Hello everyone, after the topic of bots on 888, I was asked to look at the situation in the Cash Games on WPN and I found another network of bots there.

There is a common belief that bots are inelastic for sizing and can be exploited through 1bb bets. My research reveals the opposite.

This is Human folds to Cbet out of position, after defending the blinds against one opponent, depending on the size of the bet they faced:



And this is Robot folds to Cbet out of position, after defending the blinds against one opponent, depending on the size of the bet they faced:



Whatever lines I take, bots in HU banks everywhere are more sensitive to different sizings than people.

But the situation changes completely if we start to watch multyway pots.

This is Human folds to Cbet in multiway pot:



And this Robot folds to Cbet in multyway pot:


See this inelasticity between reaction to 33% and 50% bets ?

Let's see one more example.

Here is a Humans check-fold in multyway pot as preflop aggressor:



Here is a Robots check-fold in multyway pot as preflop aggressor:


Again inelasticity between reaction to 33% and 50% bets

Now let's look at the general frequency of bets in multyway pots.

Humans:



Robots:



Amazingly, the more people are in the bank, the more often this group of bots makes bets. It looks as if they don’t have a solution for multyways and they act there according to a simplified strategy that reveals such nonsense.

Additional patterns that this group has:

1) Beast WWSF (~50% avg)
2) A high level of aggression on the postflop (Cbet river, BXB, bets vs missed cbet, strong red line)
3) Less raises (low flop check-raises, low raises against check-raise).
4) Lack of open limps on SB.
5) High folds at 3rd barrel on the river.

The final list of nicknames of bots that play identically according to the above patterns (NL50-NL1000):

CheckRiverPls
HiMomIChampion
alakrity
w0rdvirus
daggerrunner
Unassailable1
LaLaRa2211
dijkstra48
lastresistance
drunkenrage11
wielded
bobo1923
hallowedbyName
chiefbumster
Tsunami8000
AcerbateSensei
vienalgaman
golddublon
Bayaz
LordAzriel5
RumbikBubik
shootist

Total Graph (All hands played 2020, NL50-NL400 mostly):


One of the problems that I found is that if I lower the requirements for searching, then I find players playing in a similar style but differ literally by a couple of stats from the above group. I have three versions of why this happens:

1) Especially make game styles of nicknames different from each other
2) Apply a new decision-making algorithm that eliminates the inefficiencies of the earlier version
3) Part of the time these nicknames are played by real people

For example, next nicknames are very similar to the group above, but they do not coincide in some stats, I suspect them, but I can’t prove:
purus
Sharty waffles
Alcejc
Apocalypse
stongpanyid

A full analysis of differ players by stats, I did in the sheet in Google-docs:
https://docs.google.com/spreadsheets...#gid=746507886

The hands of the players from the list above so that you can verify their similarity:
https://mega.nz/file/n8lSUABR#nIcpjy...Ur8qTNo9z-zu2M

In conclusion, I want to make a small statement for those who will deal with their identification in the future. Less and less sense today to look at the intersection of players by IP and general stats like VPIP / PFR / 3BET, but it is important to look at deeper things. The distribution of the strategy along certain lines, distribute sizing, strategy distribution according to the texture of the board, timings in decision-making. The work that professional players have done in the past should now be done by security. Too precise solutions are light up. Too much randomness are light up too. Open-source to statistics is very important today. Tools for in-depth analysis of strategies are already available now - Hand2Note allows to explore any stats that are possible, and the number of stats is limited only by imagination. I am more than confident that the community itself can do this work if gets open access to Hand Histories. Most likely, bots of varying difficulty are currently playing all poker disciplines, especially at low and medium limits. Without statistics analysis and a strong poker room security service, which is able to identify them at an early stage, it is most likely impossible to resist them.

At least one in your second group is not a bot. Nice work on the top tier.

stongpanyid? He had profile similar like top tier, but fixed problems in multyway. Basically any of 2nd group have differences are not similar to the main group, so I'm not sure about them, but left them for future research.

No. I don't want to mention/point out, but he is a member here and part of our group. Not a high volume player, asks questions, etc....

Appreciate the energy you put into the work. Hope all the sites give your stuff a solid look.

Great work! Thank you for all of this, much appreciated!

But I have a question.

If I know a player is a bot or suspect it is a bot, is it against WPN rules to play the bot and try to exploit it as much as I can? Please let me know.

Thanks Alot

good job by op
here is a group of russian bots on wpn network playing heads up sit&goes from 24$ up to 500$. Here are their nikname :
spinnerwinner

autisticraven

bettercatch

G20LLD

gerda_kitty

More info here:
https://forumserver.twoplustwo.com/1...-sngs-1742797/

Thank you very much for the work you're doing. Hopefully, the sites can step up and fight this better.

One of the stats which surprised me was that they have a very high W$SD while maintaining a high W$SF. This combined with high folds in multiway pots might be indicative of some degree of collusion? Do you have any thoughts on this?

Actually im an MTT player, so i dont know how cash stats should look like. But combination of overfold to barreling and mandatory bluffs with bottom of the range and big sizings could lead to high W$SD and too agressive strategy in multyway could lead to beast WWSF. Also keep in mind that W$SD and WTSD and WWSF stats in Hand2Note (which i use) is calculated differently to HM2/PT4, they exclude preflop allins.

The fact how consistently those things run above EV is somewhat telling too.

I received reports that these players also often sit at the table with each other.

This is their Preflop All-in Graph (Big Blinds):



I suspect that they also share cards to each other.

Or on a single computer...

Sending you some info

Thanks for your hard work on this subject. This bot group has been known to regulars for a long time, but WPN has done little to directly suspend/ban them and confiscate their funds.

I'm not sure if you play PLO, but a similar analysis on the PLO field will yield similar results. I.e. their is a group of players (bot network) with very atypical stats (esp. high cbet % in multiway pots) to the rest of the population. These bots have nearly identical stats to the bot network that was banned on Pokerstars some years ago.

Like your analysis of the NL bots, the PLO bots are also running way above ev for all-in preflop hands. I also suspect they're cheating beyond just running a bot script/program, with card sharing being the easiest/most likely option.

If you're going to cheat by running a bot program, why not cheat further and gain a HUGE advantage by sharing cards. I see no reason why they would stop at just running a bot program, when the incentive for additional cheating is millions of dollars.

Fyi, most of the suspected bots in my sample are all crushing the games. I don't have a lot of my old hands/databases (bought new comp, threw away old one), but they were crushing in those samples too.

As an aside, not to sound too much like a conspiracy theorist, but I believe some of the bots are programmed to lose. The losing bots' job, when seated at a tables with 2-3 winning bots and 1-2 regulars/fish, is to enter a lot of pots, aggressively 3/4 bet and deny equity from the 1-2 players at the table not in the network.

I'm nowhere near 100% on my last theory like I am that the bot networks exists and is crushing, but having played millions of hands of PLO on the site, I do suspect this to be the case.

Amazing work OP
Everything you said also applies to the NL500-nl1k bots on Chico/Betonline, so they must be part of the same botring
They also massively run above EV, so it seems like they really are sharing holecards

Haven't read it all but if a site isn't catching actual collusion and claims to have any kind of game integrity team it's surely a huge disgrace, this all in graph is mental.

Also stongpanyid also did play on Chico/Betonline under "bloodywolf" and after getting banned tried to come back with a few different nicks, but always got banned again pretty quickly.
He is not a bot, but is using a realtime solver, 100%
Is he still playing on WPN?

Also I once played with 3 bots on the table and they all disconnected at the same time, which pretty much confirms that they are all run by a single person/ring

The same botring is present on IPoker and Bet365 Premium Tables, too.

Good work OP!

The same botring also operates on nearly every app and operated on global poker in late 2017 until early 2018. I and a handful of others players gathered enough information to get the accounts banned. There were at least 23 accounts on global and they won at least $250,000 off of the site.

At the time the only method to get money on and off the site was through paypal. Max (maximbc here) was the one facilitating the withdrawals. For those who don't know, Max also worked directly with / for WorldPokerDeals as an agent for Chinese apps. Given that he faciliated the bot accounts on global I would say it's all but certain that WorldPokerdeals was running bots on apps since Max was the WorldPokerDeals agent that handled all app deals.

I believe that this botring is the one mentioned here:

https://forumserver.twoplustwo.com/2...tfarm-1748840/

Also this: https://forumserver.twoplustwo.com/s...postcount=1757

Without looking at pt4 ive played against at least 12 of those players.

If true, WPN owes me some money imo.

Lol I’m in your second group and I’m not a bot. Still good points and and ideas overall though. I wouldn’t be surprised if the top tier had some bots

Lol, wondered if you would see this.

Great work OP. You clearly know your stuff. What are your thoughts on similar groups existing on party and stars? Does it exist across all game types? Is security too strong there for AI to function on a large scale? If it does exist, is it possible that there are greater variations between 'bots' to avoid detection?

Threads like this should be the headline news in my opinion.

I have pretty much everyone from the first group listed on my own computer as bots(uspects). I've posted about lalara and bobo before fe.
The composition of the second group is weird, one matches the first group, one is pretty much a degenerate, then there's stongpan and 2 I've never even seen.

Can pretty much for certain say that the first group should include drtyprist, mateuscornelus, estwood930, concrete3cao, chilipeppers0, urdespot, cormoran49, hankmoody84, jurassicbear56, cookiedope8 (see a pattern even in their nicknames lol?) and a couple of others I don't want to name without further investigation.

Thanks, i add them to the sheet.
They fixed problems with multyway, but in general line they are the same, to be honest looks like v2.0.



chilipeppers0 - dont seem like a bot for me, general line is different, he is more tight than this list and also i have some secret filters which i did not disscuss in the first post and he is not pass them.

I dont have data on Party and Stars. I think Stars is most clean right now, because they spend a lot on security (50mil per year?). I think its two options now:
1. Spent on security like Stars.
2. Make Hand-History opensource and free for everyone and let the community to fight scammers.

Unfortunately, some poker rooms like Party (or GG) simply ban the hand histories and pretend that nothing happens. And we can’t even verify anything.

The more time goes, the harder it will be to detect "robots" and requirements for security check will grow anyway. I don’t think it’s possible to win in the end, but we can greatly complicate the life of scammers by shifting some of the tasks to the community. This is my message - show that community still matter.

ok a stupid question
I play fair bit of blitz on wpn (50 and 200) and I don't have most of them (meaning I checked 6 random names and don't recognize a single one) in my db.

Assuming this particular botring does not play blitz just confirms how heavily they lean on collusion/card sharing, right?