can bots beat players in 'god mode'?

Anyone with programming skills sufficient to write a bot capable of playing something more complex than HU NLHE (or LHE) would just sell an auxiliary tool that circumvented calls to the camera and played canned video of a dude playing poker on his computer for as long as necessary. Or the bot would stop playing and show a dude masturbating for ten minutes.

If a killing bot would exist the coder won't sell it off course... He would use it to make millions

Watson crushed Ken Jennings at Jeopardy

I hope top players aren't foolish enough to play for free against these programs.

unless theres only one human playing

thanks for letting us know that you are aware of the fact that poker bots could easily learn from playing top players.

The bot in the article doesn't actually learn from humans. It was designed to figure the game out based solely on the rules.

This is bad news for online HU going forward. Even if sites can police 24/7 automated bots, they wont be able to stop people from using a bot to coach them into playing perfect HU poker.

How can anyone still be so clueless? Best limit hu bots are already better than best human players. No-limit HU is obviously far more complex but it will solved, at least to a point where no human can beat it. It doesn't need to play perfect GTO because best humans are so far off themselves. It's just a question of when. I give it 10-15 years.

nl holdem is usually regarded either 1st or 2nd (after plo) hardest game to solve. I saw you often keep insisting that stud games are actually hardest to master, yet any mixed games crusher always says that they are the easiest - plus, they are the only game from the mix that literally never runs online at nosebleed stakes (even nl 2-7 triple draw had its time). I think I'm gonna trust guys' like Thuritz or Kostritsyn judgement on this matter.

As for multiway pots, I don't really have any decent data, but it seems pretty obvious that they are far easier to play well due to much narrower ranges on later streets. Narrower ranges = less possible game trees (remember that number of players usually reduces during the hand, and yet ranges are drasticly narrower all the way), which makes a game easier to solve.

There's a reason that all of the biggest successes wrt poker bots has come in HU games and it's not because the makers prefer the challenge of a harder problem to solve, it's because multiway is substantially more complex than HU play. If you don't believe me, try drawing out a decision tree for a hand that involves 2 players and then do the same thing for 3 players.

no

once again, "remember that number of players usually reduces during the hand, and yet ranges are drasticly narrower all the way" - to be honest, I'm not sure about that, but I feel that increase in game-tree options due to increased number of players would be more than evened out by very narrow ranges.

Let's try considering a fictional multi-way poker game, for example (to also satisfy Sklansky's premise), a 7-card stud-high game, where every player is forced to get to 5th street, with all previous bets in the pot. Now, in that game, a correct strategy would be ridiculously tight, and also very simple - you only play the very top of your range and, of course, all the draws. Since pretty much every made hands vs made hand would be a cooler, there would be no real possibility of gaining edge here.

The only, and very unlikely, situation where bluffing would be possible would be when your made hand would fail to improve, yet you could represent a draw that got there with your board. So, the decision tree in this game would be theoretically bigger because of the number of players, but almost all of it would be trivial folds or trivial calls. Seems simplier to me.

Of course, you may say that in this game you would gain edges by making extremely thin folds/bets/whatever - but that would prove my point, since it means that achievable edges in said game are much smaller, and (once again, just a feel here, im not an expert) games with smaller edges seem to be simplier to solve.

edit: after re-reading my post, I realise that I've made a few wrong assumptions and deductions in it, however I'm gonna leave it in unchanged form to see if my intuitive reasoning has some merit

A fixed strategy programmed to exploit the leaks of the average expected opponent would be non-adjusting and non-GTO.

Juk

You're considering a 5th street situation in isolation without considering the vast number of ways that you could possibly have got there.

yeah, but in "normal" poker games there are rarely multiway pots, as early-streets correct strategy is usually pretty tight, that's why i had to make up a new game to satisfy Sklansky's idea (multiway-friendly game). Other ideas might be any limit game with gigantic antes, but i think that same reasoning applies.

Btw, I might not have been clear about in my previous post, but I'm trying to consider how i imagine GTO strategies in said scenarios.

couldn't afford to buy my m3, had to lease. poker confirmed dead

found a sample of hands from acpc and imported them



all the hands are 200bb effective

0-200bbs pot



>200bb pots


This is a small sample of hands from the logs, I found on another website here http://www.unfoldpoker.com/2014/07/

Anyway I'm more than sure that, when poker will be solved, only one player would know that... There wouldn't be any incentive to let this thing going public

URL doesn't work for me, anyone know to which URL you're being redirected?

oh hai thread,

I think I can solve this "online poker is dead" problem. Gimme a few months or a year or two.

BRB

Totally different type of computing problem. Watson's primary skill is parsing a language to determine what the (in the case of "Jeopardy!") answer means. This is a problem with a single correct solution. Then Watson needs to determine the correct matching data to formulate the question. Also a finite solution. The correct responses need not take into consideration what the other players are doing, or what their responses might be; Watson just needs to make the fastest response. No element of chance in" Jeopardy!"

Sweet, with this kind of computing power he should be able to get very close to GTO with finely grained betsizes in NLHE several times over and then find cure for cancer. That being said he could surely spare a core hour or two and tell us how far away is his bot from GTO (assuming only few betsizes or something). It's fast to compute, it's necessary test anyway so he has the code somewhere if he knows what he is doing. It's also very specific measure which would be good estimation how good his code/bot is.

No man I'm not exagerating at all. I tried to play 500z today but it was impossibru. Too many nuclear weapon bots around. They took my whole ****ing bankroll, along with the bankrolls of at least 7-8 other 500z regs. I'm telling you guys this is it, it's over.

People say bots this bots that but this is the real bot. It's here, and it's eating regs alive. one by ****ing one.

I'm pretty sure you posted on snowie's thread about the same idea... and seems you know what you are talking about.
However don't you mind explain how can you compute that to a noob like me?