Haven't played in a while but after reading some bot discussion here about bots on IPoker i was wondering how true this is and if the bots are used for Husngs?

I was going to deposit with IPoker and play the $10 HUSNGS but now not sure after reading all this bot discussion, I cant play on poker stars atm btw.

I think HU is a terrible format for bots

Can you link to the discussion that worries you?

HUSNG bots haven't been an overly concerning topic in the past, though unmarked bots are always a concern for the majority of human players.

I seen some discussion in the HU SNG REGS THREAD and searched Ipoker bots and read some more there, but wasn't sure if it was full ring or Hu

iPoker thing was FR I believe and pretty long ago (I think when Playtech bought PokerStrategy a lot of the botting was cracked down).

The recent Regs thread discussion was simply a poster feeling that 3 players on a US facing network (not iPoker) were potentially bots.

The important thing is to keep in mind allegations from confirmations. It's far from certain the three guys the person in the Regs thread was talking about a day or two are actually bots.

There are definitely bots out there, but most of them are only decent enough to breakeven overall before rakeback, and usually only beat losing players. They should still be reported, but it's not a situation where the opponent that is amazing and crushing you at $10s is a bot (a bot that was the best $10 player would've been playing higher stakes awhile ago). It's not yet that sophisticated by most evidence we've seen.

And do keep in mind that probably over 90% of the time people publicly talk about bots, the situation is actually a human and not a bot.

actually researchers from uni of alberta made a pretty sick bot
https://www.youtube.com/watch?v=ScXX2bndGJc

quite the oposite. GTO strategy for heads up is the easiest one, especially 25bb deep. hyper husng is the best format for bots.

even if this is true, GTO strategies are nowhere near from being solved.

Bot are nothing to worry about at this time for several reasons:
They are extremely difficult to program. It would be pretty rare to find a individual with all the required skills. Also, Someone with these required skills would do much, much better working in an other field.

They would be very unreliable. Just think of the number of time your connection mucks up, you lose connection to a server and other glitch...

Finally, they would be quite easy to spot and poker site have securities in place to prevent anyone from even trying. Even if someone was jumping all these hurdles, the bot could only play low stakes (may be low mid stakes too) as regs would easily spot them at higher stakes.

As for the few successful bot created by universities, like the one mention above, at the end Humans always end up beating them quite easily. There were some bots created that performed even better than the one mentioned above and there were also quickly beaten and ridiculed by good HU players.

It's also worth noticing that the few successful one were cash game bots. Tourney bots are even more difficult to program.
IMO this is not something to worry too much about.

- I agree that hypers are a good format for GTO-based bots, but solving poker is at least somewhat challenging, and I imagine most bots out there are simpler rule-based things. A simpler format, like full ring, is probably a better choice for these.

- I disagree with the sentiment that bots that only beat losing players at the $11s or whatever aren't a big deal. They could easily take a lot of money out of the poker economy that would otherwise be won by winning players and/or trickle up to higher stakes.

These bots play their standart play and dont pick up any reads or adapt (atleast the ones i know). I think this is terrible in Heasd Up.
Micro Full Ring Games where they can play their ABC game without too many people noticing it is the best format for them.

its true that the game is nowhere near to gto
note that it shudnt be the goal tho, its all about adapting
botwise the toughest challanges are in computional field atm

afaik a key to beat that bot was to balance ranges and mix strategies. guess that might apply to almost all opponent modeling type bots

I think your first point is very valid, but dont really agree with the last points i think.

I dunno how hard it is to program stuff but i imagine if a very good poker player would partner with a very good programmer they could create a very dangerous bot. It might take too much time/skills too program, but im thinking of some kind of bot that can play a combination of near-gto strategies, and bayesian exploitation strategies. Imagine you have a program that you could tell to play any strategy that you like, and the programm would execute it perfectly. Furthermore the program would be able to learn himself to play better by scouting his own database to come up with max exploitation strategys against the population, or against specific players. You could learn it to play very balanced against new players but perfectly adjust his game based on Hud statistics when more data comes available, you could learn it to take gameflow effects or timing tells into account. Of course you could program it to switch between alternating strategies at random time intervals, or after x games. You could program it to play a different strategy vs any player that you have some info on. You dont want people to know you are a bot so you make all betsizings and timings as random as a human would make them. You can even program it to make mislicks every once in a while, to decline faster after a loosing streak etc etc, You could train it for thousands of games on low stakes and keep tweaking and improving it untill you have found all the right balances and have created an unbeatable moneymaking machine, and nobody would know you are a bot..

Of course it would take a ton of time and skills to program all this but fundamentally i think the human thought process in poker is not something that a computer wouldnt be able to do much better. It wouldnt suprise me if bots will become a serious problem at some point in time in the future.

The university bots are easy to exploit because they just play one fixed strategy at a time and afaik they dont use expert knowledge. Also when you know you are playing a bot its much easier to test out a lot of random stuff

As for tourney bots vs cash bots, wouldnt full ring tourney bots be easier to make money with than full ring cash bots? I imagine its very easy to program a computer to make perfect preflop ICM ( or any other tourney model) descisions, and that alone would give u a big edge if the rest of the program is solid enough.

How anyone can say bots are not a huge concern is beyond me. I remeber a thread where bots beat 6max cash games if i can remember correctly for hundreds of thousands of dollars... if they can do that in a deep stack format what do you think these things can do in a short stack format...

Totally agree

It was just one month ago, where ive reported one guy at 60s hypers hu, because i thought it was a bot.

-> here the message of pokerstars

Its going to be a big problem in the near future. Keep an open eye and start reporting guys, which look suspicious. Ive detected lot of more bots outside of pokerstars too (up to 100s ) So there is def a BIG problem out there with bots. So pls dont spread out msgs like "no worries about it", because its just not true and will make ppl feel too safe (talking especially about lowstakes guys, who might feel there are some bots and just dont report any1 because msgs like those genher etc posted)

the bot you reported couldve easily been a human ******, too. if it didnt fail so hard in some obvious spots, which made it really easy to detect him fast. i think it only took a few days until he was banned, people noticed really quickly

sorry, but dont get your point?

i agree with you that its going to be a bigger problem in the future, if they tweak the obvious ******ed stuff

Everything HU up to around 15bb is pretty much provably solved via nash equilibria, I dont think 20-30bb is too far behind it.

Granted this is just aiming for unexploitable play, not a bot thats a world beater. But even it was a perfect player, how big of an edge could any person or bot have 30bbs deep unless their opponent is just awful?

Tournament poker is a toy version of the real game. Of course its going to be solved first.

the nash equilibrium you are thinking about, is a solution for primitive game of tree size of 5 nodes. the true full scale poker 15bb deep has a millions of nodes, thus it's nowhere close to be solved in a sense of game theory

I've solved games with millions of nodes, but the real number is much larger.

Anyhow, I agree that the game isn't truly solved above very short stacks (certainly not at 15bb).

But that kind of misses the point -- a bot needn't be anywhere near perfect to take a lot of money out of the poker economy.

Pretty sure poker below 20bb has lower game complexity than checkers, which was solved in 2007.

Yep. If I remember correctly, people have completely solved HU up to something like 2bb deep. As stacks get deeper, the game tree also gets bigger really fast and becomes very hard to solve using today's technology.

even if you are able to solve millions node games, thousands of hands and runouts, there is a 2nd problem, opponent modeling. for limit texas holdem there is ~10^18 parameters that has to be learned (according to Southey work). no limit poker has far more complex betting rounds, so guess the numbers are way higher.

consider learning all that parameters baesian way. what opponents are you able to play for more than 1k hands?

ye, of course.

opponent modelling? what exactly do you mean by "solve"?

by solving actually i mean both GTO (probably like continuum strategies) and develop procedure to find out the best response for given unbalanced strategy.
the first problem is only computational, and the 2nd one requires a way to model opponent strategy. there are few ways, but none is fast enough to be effective short term (to play a human)
i'm not an expert, just made a small research