Is there a perfect NLHE Bot? Could there be?

[Sushiglutton]

Yes it is possible to program a perfect NLHEM-bot. It´s a result from game theory that there exist a perfect randomized strategy for any game like poker. For every possible situation the bot should choose a random action with different probability. For example with KQ, an ace flops and there is a big bet the correct strategy might be something like: Fold(75%) Call(15%) RaiseX$(1%) RaiseY$2.3%)... (given all other data like position, potsize, stacksizes etc). But one thing this bot doesn´t care about is psychology (playerstyle etc). Therefore a human would still do much better against poor players. This bot would beat ALL players but it wouldn´t win maximum.

However NL is such an extremely complicated game that no one has been able to determine the perfect strategy (not even close). But it has been proven that it exists (theoretically).

To summarize: Yes there exist a "perfect strategy" that for every possible situation (cards, number of players, stacksizes, previous actions in the hand etc (but NOT based on players actions in previous hands!) ) give a probability to every possible action and then choose one at random.

(please forgive my poor english. foreigner...)

[sjb]

There's one point that I think some folks miss here. Hold 'Em is (discouting the rake) a zero sum game. The only asymmetry in the game is position, so the "value" of the game, as determined by game theory can only differ by position, and the sum of the values across all positions must be zero.

Therefore, the "optimal" stragety identified by a Nash equilibrium for the game really only tells you how to to break even. So, if you're going to build a poker bot that *wins*, you'd have to make it so it can recognize when other players deviate from one of the equilibrium strategies and then determine how to best capitalize on that deviation - that's where the money would come from. Game theory doesn't tell you how to do that, because one of the fundamental assumptions in game theory is that all players are equally able to analyze the game to determine their optimal strategies.

So, to build a "perfect poker bot", you'd have to go way beyond a game-theoretic optimal strategy (Nash equilibrium) to work out a model of how the other players at the table *actually* play. That problem is *way* underdetermined. Hence, I'd argue that no such bot is possible.

[Borgland]

[ QUOTE ]
[ QUOTE ]
No computer can comprehend human emotions. You can program the correct math but not the emotional variable.

[/ QUOTE ]

Are you sure?

[/ QUOTE ]

Yes, because there is a difference between "on going research" and a finished product.

Aaron Brown mentioned a few different examples of bio feedback bots in another thread, but the best example was 80% accuracy. We're talking about the perfect bot. And those werent taking into account outside factors other than the poker game itself.

Some examples:

While I'm in a hand Meg Tilly sits next to me wearing a low cut dress that showcases her magnificent breasts. She smiles at me and says "Hi". Now obviously I find her hot so my heartrate increases, breathing speeds up, my body temparature rises, and I start sweating.

Or I owe some guy a lot of money. I'm in a hand and he walks up and demands his money or he'll hurt me. I dont have it so my heartrate increases, breathing speeds up, my body temparature rises, and I start sweating.

I'm bluffing to a huge pot against a flush draw and overcards. My heartrate increases, breathing speeds up, my body temparature rises, and I start sweating.

And playing online there could be literally dozens of reasons that heartrate increases, breathing speeds up, body temparature rises, and players start sweating that noone but the player would know why because noone can see him.


How could a bot or the fanny pack Mr.Brown mentioned tell the difference of what's causing this? My arguement is you cannot program a bot to tell the difference because it cannot comprehend human emotions. It would have to have human emotions itself and so far that only happens in Star Trek (Commander Data).

[quote="Snoggins"] The "Deus Ex Machina" post ruled.

[/ QUOTE ] I thought so. [img]/images/graemlins/tongue.gif[/img]

[donkeyradish]

Poker is not chess. There is no way to determine what "perfect play" means unless you could second-guess the thought process of all your opponents. In order to do that you'd need a vast amount of data in advance (more than can be collected in a reasonable time).

For every strategy there is a counter strategy. I think any poker-playing robot could always be defeated by a "better one" that exploits its predictable behaviour.

[b33nz]

Pardon me, but if there would be a bot that always makes the correct mathematical move, then...

Would the bot ever check-raise on a bluff?
Would the bot ever raise on a bluff?
Would the bot ever continue betting on the turn on a bluff after its bet got called on the flop?

If you said no to any of these questions... then a perfect bot is not possible. And technically, the mathematical move is to never check-raise bluff if I'm not mistaken. This move requires art, not mathematics. It's what you know about the player and his tendencies.
So the bot can be programmed to check-raise bluff either randomly or everytime against very agressive players. Pros, even halfway decent players, would pick up on this and only bet with good hands, or reraise bluff the bot on a semi-bluff.

Think about it, if the bot is raising on a bluff every hand... that can't be the correct play. If it is raising 20% on a bluff... it is making his moves random, which is also dumb because then you can make a continue bet 25% of the pot and see what the bot does. 75% when he has nothing he will fold when he will be getting 5-to-1 odds.

Also think about if the bot bets preflop, you call. Then bets on the flop, you call. If the bot was bluffing on the flop... how can it know whether to keep bluffing on the turn? I don't think it can. So that means you always call on the flop and see what the bot does on the turn. If it checks, it has nothing, if it bets, it has something.

Bots are definetley more limited than the human mind when it comes to deciphering things. Why? Because a human has created it! Some of you guys act like the bot is going to be an unstopable T-10000 at the poker table... but it really isn't, not if you're a halfway decent poker player.

[Snoogins47]

Your argument here pretty much amounts to "I don't understand how to program a successful bot, therefore, it is not possible."

If you raise preflop, get called, bet the flop, and get called, how do YOU know whether or not to fire another bullet?

[b33nz]

Through experience and just how you feel about your opponent. Once you gain enough experience, you will be playing much on instinct, something a bot doesn't have. This is how the pro's play and I think that's a fine way to play.

[gaash]

On any given day anyone can win... I can go all in every hand and still have a chance to beat even the best players. A perfect bot is difficult to create because people play imperfect poker, and 'perfect' play involves taking advantage of these mistaskes.

www.aces-up.net

[Scotty.]

The bot's bluffing and semi-bluffing tactics would be programmed according to perfect game theory. It will take the size of the pot into account, and bluff or semi-bluff at a frequency so that your EV is 0 whether you decide to call down or fold. This is a very simplified explanation of exactly what it would do, but that is the basic idea behind it. Read the bluffing and game theory chapter as well as the optimum bluffing frequency section in TOP for more information on this.

[WhoIam]

Theoretically, you could train a neural net to play NL at an above-human level. The most important difference between this and most of what has been discussed, is that instead of being programmed with formulas and so forth, it would play intuitively, the same way top humans play. We say that NLHE is about psychology, but what we really mean by this is pattern recognition. "Reading" people is nothing more than recognizing paterns in their body language and betting patterns. A computer could theoretically pick up on patterns far more subtle and complex than the human brain could process. Even a player like Phil Ivey, who is probably unreadable to any living person, would have detectable traits and could be defeated. The neural net would be infinitely adaptable to any level or style of play.