Live HU vs Phil Laak (spew?)

[dbirider16]

I imagine that if someone could make a bot that bases its decisions off of a huge pokertracker database, it would play very profitable at various limits. It would know every players tendincies for how they play in every hand/situation and capitalize off of it.

[spino1i]

It seems like you got a hopeless amount of money in on an extended bluff, but maybe this move is above my head.

[PJo336]

no whats above ur head is the rest of the posts u obvi didnt read if ur still commenting on HIS play of a hand which he SAID he didnt play..AHHHHHHH PAY ATTENTION

[AndyatSD]

[ QUOTE ]
Andy, why do you think computers have been beating the best chess players for years, while computers have not even come close in a seemingly much less complex game such as poker. It would seem there is an element in poker that computers are just not able to crack.

[/ QUOTE ]

Warning: Long. The tl:dr crowd can safely skip.

You propose a given - that Chess is a more complex game than Poker. I would challenge that in the context of decision making. Here's why:

See, I've always thought of it that Poker is a far more complex game than Chess for computers to calculate, and that's the 'an element in poker that computers are just not able to crack'. And in my post - I did not say that computers can not crack this - it is just my understanding, as I'm in the software development business, that the current computers accessible to the public will not be able to perform the calculations required for poker in any reasonable amount of time.

Chess has perfect information. Each time a move is made the decision making process starts from a 'from that point forward' basis - where calculating that particular moves and future moves do not necessarily depend on information from previous moves and certainly there isn't an element of uncertainty - past information is all present there for everyone to see. The board is what it is - there are no hidden information. This takes the decision under uncertainty aspect out of Chess, and for that reason along will make it so much easier to calculate.

Poker is much different. There remains unknowns that you can only approximate. For example, you never reliastically *know* 100% what two cards a person has. The most you can do is to create a range and as you collect more information, refine this range. To cut a long story short, the amount of premutations that a computer has to chug through to define through and then actually calculate is just too vast. Partly this can be thought of as someone not coming up with the proper algorithm yet (that may alleviate some computational issues by grouping types of hands together or whatnot), but I still attribute more to just pure lack of computational power at this stage of the technology game. And this is only on a pure math basis - not even really taking into consideration that the statistical modeling can vastly change with things such as PT data.

Consider a pretty simple situation of Button open raise, sb fold, computer is the BB and decides to call with T7o. Board is K94 with two diamonds. At this point, to calculate the optimal strategy of whether to come out betting with intention to fold to a raise, bet with intention to call a raise, bet with intention to 3 bet (actually, some of these actions don't have to necessarily calculated until the first raise), check with the intention to raise, check with the intention to call, or check with the intention to fold - the decision-making process would start with something like this:

WARNING: Completely made up hypothetical #s coming into play only for proof of concept.

1) Evaluate the likely hand of Button.
Since it is impossible at this point to determine what exact two cards the button has, it is required to put the button a set of hands. Obviously the button would raise with any premium, but what else would he raise with? Any pocket pair seems reasonable, any Ax seems reasonable, if this is a strictly mathematics player it's not hard to compute the range at this point. But what would the distribution look like? The hands with a K, T, 9, 7, & 4 goes down proportionally since those 5 cards are known, the pairs obviously less likely than the unmade hands by the 4:3 or whatever the #s is (I never said I was good at remembering the #s). To be even more granular, the hand combinations with diamond as one of the suit goes down, and yadda, yadda, yadda. You end up with something like:

AA, QQ, JJ, 88, 66, 55, 33, 22 - 1% [each]
KK, TT, 99, 77, 44 - .5%
AK, AT, A9, A7, A4 - 1.15%
AQ, AJ, A8, A6, A5, A3, A2 - 1.3%
KQ, KJ, K8 - 1.15%
KT, K9 - .9%
K8, K6, K5, K3, K2 - .7% (because maybe the player is less likely to raise those hands from button unless they were suited, or whatever).
T9 - blah %
T8 - blah blah %
J9 - bleh bleh %
27 - ??? %

You get the point.

2) Decide what to do
You (the computer) evalute the board versus what you determined in 1 to figure the best exploitive strategy (read Mathematics of Poker by Bill Chen if you haven't) against the weighted average or whatever in order to come up with what the 'best decision there is to do in this situation'.

Then after you make the decision, and BAM ,the turn comes. All of a sudden you have to go through this process again, re-evaluate and calculate the different likely weighted distribution process and then to add insult to injury you have to take into account the preceding action from the flop.

All this so that the computer can then chug chug chug the EV calculation to come up with 'optimal strategy'.

I'm not explaining this very well - but to put a long story short I think it's complicated.

In a practical sense, the human can process a lot of this information on an intuitive level and 'guesstimate' and make a snap-decision. It won't be optimal, but the 'better' player you are, the closer.

So that's why right now it's still humans winning and dominating the games. In 10-20 years? Who knows.

Also - as a disclaimer (that should have gone on the top) - I reserve the right to be out-right wrong on this, and I will directly attribute that to my lack of knowledge as I did not choose to specialize in mathematics or poker. This is a hobby - I only have hobby-est level enthuasism for the game. The days when my buddies and I get together and draw game theory trees to analyze the best moves to perform for poker are a thing of the past. : )

~andy

[iveyleague24]

I thought he had pocket kings

[stinkypete]

so was phil valuebetting or bluffing the river?

[CallMeIshmael]

[ QUOTE ]
Technology is 'doubling' on a daily basis, so there is a theroretical possibility.

[/ QUOTE ]

you are aware that means technology is a billion times better today than last month, right?

and, a year from now, it will be like more than a google times better

[daryn]

valuebluffing

[SomthingOrOther]

so if he checks behind computer is owned?

[StrictlyStrategy]

[ QUOTE ]
I'm of the school of thought that since Poker is nearly a pure mathematics game

[/ QUOTE ]

If I could create a bot that would follow whatever logic you gave it, and I played it HU and whatever stakes I wanted -- would you let this happen?