looks like this:
1) High WWSF (48-49)
2) No open limps SB (<1)
3) CBet Flop IP below average (~ 59)
3) CBet Turn IP below average (~ 43)
4) CBet River IP above average (~ 60)
5) Reraise against a check-raise is practically absent (<3%)
6) BXB IP line is above average (~ 49)
7) Folds agains CBET on the flop OOP less than average (~ 34)
8) Fold on CBET on the river OOP more often than average (~ 64)
9) For a preflop caller, a check-raise without a position on the flop is below average (~ 8) and a raise in the position against cbet the flop is below the average too (~ 5).
10) Extremely high level of aggression against several checks (~ 64) bet on the turn per caller preflop in case the preflop aggressor did not bet either a flop or a delayed bet on the turn.
11) A large number of river overbets (~ 14) on average, but those bots that play more expensive (NL100-NL200) morbusmortis, lemmetakeit, Nozomi2716 no longer have such frequent overbets.
12) In multiway-pots there can be both a loose strategy or a very tight one. I put the nicknames in the table below from the most frequent CBET Flop in 4-way pots to the rarest.

A comparison of statistical indicators of the list:
https://docs.google.com/spreadsheets...8rOJXjbcjbcjbc

As in previous times, I assume that someone from the list may be in by accident, due to the dispersion and wiered choices of their strategy. But I think, that combination of the above factors is likely reveals a robot.

List of suspected AI users (April-June 2020):
morbusmortis
skuerpents
Sanzparil
Zukiaz
lemmetakeit
Stuffbluff
Cleverdore5
Zukzuk
Wheelle
Panda silver
shutoff
Strikemilvus
Rooorshor
Nozomi2716
skinnyNinj
Ceoperp
Djhankee
Mustar
Cingler
gutburglar
Gimira

Hand History of the list:
https://mega.nz/file/zhcWwYZb#3oVuNQ...At5Gt7J_fdA2IU

To summarize, I want to admit that it is easy to reveal automated bots nowadays, despite the fact that they are trying to disguise themselves. If you have suspicions of someone who might use a bot or a tips software, please write to me, I will study and it may be possible to reveal a new group.

In conclusion, I want to thank the guys from Pokerpopup.net for providing the "Analityc Package" popup that I used to identify a group of bots and Smarthand.pro for providing statistics for analysis.

Aside from possibly Stars and networks we're it's not possible to get hand histories, I think you could do this and would be able to find that this or other bot rings are present on the majority of most medium to large poker networks.

where*

I could check, but I dont have HH anywhere except 888 and WPN. If someone send to me, I will apply my filters to new database, its pretty easy now after all work done.