CBet ranges for bigger sizings, by any human, always seemed stronger, with overfolding being a good adjustment. But, if everyone knows this and overfold, then consciously playing a strat that overbluffs when using bigger sizings might be the most +EV?

*Especially if the pool makes the effort to overfold by default, and know optimal x/r frequencies go down as sizings increase?

I've had large cbet sizings for awhile as overbluffed especially on flops where a small bet is preferred in solverland.

I always thought it was just a group of bad regs, but makes sense they could be bots.

The difference in river play vs fish and regs is quite large in some spots so they're probably adjusting to VPIP or VPIP-PFR. I didn't see many river calls that seem terrible vs population which makes me to think they just have a set of static strategies.

In a lot of nodes they'll blast the worst combos or float garbage, but in others it makes rare unintuitive plays that don't seem based on a set of rules, but usually are solver approved.

Thank you for doing this. It's quite depressing.

I'm finally admitting to myself after 17 years that online poker is indeed a bad idea and that sucks bc live player pool trends toward having very bad breath generally

I was wondering the same thing. This idea and your next post about the pool overfolding to large sizings would both explain why the bots are playing the way they are.

Betting with large sizings and a weak range could also be a simple exploit based on HUD stats. Maybe the bots are programmed to only use this strategy against opponents whose stats suggest that they overfold?

Just because the bots overall stats come out this way doesn't mean they are playing this strategy against all player types.

Oof! Thank you for the explanation, it's kind of what I assumed. It's interesting that GG and ACR regs are still grinding/winning though, I wonder if the botting problem there is less than at ignition with anon pools? Ignition is an absolute cluster **** at the moment, I don't know of a single reg at $1k/$2k that is currently winning (without using an incredible amount of game selection). If the site isn't unbeatable it is pretty close to not being playable.

Does anyone know of any more history/information in regards to botfarm, who's running it etc? (someone mentioned Russian mafia in another thread but not sure if just hyperbole).

First mention here in the forums:
Corporation BotFarm - the biggest bot farm in the world?

Some company files uncovered:
Bot farm uncovered and exposed.

Old website:
Neo Poker Lab

New name and website:
Deeplay

I dont know if I should post this, but around 1 year ago I was very sure almost every "bot" on wpn was showing cards on river lol. They were operators, not bots ,95% sure. Since then some guys I know 100% are regs start showing cards, so I was not so sure. But I would be very suspicious about random regish accounts showing cards. And also I dont think they were amazing at 3way pots, but I played 80%+ hands at 1k so not so much 3way pots so idk.

Trying to explain the gist of this thread to my buddy who dumps small money on these sites is funny. He prefers heads up play and he almost always get torched, its no telling how many bots have been slapping him around. I knew online poker was marginally profitable without HUD/other technology in early 2000s and knew that online poker was never going to be the same when computer programming crowds found the profitability in the game.

I had only checked river, looks like they showed a fold 3 times so 1 in every few thousand hands which is way less than random players

6h6c hand was Nov 6, following hands were Nov 3 and Oct 16. Data span from Sept 2023 to Mar 2024 so it doesn't seem random

Don't think the problem is that big at all from my knowledge on ACR. I would say it's rare to even have 1 bot at the table at nl400+ on acr.

Where as on ignition its sort of debatable how many regs are actually human. I got my own cfp and I'm friendly with both a couple of American regs and some cfp owners and to my knowledge like 90% give or take stopped playing there. But I also know one guy that played it as his main game recently and you can still play plenty of volume if you want. Which to me is really absurd, it kinda puts into perspective how infested the network is with bots. If 90% of regs quit in your player pool your win rate should be through the roof and games would barely run. Where as clearly this ain't the case.

At least on Ignition — every single wild hand has another player at the table who folded relevant hole cards.

Ive played a bit with the ACR bot profile and I believe it plays similar to bots I’ve played on Chico (~2019). For me, it plays nothing like gto. I believe it’s designed to attack population tendencies. I can discuss w/anyone via pm but Idw post hands here.

TY

What do you mean it doesnt play anything like gto? I think its rare that the bots make obvious punts. I have looked at sims where people questioned their lines/sizings and it usually ends up being super small mistakes at equilibrium. Since it kinda is presumed to nodelock and solve strategies in real time automatically, I could see a lot of lines being justified.

Also the bots have developed a **** ton since 2019. I think you should be careful about sharing information about how to beat them fwiw (at least to accounts that are new/unknown/few posts). It's just a matter of time until it reach creators or enough humans exploit them and they will make a update that fixes the leak and they get harder to identify them.

The creators are 100% reading the forums which isn't that weird considering they are making millions consistently.

The bots across WPN and iggy have very clear imbalances that you won't see if you alienate one hand and punch it into a solver (different splits pre and post) so looking at one solve and seeing EV difference between plays assuming that one line they play is balanced is an inaccurate outlook.

They have 123 employees associated with them on linked in, doesn't mean they have 123 employees right now working for them, but wouldn't be surprised if the number is still pretty high, maybe 30-50ish.

They have a bunch of data scientist and engineers, dev and ml ops engineers. Even a scrum master to make the project management for the scamming of other people as smooth as possible, rofl.

From my humble opinion (I work as a data science consultant, so I have a bit of an idea what they are doing) poker is obviously kinda f***ed if you have to deal with industrialized bots, where a pretty big team of experts is working on them...

GG can barely detect superusers, meanwhile this deeplay company has own teams working on implementing ml algorithms to scam other people.

Realistically speaking you can't really fight this, as one would probably have to put up the same resources like they have.

Only the sites could combat this, but as we saw with gg, they don't care and or are complicit (like maybe acr, who knows).

Edit: well, found the old discussion from 2022 here on 2+2. According to their website they have 500 employees, lol.

They don't even hide it really:

having knowingly played a ton of volume against them this year i can confirm this firsthand, the splits are pretty radical on most nodes i've tracked

also their general disregard for blockers is quite interesting, thinking of the 75o cold 4bet that patrick howard posted. assuming that this has to do with some bunching effect when another bot(s) has folded relevant broadway cards in earlier positions, or could just be random frequency aggression to squeeze out the 3bettor

Just realized the "win" folder includes the Jan 9-16 losing period, a lot of the indecipherable hands were from then

Most of their leaks are by design and the developers would rather not have to fix them. Even if it's public they're only incentivized to change once a large % of regs adapt which I don't see happening.

The problem is you can't counter them without realization they're a bot, but some of the obvious signs they could patch without losing too much. And fish won't adapt so they'll continue to get brutalized, even more with every update.

I think the only real solution is for sites to ask winning players to film themselves playing or get banned. I don't see the operators getting around this without using identities of decent players, and even then the sites could detect them playing differently.

No it isn't a inaccurate outlook. They operate the same way humans do, they have their baseline which is kinda gto-ish. Then they add exploits on top of that very much like a human do. Which is far more dangerous, as of now I dont think their game compare to the levels of the best humans if we are talking 6max. But as they improve they will likely become better than guys like Linus and Davy. I've played vs bots at the highest stakes that run in 2024 and i've played vs oborra and the early 500z bots. The only reason I've never experienced them to be tougher than the best humans is because they haven't found the perfect dynamic/static strategy. Every bot have either been too gto / theoretical or too exploitative. It's only in HU poker or possibly reg wars without a fish that a bot would perform better than the best humans, for now.

Pulling off right frequencies like you guys are talking about is completely trivial unless your opponent starts punishing you for doing so.

Basically everything you wrote is wrong other than the fact that you’ve found solver justifications for some lines, I can’t really dispute that. But yeah not rly too interested in helping you making money so gonna put a lid on it

Lol, im in the group with tyler. I've talked a lot about the bots and their strategies, im not looking for your or some randoms help. To you or anyone else who think im making things up im happy to bet very big on it. Anyone who think they can beat 2k ignition for more than 5evbb for meaningful samples hit me up. Considering I know plenty of 8-10bb winners that stopped playing there it must be a piece of cake for 2p2 nvg.

Feel free to dm.

Hmm honestly - if I run a poker site or an app, and I had no ethics, wouldn't the best approach to simply work with the bot farm, take a cut of their winnings, and give them instructions on what games and volume (and even quality of play) I want them to run and when?

I could even cut my bot detection and leave it to THEM as part of our deal as they have the best expertise on how to detect other/unsanctioned bots.

The add has a point - this would be much easier and much more profitable and customisable than having regs who maximise profit at all times, I need to deal with their withdrawals, complaints, time zones etc to keep games running ? Talk about a win win!

sure, unless some regs figure it out, make it public and recs catch wind of it and decide to take their business elsewhere, seems unsustainable on the long run

Well you could lower your risk considerably by, let's say, just off the cuff thinking here - making sure you battle data mining, hand history sharing in stables as your main priority!

Where my fellow tinfoil hats at