This report exists solely thanks to datamaning and mass data analysys (MDA)

I have examined cash games on GGPoker over the past six months and conducted an analysis of bots based on the y2da report. I have identified 15 bots (11 inactive and 4 active).



Results:



What Identifies Them as Bots:

1) The combination of WWSF and WaSD is remarkable. These statistics are unusually high, which is challenging for humans to achieve. Currently, it's rare to find a regular player with this combination.



2) Their 58% WaSD suggests that their checking ranges out of position (OOP) on the river are strong enough to win at showdowns when opponents check behind, with an average fold rate of 47% against bets afterward. Most human players have higher fold rates.

Check OOP River Total (1x1 Pot)


3) Their construction of a C-betting range on the flop implies weaker ranges in larger bet sizings, leading to higher fold rates (around 50-55%) versus check-raises after C-bets between 45-90% of the pot.





4) In single raised pot (SRP) scenarios, their C-bet sizing in three-way pots does not fall below 40% when facing at least one opponent in position against them. However, this detail was omitted from their solution but may be included in future updates.

SRP Raiser Cbet 3-Way Pot vs OOP/OOP (SB+BB) - Small Betsizing exist


SRP Raiser Cbet 3-Way Pot vs IP/OOP (IP+SB/BB) - Small Betsizing does not exist


SRP Raiser Cbet 3-Way Pot vs IP/IP - Small Betsizing does not exist


**** happens that they forgot to write it into the solution, probably add it in the future.

5) Their winnings graph shows a minor downturn occurring between January 9-16.



During this period, they experienced significant losses and played many random hands, which is improbable for a skilled regular player with such a graph.



For a detailed mathematical proof, refer to the link provided.
https://mega.nz/file/Mj5gmJwB#KJigZv...EWZ1KDe0tXscyE

You can also access hands played by the bots to verify and develop methods for detecting them in the future.
https://mega.nz/file/YqAW2aDI#3dPlzT...UAxYz7r5VeLijY

I am sharing this work to illustrate the importance of openly disseminating data about the game and the advantages of employing specialists in poker big data analysis. While some consider poker player education schools detrimental to the gaming ecosystem, they also produce data scientists. Following my reports on bots in the Winning Poker Network (WPN), independent teams have emerged to detect and report them, thereby fostering fairer gameplay. I hope this trend continues, with more teams working to identify cheaters at the tables. I will continue conducting research and posting findings on suspicious cases on GGPoker to contribute to a more transparent and fair gaming environment.

There can be no fair game when crucial information is concealed.

thanks a lot
wonder what happened beetween 9 and 16

but why does it say pokerstars in the hh?

Nothing good. Means it's not stack sizes or misreading cards that is causing this. But it's bad strategy vs Fish raises (judging from examples).
Unless there are similar missplays vs regs, it means they have ways to id fish. And then they have exploit building algo that misfires because there's not enough hands like that. Or just poor mapping vs big reraises.
Another thing that's not good is that they went a week w/o noticing. So nobody is even observing this BS, it's straight up bots.
20bb/100 pre-rake exploitive bots. Fkn nice. Let's ban mining, so they can destroy in peace. Another episode of GG cares.

It's because of you cant import any format in h2n i think, it's common you need to convert them to a different format to import the hands.

You can import gg hands using the h2n website converter for that. With h2n4 you can import them just like any other sites hands.

Thanks for your work Tyler!

Two questions:

* If these 15 accounts have been identified as bots, does that mean all the other accounts in the sample you analyzed are confirmed not bots, at least of this type, or are these just the bots you have identified so far?

* Given the high winrates in games where that is uncommon, is it possible that these accounts are cheating in some other way? Do they play on the same table for example? I wonder is hole card sharing a possibility.

I played a lot with almost all of them.
Kazakstan, Russia and Peru are their origins.
They are all more aggressive than the average reg and use very big bets (based on my notes). Don't know if they are connected between them.

are they from the same botfarm as bots on ACR?

Why is it hard for human to achieve high WWSF and WaSD?

Great job, sir!

As someone building models for bot-detection for other sides this is highly appreciated.

Jesus christ, even bot magic made 89k but paid 120k rake

Well baseline assuming you only play hu pots would be to win 50% of the pots when you see the flop assuming both of you play same preflop ranges. Then it gets harder assuming you play multiway pots. If you just bet/raise a lot you will get high wwsf but worse wasd because you wont have the best hand very often if you reach a showdown. Most strong regulars end up being 48-52 depending on how much they coldcall/overcall. I don't think there is a single reg that has 54 wwsf on highstakes or even midstakes from what i know. The won at showdown is also extraordinary high. It's either the most advanced/best MDA/GTO-hybrid bot ever or some type of collusion

Botfarm is the name of the russian company that produces them.

it has to be collusion, these showdown and wwsf stats are insane

If they're colluding like on Ignition (trying to go 3-way) WWSF would go way down

For card sharing you'd often know of blockers that disincentivize bluffing, W$SD makes sense with better bluff catching decisions but I'd think 2 random cards would be a fairly small difference

Does anyone know if the bots on Bonition are also using large flop sizes?

In other bot-scenarios (PLO though) we saw the same... It was also because of extreme aggression to exploit population tendencies

After importing the hands (besides the losing period) I'm very confident collusion isn't a significant part of their strategy. Not gonna say how I think they might be programmed but they're extremely impressive

Filtering for multiway they don't tone down the aggression. Looking at a bunch of hands I didn't see one hand with 2 bots seeing the flop together multiway, rarely even being at the same table


Some other things I noticed:

- Looking at turn probes, the vast majority of the bets were in increments of 0.5bb. Bet turn after donk flop was almost always a 1bb increment

- SRP donk is 8.1%, never using the size most regs who donk use (1/3)

Most regs (and solver) would donk low boards a lot more


- Huge overbluff with very large sizes, this is all river bets. From the hands I looked at they don't respect blockers much if at all


- BadWolff folds to a river raise and then shows his hand, only showdown of a fold in the database


- Zero turn XRs with air or a draw

- Very low XR and XF in 3bps especially vs 1/3, lots of floats followed by a turn donk


- Pre stats are 25.5/19.5/12 when 6 players are at the table

the J8dd is wild, how the hell

Yeah I don't get it at all, maybe river is +EV but it can't be enough to make up for the turn call

Found some more fun ones looking at multiway (where they're winning at 65bb/100)

I don't know how deliberate these plays are versus being a simple extrapolation from their heads-up strategy

Has anyone who has studied these hands actually found a solid counter strategy? (not asking for it to be posted here), they feel pretty unbeatable on ignition.

They stopped doing this towards the beginning of Feb, no more of those crazy collusion hands so I assume if they are sharing hole cards, it's just part of card removal.

They were crushing a tough known pool seemingly without colluding, I can only imagine how much $ they've taken off Ignition. I assume the same or similar profiles are also running on ACR.

From what I understand the bots are running some type of program similair to what they said wolfeating/2btc/stongpanyid guy did. So it's a superquick solver that calculates spots in real time but also do some type of profiling/explo profile. I would assume they are kinda one step ahead at all times, so even if you find a good counter-strategy it will never last long term.

The stats makes me kinda curious, I wonder how hard it would be to replicate those stats. Untill this thread by Tyler I have always been very confident a strong human reg on his main stake would always beat a strong bot. These bots have to be the biggest winners on the stakes they played, which historically have never been the case.

It's very interesting, are they using just data like MDA to exploit general pool tendencies by doing same stuff over and over to all players or it's more advanced and their actions change user to user to specifically exploit tendencies of a single player

Anything we can actually learn from these bots and can, as humans, implement to gain EV? Don't have the MDA to back it up, but always heard pool overstab when checked into in pretty much any node, so having a range that does not overfold OOP OTR seems good (per item 2 in the OP).