Will computers dominate poker as they did with chess?

[Robmaf]

Lots of people accused Kasparov of a scam. Most probably he was paid to lose the game. Same as Kramnik when he lost in one move. And it was for the first time in his career. Chess computers are really strong, its counting is extraordinary. In poker computers have much less chances. I can agree poker computers will count much better. It can also be more stable, play without emotions.
Watching poker, playing poker I have never seen or even heard of poker computers taking part in live tournaments like in chess is.
I think that beating the best poker players will be possible when computers will start reading emotions and thoughts perfectly.

[RIIT]

[ QUOTE ]
Lots of people accused Kasparov of a scam. Most probably he was paid to lose the game. Same as Kramnik when he lost in one move. And it was for the first time in his career. Chess computers are really strong, its counting is extraordinary. In poker computers have much less chances. I can agree poker computers will count much better. It can also be more stable, play without emotions.
Watching poker, playing poker I have never seen or even heard of poker computers taking part in live tournaments like in chess is.
I think that beating the best poker players will be possible when computers will start reading emotions and thoughts perfectly.

[/ QUOTE ]

Well the reason you've never seen a pokerbot play in a live real money tournament is because it just happened for the first time ever last week:

2007 PBWC RESULTS

For those of you who don't agree that it was the first time:

The Golden Palace event in 2005 doesn't count because there were no real money entry fees paid by either the botters or by Layne Flack (i.e. gambling was not involved) and the initial rounds did not include human players.

The UofA duplicate poker event earlier this year doesn't count for the same reasons (i.e. gambling was not involved) and the event was not open to the public. It had very little to do with real world real money conditions. The event would have been much more significant to me if the UofA would have simply used their $50k budget as $1k prop bets against any and all opponents (human or bot) for some fixed number of hands. I'm less interested in the sterile lab like conditions.

It is not real poker unless financial losers are created.

RIIT

[RIIT]

[ QUOTE ]
lol obv its logic that rake does not make it a zero sum game anymore, i think this is so obvious nobody really mentioned it.

this won't be a problem because if you are human and not the perfect bot, you will lose more than rake, because its not possible to play perfectly optimal as a human because our randomizer sucks.

look at roshambo. no player will ever turn zero ev against a computer if he doesn't know if hes playing one or not...

in a raked game the optimum is the same as in a non raked poker game so who cares?

and we already stated out that we do not know what happens if there are multiple players because of collusion, but it will most likely not really create problems because of the reasons i allready argued (everbody more or less going for max ev, lots of possible strategies and so on and so on.) even though it's not proven

but dude this is so far away from your first statement that i really dont know what your point is really...

but i'm glad you brushed up your gametheory skills in process of this discussion...

[/ QUOTE ]

http://forumserver.twoplustwo.com/sh...50231#12350231

My apologies for the confusion created by using incorrect terms. I've been meaning to say "pure strategy" instead of "absolute/static strategy"

So what I've been trying to say is that for every pure strategy in HU holdem there is at least one other pure strategy that trumps the first.

RIIT

[indianaV8]

[ QUOTE ]
[ QUOTE ]
As it had been mentioned several times in this forum, the existance of equilibrium has been proved back in 1950. And of course unbeatable holdem player exist. The only question is how far is the research from computing it.

[/ QUOTE ]

The proof was for finite games. Is no limit hold'em discrete finite or continuous infinite? Was it really a proof for all finite games? All the games presented are static. Hold'em is dynamic.

[/ QUOTE ]

No limit holdem is discrete finite in all discussion here. No poker site allows you to bet <0.01
Would be irrelevant for any practical considerations anyway.

What was it exactly:

Proc Natl Acad Sci U S A. 1950 January; 36(1): 48–49.
Equilibrium Points in N-Person Games
John F. Nash, Jr.*
Princeton University

http://www.pubmedcentral.nih.gov/pic...p;blobtype=pdf

[indianaV8]

RIIT -- (I cut the whole quote over quote over quote stuff)

The HU case I leave aside as you agree here. Only clarification - it's the same case for HU NL holdem, not only fixed limit.

Rake is irrelevant to the discussions in this thread, you can remove it from any discussions.

Same for the luck factor, cards to come (you asked - "irrelevant to what?") - it's irrelevant to almost any theoretical poker discussion and 100% irrelevant for the topics lately discussed in this thread.

The left open point (from your perspective) is multiplayer games. The situation here (to my knowledge) is:
1) Nash equlibrium (not one, many) exists (proved)
2) Which means: a set strategies exists, such that no player has an incentive to deviate from it
3) So question is: do we assume there are players that don't play to win? I.e. don't play to maximize their profit / minimize loss; by exploiting the field how much players deviate from equilibrium and so on.

Here, depending on the distribution and position of these players some optimal (NE-based) players may have advantage or disadvantage over other NE players.

Here I think the equlibirium play will still be good although I'm not 100% certain. My reasoning is that if we have players that don't care to win (e.g. say player2 is maniac that always goes all-in in 3 players game) we can expect that randomly and uniformly this player will be 50% on our left and 50% on our right, hence will not change the big picture.

So all in all, I believe while the above sitiation will lead to incentives for a calculator to play explotive much faster based on much smaller set of assumptions and opponent modeling, rather than play the NE-based play, the deep-blue effect will be the same.

But again (for all observers) - I don't expect this to happen soon, in the next 5 years, and I believe that poker (at least NL HE) will decline first because of other reasons than bots or computer aided play.

To say it with other words: I hope to have at least 5 more profitable years for my NE bots -- i'm kidding of course ;-)

[Paxinor]

100% agree in every point indiana made...

[questions]

Here's a random thought I have about poker bots/a deep blue poker computer program:

In order for a bot to win (in the nightmare scenario posited by the threadstarter), a human player must lose to it. Therefore, the sites must continue detecting and prohibiting bots. But even setting aside the notion that the sites can do this, if a PLAYER detects that they are playing against a bot, they can decline either to play at a table with one, or else simply decline to enter into a hand against a bot, except perhaps with aces.

[indianaV8]

[ QUOTE ]
Here's a random thought I have about poker bots/a deep blue poker computer program:

In order for a bot to win (in the nightmare scenario posited by the threadstarter), a human player must lose to it. Therefore, the sites must continue detecting and prohibiting bots.

[/ QUOTE ]

(I use this not only to answer, but also to expand on my point)

This is the same reasoning as to say: In order for a pro player to win, there should be someone to lose, therefore sites should prohibit winning players.

Sites need to prohibit bots in the general case, as otherwise their T&C are going to be inconsistent. But a single non-multi accounting bot is no issue for the site and for the players (at present).

You can look at bots as emerged winning players, nothing more. Sites need to fight with multiple accounting (which is another issue, also other forms of cheating but they're irrelevant in this discussion).

Winning player (bot or not) that is able to play on every table certainly ruins the game. Otherwise how big do you think is the difference between bot and winning player (assume no multiple accounts)? There is none. Many poker sites have just about 10 tables active at given limit. A pro can easily play on 12 tables i.e. populate all tables as the bot.

The bot can play 24 hours - but this is easly identified by sites and other players.

To summarise, if you assume sites will ban any obvious non-human play (like 24 hours), and they follow seriously on multiaccounting and other cheats, bots are not an issue by itself. Nowadays, bots are build as you build your strategy. So a lot of bots are losing one, I believe the ratio is the same as winning vs losing players.

Nowadays there are (rule based) frameworks that allow easily to plug in play profile. These are limited, and in the bot case this deviates to one of two things:
a) You have personal profile that you tune continiously and keep for yourself. This has some chance for success, but it will take you quite some time - similar to the one you spend to learn the game and play it without the framework. And at the end / at some point you'll hit numerious limits that will hinder your game.
b) Some guy comes out and claims he has a winning profile, and for whatever reason sells it. In actuallity this is a losing profile. In some cases it is not, but in this cases it does not stay winning for a long - once players get it and start to use, all pro players adjust and start to profit against it, and the profile soon becomes overall loser.

Point b) is due to the rule based nature of the bot frameworks, hence if NE and more advanced generic game theory-based bots are developed (for which the research has to progress first), this will be the deep-blue effect, independnat from botting.

[ QUOTE ]

But even setting aside the notion that the sites can do this, if a PLAYER detects that they are playing against a bot, they can decline either to play at a table with one, or else simply decline to enter into a hand against a bot, except perhaps with aces.

[/ QUOTE ]

The "aces only" setup is not possible. If a player enters the game, he need to post the blinds (eventually) and then if he folds everything except aces, that will be bad for him.

If the research advance, and the forms of poker that exist today do not decline in the next many years, then "helpers" will be sold in the form of physical hardware calculators.

[questions]

[ QUOTE ]
If the research advance, and the forms of poker that exist today do not decline in the next many years, then "helpers" will be sold in the form of physical hardware calculators.

[/ QUOTE ]

Good response - worst-case scenario (IMO) is that if a poker deep blue is created (and you just KNOW that there are people out there who will not stop until one IS created), then simply another new competitive computer endeavor will emerge.

How far away is a poker deep blue, or even a junior poker deep blue??

[RIIT]

[ QUOTE ]
Here's a random thought I have about poker bots/a deep blue poker computer program:

In order for a bot to win (in the nightmare scenario posited by the threadstarter), a human player must lose to it. Therefore, the sites must continue detecting and prohibiting bots. But even setting aside the notion that the sites can do this, if a PLAYER detects that they are playing against a bot, they can decline either to play at a table with one, or else simply decline to enter into a hand against a bot, except perhaps with aces.

[/ QUOTE ]

And what if one or more bots loses to a better bot? Where is the victim here?

So many of these types of scenarios offered always presume the human to be at a disadvantage.

RIIT