In theory, perfect poker is possible.

[adamstewart]

Let's say that all of the major online poker sites have complete databases of all the hands that were ever played. Accordingly, they also have the potential for the world's largest "PokerTracker" database, on every single player that plays - especially if all the sites combine their resources.

When we play poker, it can be broken down into a series of individual decisions. Each decision has limited options, namely check/fold, call, or bet/raise. In theory, there is a "superior" or "best" play at every decision - accordingly to mathematical odds, equity, etc.

Now, one play may be "superior" against a particular opponent. Whereas other times in the seemingly exact same situation a different play may be superior, because it is against a different opponenent.

An analogy may be drawn to blackjack, where it may be correct to "hit" if the count is beyond a certain number, or "stand" otherwise. One's "correct" play depends on the count of high/low cards that are left in the deck.

In poker, this is where "reads" or "player stats" enter the equation. It could be either calculated (or analyzed empirically via the online sites' extensive databases), what play is "correct" against any given opponent(s) in any given situation.

For example, against a 56/4/1 opponent, raising may be the best play in a particular situation. However, calling or folding may be superior against a 14/7/2 opponent. Again, all of this could theorectically be analyzed empirically using the online sites' databases (and perhaps backed up with mathematical proofs?).

Imagine all of this information and database statistics were applied to a computer program that could go through an algorithm in order to derive the best play at each an every decision at the table. (Remember: this program has access to the largest "PokerTracker" database in history). Admittedly, this would be a labourous and difficult task, but nonetheless it remains possible.

As a result, you have a machine that plays perfect poker - or as good as one could ever hope to be.



Adam

[NSchandler]

Well in theory, "perfect poker," that is, an unbeatable mixed-strategy Nash Equilibrium, can be applied without knowing the tendencies of your opponents.

Knowing a player's style helps only because deviating from "optimal" strategy can win more money than playing "properly" if your opponent is making mistakes.

You don't need to know anybody's tendencies to play unbeatable poker, but you won't necessarily be winning as much as you could.

[twindrake]

In theory, a perfect life is also possible...

[Matt R.]

I think this argument is giving too much weight to the idea that an opponent will always play like his/her statistics suggest and that a few simple numbers can give an accurate description of what a player will do in every possible situation. I don't think it's possible to have a perfect "poker playing machine" simply because there is so much short term variance into how an opponent will play due to things such as tilt or him/her making small adjustments to THEIR opponents (you). For example, for a given situation, let's say an opponents statistics suggest that you should raise in that instance every time (and this will be profitable in the long term). However, any capable human opponent will eventually learn that you're making the same play against him in this instance and will adjust to that, making the correct play against him now different than what his statistics would suggest. These subtle adjustments in play make it impossible to empirically evaluate "perfect" poker.

[wheelz]

There are so many different variables involved in every decision. I mean, I had a hand the other day against a calling station with a total AF below 1 who bluffed when I checked to him on the river on a 4 flush board. Your machine would auto-fold the river based on the stats that were shown, but sometimes the stats just aren't enough. There's so much more to the game.

You'd need a machine that was able to predict the future to play perfect poker, not one that could read past results.

like Matt R. up there said just before me...

[adamstewart]

I think you guys are missing my point:

If our opponents' cards were flipped over, it would be possible to play absolutely perfect poker (Fundamental Theorem of Poker).

That is not possible.

All that we have to go on is probability. And the probability of our opponents' holdings based on:

(i) their various statistics
(ii) empirical evidence, given past hand histories


Not to mention the past histories of players with similar statistics in similar situations. It's all there, empirically available in the nearly infinite amount of hand histories the sites have access to.



Adam

[Voltron87]

no, it isn't, if you take into account that opponents will not always do the same thing in similar situations, there is a certain randomness to many decisions. you can't always predict exactly what an opponent will do, which is needed to play perfect poker.

[Zetack]

[ QUOTE ]
no, it isn't, if you take into account that opponents will not always do the same thing in similar situations, there is a certain randomness to many decisions. you can't always predict exactly what an opponent will do, which is needed to play perfect poker.

[/ QUOTE ]

I think what he's saying is that if you knew that he always did something say 93 percent of the time in a certain sitution you could determine the optimal play against him knowing that, not that its necessary for the oponent to do the exact same thing every time in a certain situation.

--Zetack

[wheelz]

Right. I suppose you could then.

[CallMeIshmael]

Adam, though I do agree with you that, in theory this is possible, I contend that we do not, at this point, have the adequate HH database.

Poker situations are so varied, that even for players with 1 million hands (clearly getting towards the biggest online players) they have only seen some situations a few times.