Использование ИИ на 888 Poker

]

Today I have found a group of players, that uses similar/identical AI (artificial intelligence) for decision making in schedule tournaments on 888 poker site. While recently 888 has banned many players with “Using AI” ban reason, I want to help 888 and clean tournaments from cheaters.
Common alias diagram for the group look like that:

All hands were played in 2020.

Common signs of using AI for the group of players:

1) All players have big win rate BB/100 and strong red line for not showdown win, this fact is not accusation, but it shows similarity in the group and connections between players in the group.

2) Big percentage WWSF, that obtained by playing in combination of GTO-style and using some good adjustments to the field of all players in tournaments.

3) Cbet Flop in position, around 75%, that close to solver’s suggestions.

4) Big Cbet River in position, around 66%, they bet often using Big sizing.

5) Often, they use line Bet Flop(B), Check Turn (X), Bet River(B). Overall frequency (BXB as Raiser IP 62%. This strategy is rarely used in tourneys, beside players in my poker school, only few people using that play.

6) The group standard Probe Bet River is around 43%. How can you find and use strategy for high frequency BXB strategy and do not use High Probe bet River frequency for exploit strategy? For me it is a mystery, because to bluff with a Probe Bet River with a weak hand in average is more profitable than BXB strategy? How 44 Topregs (Very good players) have found to use light BXB and have not found to use light Probe Bet River, that suspicious. Programmer who has coded the brain for AI, should get some punishment for that.

7) All of them do not use limps from SB. At this moment of time, SB open raise preflop 57%, Open-limp 13% and 30% open fold. My version of this, because Limped pots are complicated to solve because of deep stacks, and the programmer has decided to forfeit the calculations. On other hand, probably they have not calculated limps yet and basic calculations was without limps and limps appears when they see some trigger.

8) Big Blind donk, after BB defense the use Flop Donk Bet(lead bet) 5% of times. Usually human players try to simplify the strategy of play and do not use donk bets, because in human world use these frequencies and at same time to be balanced is a extremely difficult task. But AI easily may use solver decision, that recommends to start flop play with a donk bet with 5% frequency.

9) Because some part of strong hand goes to the Donk Bet on the Flop, Check-Fold on a flop is slightly big, around 48%. That frequency is not common for players who like to fight for pots.

10) Low Check-Raise on the flop 10%, which is too low for active players with strong red line.

11) Another thing is overfold to a third barrel (continuation bet River) against preflop raiser INP, they fold 72%, that makes sense, because players on 888 site bluff not so often than they should. This is a good adjustment, that most people in the group have, but data sample is not big for every player in the group.

12) Low frequency 3bet BB vs SB in deep stacks. Frequency is 8% for all players in the group, that is lower than 3bet BB vs BU (11%). I do not understand, why that strategy is implicated.

13) Very good feature that show similarity is bet with missed cbet from preflop raiser. In HU pot, bet against missed cbet is 56% in average, in 3way pots in same line of play, bet vs check from preflop raiser and blind check (SB or BB) is increased, 67% in average

14) Big Check flop (X), check turn ( X), Bet River (B) strategy 52%. That is also rare strategy as BXB.

15) Big cbet flop OOP without position (64%), this is not a default strategy for regular players in nowadays. But at the same time, Check-Fold on the flop while we are raiser OOP without position is (42%)

After I have found common similarities in that group of players, I have entered expression below in Hand2Note in specified format and got the list of players which have these stats.

Screennames of players:

PaperBush
Lonelystrang
StrangeRouge
Quadrick
Halui
AllCash90
Rulle
Drozd676
MistrJackpot
Countrylove
SonOfWinter
ZinZiber
Pro100Bill
AnneB0nny
Malygos1987
CunningGreen
mr.Robbot
shadowmoon3
ropecutter
stopANDg0
Celebrimbor3
araikvip
Hard1ine
Desired
YupYupMan
forsaknbyall
Fokerok
Discharger4
Acefighter86
Gueekin
flashTrash
Helilga
Gidrilla
notionalist
Bloodyreveng
bullmad
AkaneTsunemo
Fourniti
neopet4
Truemaster89
GunsandGret
Timbio
SloanWall
Yasal

I have uploaded all statistic on a file server, you can download it and use in in Hand2Note and see how similar these players are.
https://mega.nz/file/mxsBHA5S#PrljySTXIBlsxDqtwyWQILH7uaXJV9XA1hegUmhAGfg

I am afraid that the list of players is not full, that list I have got after 3 iteration in the Range Research in combination of different entry conditions. The problem is that to get new screen name is much harder and requires much more resources than to get first 10 screennames.
Because It was a shot in the dark, I believe that for some players it is a mistake and they are not part of that gang, but like these players on randomly played distance. Especially some players’ static has slightly difference: red line drops faster than usual, check-raise flop is 15%, 3bet BB vs SB 15% etc. All these differences is easily explained by variance, but on the other hand player might be not involved in this gang. That is why I created a google table, where I showed key patters (statistics), where you can see what and how is different for people.
https://docs.google.com/spreadsheets/d/1dI7xnyK6V7bzEXwNSroC89TKOlEluJqBI0bAwhwFXxY/edit?usp=sharing

It is very sad that many poker sites started to forbidden tracker programs. From information that some user shared on one of the Russian language poker community forums gipsyteam.ru. Bot creators using screenshots of hand history to convert them in text format in converter, to understand how people play against them and how to set AI. On other hand we (poker community) can not fight back, because hand history and tracker programs are forbidden.