Monday, November 3, 2014

"The Brain"


"Texas Hold ‘Em Heads Up Poker" machine play the limit version of the most popular poker game around.  The casino owners obviously think that no human opponent could be able to defeat "The Brain" on a regular basis.   However, the manufacturer estimated that only 100 around the world will be able to beat it on a regular basis.  But that remain to be seen.  Some accomplished world class poker players and poker theorists think otherwise. 

I’ve seen about 10 different [heads-up] pros playing the bot, all with their own special strategy which they are convinced is the one that beats the bot. I feel less than 95 percent certain that they are all losing players against it, and at least one of their assumptions has always been wrong. I feel after 100 hours of play against it I can break even against it. I think I’d rather play craps, though.”  ---  Anthony Rivera, high-stakes limit pro and World Series of Poker bracelet winner

You vs "The Brain"

Non adapting strategy "artificial neural network" 

"Texas Hold ‘Em Heads Up Poker" machine was originally designed by Dahl to learn while playing and adjust his strategy to the opposition. If, for example, an opponent folded a lot, it played aggressively; if it faced aggressive play, it tried to trapHowever casino commissions mandate that a gaming machine cannot change its playing style in response to particular opponents. 

Playing a fixed strategy, "The Brain" isn't able to change its tactics to take advantage of the bad play of a novice player. A skilled human poker player or a computer algorithm with opponent modeling capability should expect to win at a higher rate than this game against especially weak opponents.  So the poker game must play a World Series of Poker champion the same way it does a neophyte.  Nonetheless an unskilled player will lose more against "The Brain" than a skilled player. 

Perfectly defensive game

What seem to be a disadvantage turned into a huge advantage. Due to the fact that in theory a non adapting strategy could be unbeatable.  If its play away from its approximation of game theoretic optimal strategy to take advantage of opponent mistakes, then it itself is playing less optimally and may be counter-exploited.
With that being said, since the machine bot needs to be stable and can't figure out weaknesses in the opponent and find ways to exploit those weaknesses.  So instead, the effort were put on defense. The basic idea behind its play technique is "to prevent itself from being exploited." "The theory behind it is almost paranoid."  So that it would teach itself to play a perfectly defensive game. Rather than steer it to study its opponent and try to capitalize on weaknesses, the net was directed to make itself as hard to beat as possible.

"If it "adapted" it would open itself up for explotation. In theory a non adapting strategy could be unbeatable. I played it for a few minutes and it was quite aggressive and tricky. It even check raised bluffed the river once. At first glance it appears to me that only very good players will beat it, they won't beat it for too much, and that the correct strategy to beat it is a bit different than against typical humans." -- David Sklansky, respected poker author and mathematician

Texas Hold ‘Em Heads Up Poker "The Brain" strategy

Hyper-LAG / maniac
+ Raise almost every time from the Button
+ Raise called button from over 75% of the time from the BB.
... Sometime 3-bet weak hands like T5off (negligible)
... Occasionally fold on the Button or when raised in the BB (negligible)
- Fold more hands from the SB on most streets than from the BB

Very LAG
+ Routinely raise with middle pair and a weak kicker
+ Call most of the time a check-raise+ Rarely 3-bet the flop with weak overcard, no straight, flush or backdoor possibilities
+ Often fold to a bet if a high card comes on the flop
+ But he will bet the high card on the flop
+ Most likely to fold on the flop

Tightens up
+ Check-call with middle pair and weak kicker
+ Check the turn most of the time 

+ Call with a piece of the board
+ Call down with Ace high unless the board is scary
+ Likely to call with King high
+ Sometimes call down with weaker hands (J high) on a bimodal (monster or miss) board.  Like a pair on the board.

+ Reasonable adjustment to what the opponent might hold based on the board.
+ Seem to play top pair / top kicker more slowly on board with flushes and straights possible
+ Seem to play high card / no pair more aggressively on paired flops.  Check-raise bluff
+ More likely to check-raise if opponent put the last bet on the previous street.
+ Doesn't fold too often.

Why is "The Brain" so good?

a) Defense. The basic idea behind its play technique is to do everything it can "to prevent itself from being exploited".  "The theory behind it is almost paranoid," 

b) Unpredictable. The pokerbot use knowledge gained from billions of staged rounds of poker fed through neural networks, and the result is an unpredictable poker player that can win almost every time.  Three different banks of knowledge are used depending on the gameplay scenario, but the basic idea behind its play technique is, as I said above,  "to prevent itself from being exploited." "The theory behind it is almost paranoid." So "The Brain" unpredictibility and three different game plays are part of his perfectly defensive game. 
Here are his 3 tag-team fighters alternating against an opponent.
    * The first is a neural net with optimal number of bluffs and can do anything in anyone hand.
    *The second play a slightly different style.
    *The third come into play when the opponent has a reduced stack.

Aggression level might change at random moment

c) Aggression.  "The Brain" is the aggressor. Almost never check-calls or simply matches an opponent's bet without a raise. But give credit to your hand when you raise and reraise.  Far too aggressive and steals far too many pots to get beat on a regular basis.

Originally published in:

No comments:

Post a Comment

Your comments are always appreciated