Re: Live Updates of the Laak vs Computer Match HERE
 

Computer won the 1st 125 hands. +300

Session One
$10/$20 Limit Texas Hold'em Seat 1 Seat 2

Private Room Polaris ????? Phil Laak ?????

Public Room Ali Eslami -$75 Polaris +$75
After 125 hands.

Re: Live Updates of the Laak vs Computer Match HERE
 

[ QUOTE ]
The Duplicate matches should give the humans the edge, unless the computer is programmed remember and strategically use the hand histories.

I imagine the human players will easily remember when big draws hit/miss etc on certain boards..

[/ QUOTE ]

if you read the format you would see that each human only plays one side and has no awareness of the other match's results. obviously the computer is not made aware of the duplicate nature of the hands as well. so basically there is no such issue.

Re: Live Updates of the Laak vs Computer Match HERE
 

[ QUOTE ]
[ QUOTE ]
furthermore, NL HU poker has been SOLVED for ~12 BB or less.

[/ QUOTE ]

What is this about?

[/ QUOTE ]

Exactly what it sounds like. there is a strategy that no matter what strategy the opponent employs, the best the opponent can hope for is to be break even long term. the nash equilibrium strategy was found.

Re: Live Updates of the Laak vs Computer Match HERE
 

[ QUOTE ]
The Duplicate matches should give the humans the edge, unless the computer is programmed remember and strategically use the hand histories.

I imagine the human players will easily remember when big draws hit/miss etc on certain boards..

[/ QUOTE ]LOL illertateaments

Re: Live Updates of the Laak vs Computer Match HERE
 

Maybe a better match would be boob Bush vs Phil

Re: Live Updates of the Laak vs Computer Match HERE
 

I think it's cool that they've put this together, and I'm even more excited now that I've seen that it's a duplicate match. I'm looking forward to finding out on Wednesday morning how this all wrapped up, but I think the computer will acquit itself quite nicely.

Re: Live Updates of the Laak vs Computer Match HERE
 

FYI ALL,

5 years ago, Phil Laak and the Computer had a similiar match. Phil won the match, but estimated he had less than a 5% edge.

Proccessing power and memory limitations played a big part in the ability of the AI to run. Now 5 years both the technology and research have advanced tremendously.

I believe the computer has the slight edge this time around, and will win the tournament. In another 5 years the computer will be mostly unbeatable with maybe a dozen players or so standing a chance.

Re: Live Updates of the Laak vs Computer Match HERE
 

lol limit poker.

Re: Live Updates of the Laak vs Computer Match HERE
 

[ QUOTE ]
FYI ALL,

5 years ago, Phil Laak and the Computer had a similiar match. Phil won the match, but estimated he had less than a 5% edge.

Proccessing power and memory limitations played a big part in the ability of the AI to run. Now 5 years both the technology and research have advanced tremendously.

I believe the computer has the slight edge this time around, and will win the tournament. In another 5 years the computer will be mostly unbeatable with maybe a dozen players or so standing a chance.

[/ QUOTE ]


i think power is much less of an issue in this case, than it is for say Chess. this is because Chess has been researched seriously by PHDs for years (like 40-50 years), whereas poker has gotten very little attention. i think most of the weaknesses of the program are almost certainly on the software side for now, and basically could be overcome very easily just by pouring more man hours and research dollars into the project.

there are literally thousands of chess engines that have been built from the ground up. i doubt there are even a a couple dozen functional poker engines that are beyond their initial development phases.

nothing should actually be that hard to calculate when it comes to poker anyway. you don't need to search 100 million board positions. you could probably run a million pokerstove type calculations in a second or two. i think coming up with an algorithm to determine your opponents range given his play is the most difficult problem. that isn't specifically difficult from the number crunching side but rather from algorithm development side. once you have a range, figuring out your equity and what your action should be based on that equity shouldn't be that hard.

Re: Live Updates of the Laak vs Computer Match HERE
 

wake me up when they make a deepstacked nl cash bot

Re: Live Updates of the Laak vs Computer Match HERE
 

[ QUOTE ]
FYI ALL,

5 years ago, Phil Laak and the Computer had a similiar match. Phil won the match, but estimated he had less than a 5% edge.


[/ QUOTE ]

What's a 5% edge mean? Since most decisions are no brainers (find a raise with top pair, fold if you miss the flop), it sounds like a lot.

Re: Live Updates of the Laak vs Computer Match HERE
 

[ QUOTE ]
[ QUOTE ]
FYI ALL,

5 years ago, Phil Laak and the Computer had a similiar match. Phil won the match, but estimated he had less than a 5% edge.

Proccessing power and memory limitations played a big part in the ability of the AI to run. Now 5 years both the technology and research have advanced tremendously.

I believe the computer has the slight edge this time around, and will win the tournament. In another 5 years the computer will be mostly unbeatable with maybe a dozen players or so standing a chance.

[/ QUOTE ]


i think power is much less of an issue in this case, than it is for say Chess. this is because Chess has been researched seriously by PHDs for years (like 40-50 years), whereas poker has gotten very little attention. i think most of the weaknesses of the program are almost certainly on the software side for now, and basically could be overcome very easily just by pouring more man hours and research dollars into the project.

there are literally thousands of chess engines that have been built from the ground up. i doubt there are even a a couple dozen functional poker engines that are beyond their initial development phases.

nothing should actually be that hard to calculate when it comes to poker anyway. you don't need to search 100 million board positions. you could probably run a million pokerstove type calculations in a second or two. i think coming up with an algorithm to determine your opponents range given his play is the most difficult problem. that isn't specifically difficult from the number crunching side but rather from algorithm development side. once you have a range, figuring out your equity and what your action should be based on that equity shouldn't be that hard.

[/ QUOTE ]

No, the particular problem has been at "adaptive" play by the computer and analyzing and calculating its opponents possible moves based on hundreds or thousands of previously played hands. Anyone with a large db on PT will know that it can bring an older machine to a crawling halt.

When the "bot's" AI is dynamic changed with each new hand played, the hardware is definately a limitation. Of course it is not prohibitive and if the UAB had millions of dollars to throw at the program it wouldn't even be an issue, but with moderate funding and 5 year old technology, yes, its a limitation.

Re: Live Updates of the Laak vs Computer Match HERE
 

Is it over?
Have...Have the computers won?

Re: Live Updates of the Laak vs Computer Match HERE
 

[ QUOTE ]

No, the particular problem has been at "adaptive" play by the computer and analyzing and calculating its opponents possible moves based on hundreds or thousands of previously played hands. Anyone with a large db on PT will know that it can bring an older machine to a crawling halt.

[/ QUOTE ]

This seems like a very easy thing to work around though, especially for HU play. Considering how people generally use PT for little more than 3 numbers - VPIP, PFR, AF - you could get a pretty good baseline read just keeping track of those stats. If you get rid of the hand histories and most of the legions of other stats, this wouldn't be that hard to follow. And yes, people obviously are going to have other reads that aren't based on just those numbers, but I bet there are ways around that, too.

Re: Live Updates of the Laak vs Computer Match HERE
 

[ QUOTE ]
Is it over?
Have...Have the computers won?

[/ QUOTE ]

After 425 hands Ali is up by a single bet, and they won't say how Phil Laak is doing until later.

Re: Live Updates of the Laak vs Computer Match HERE
 

phil laak hand from 2k limit event

phil opens qq, gets 3bet. he calls.

flop AK6. he checks, guy bets.

phil : if i fold will you show me?'
guy : 'ya'

phil folds qq faceup
guy shows 55

whole table laughs at phil

Re: Live Updates of the Laak vs Computer Match HERE
 

[ QUOTE ]
[ QUOTE ]

No, the particular problem has been at "adaptive" play by the computer and analyzing and calculating its opponents possible moves based on hundreds or thousands of previously played hands. Anyone with a large db on PT will know that it can bring an older machine to a crawling halt.

[/ QUOTE ]

This seems like a very easy thing to work around though, especially for HU play. Considering how people generally use PT for little more than 3 numbers - VPIP, PFR, AF - you could get a pretty good baseline read just keeping track of those stats. If you get rid of the hand histories and most of the legions of other stats, this wouldn't be that hard to follow. And yes, people obviously are going to have other reads that aren't based on just those numbers, but I bet there are ways around that, too.

[/ QUOTE ]

LOL, not laughing at you, but that could be where a lot of the confusion between bots and human performance lies. The idea of programming a bot is not to mimick how a human would use PT. The bot is going to look at a flop, analyze all possible holdings its opponent could have, go back to the database and compare all the hands the opponent has played with those starting cards. Its going to compare how the opponent has played on similar boards. Its going to use every bit of information a human cannot simply recall or calculate on the fly. Parsing a large database hundreds of times on a single street just to make one decision can bog down a machine.

Re: Live Updates of the Laak vs Computer Match HERE
 

[ QUOTE ]
phil laak hand from 2k limit event

phil opens qq, gets 3bet. he calls.

flop AK6. he checks, guy bets.

phil : if i fold will you show me?'
guy : 'ya'

phil folds qq faceup
guy shows 55

whole table laughs at phil

[/ QUOTE ]

I know he's not much Yeti, I know he's screwed up in the past, but he's mankind's last, best hope, and we are all counting on him.

Re: Live Updates of the Laak vs Computer Match HERE
 

[ QUOTE ]
phil laak hand from 2k limit event

phil opens qq, gets 3bet. he calls.

flop AK6. he checks, guy bets.

phil : if i fold will you show me?'
guy : 'ya'

phil folds qq faceup
guy shows 55

whole table laughs at phil

[/ QUOTE ]

how isn't this standard

Re: Live Updates of the Laak vs Computer Match HERE
 

Does the computer have the ability to adapt to the players playing style? Or will it always play out a hand exactly the same against both players if they both make the same moves for the hand?

Re: Live Updates of the Laak vs Computer Match HERE
 

My new theory is Ali only took part in this because he thought he would get to play heads up limit against Phil Laak at some point. Imagine his surprise to find out he would be playing 10/20 limit against a computer the whole time [img]/images/graemlins/frown.gif[/img]

-DeathDonkey

Re: Live Updates of the Laak vs Computer Match HERE
 

[ QUOTE ]
Does the computer have the ability to adapt to the players playing style? Or will it always play out a hand exactly the same against both players if they both make the same moves for the hand?

[/ QUOTE ]

It adapts.

Re: Live Updates of the Laak vs Computer Match HERE
 

[ QUOTE ]

No, the particular problem has been at "adaptive" play by the computer and analyzing and calculating its opponents possible moves based on hundreds or thousands of previously played hands. Anyone with a large db on PT will know that it can bring an older machine to a crawling halt.

When the "bot's" AI is dynamic changed with each new hand played, the hardware is definately a limitation. Of course it is not prohibitive and if the UAB had millions of dollars to throw at the program it wouldn't even be an issue, but with moderate funding and 5 year old technology, yes, its a limitation.

[/ QUOTE ]

Yea you are correct when you are talking about tackling the problem in that manner (using neural networks or massive databases of past play) i think most engines that currently exist aren't using those methods however. infact, i am pretty sure there is little benefit in having 100k hands on a player rather than 10k, you only gain a bit more information about the player at a huge computing expense. of course is you had ample computing power the question would be somewhat moot, but for the most part optimizing a decision making algorithm is the tricky part these days (based on what i know, which is admittedly limited). they are taking a lot more short cuts not unlike the ones suggested by previous posters.

Re: Live Updates of the Laak vs Computer Match HERE
 

[ QUOTE ]
Does the computer have the ability to adapt to the players playing style? Or will it always play out a hand exactly the same against both players if they both make the same moves for the hand?

[/ QUOTE ]

It's probably adapting. The group has been some work on opponent modeling so you'd think they'd integrate it in.

Re: Live Updates of the Laak vs Computer Match HERE
 

[ QUOTE ]
[ QUOTE ]
phil laak hand from 2k limit event

phil opens qq, gets 3bet. he calls.

flop AK6. he checks, guy bets.

phil : if i fold will you show me?'
guy : 'ya'

phil folds qq faceup
guy shows 55

whole table laughs at phil

[/ QUOTE ]

how isn't this standard

[/ QUOTE ]


it may have just been Axx i'm not sure. it wasnt off the charts awful but he should have def capped v the guy preflop. it was just hilarious how he proudly turned over two queens, then when the guy showed 55 he pulled up his hood and put his headphones much to everyones amusement

Re: Live Updates of the Laak vs Computer Match HERE
 

[ QUOTE ]
[ QUOTE ]
phil laak hand from 2k limit event

phil opens qq, gets 3bet. he calls.

flop AK6. he checks, guy bets.

phil : if i fold will you show me?'
guy : 'ya'

phil folds qq faceup
guy shows 55

whole table laughs at phil

[/ QUOTE ]

how isn't this standard

[/ QUOTE ]
It's LHE so he's getting eleventy bazillion to one from the pot with 3rd pair?

Re: Live Updates of the Laak vs Computer Match HERE
 

[ QUOTE ]
[ QUOTE ]
[ QUOTE ]
phil laak hand from 2k limit event

phil opens qq, gets 3bet. he calls.

flop AK6. he checks, guy bets.

phil : if i fold will you show me?'
guy : 'ya'

phil folds qq faceup
guy shows 55

whole table laughs at phil

[/ QUOTE ]

how isn't this standard

[/ QUOTE ]


it may have just been Axx i'm not sure. it wasnt that awful but he should have def capped v the guy preflop. it was just hilarious how he proudly turned over two queens, then when the guy showed 55 he pulled up his hood and put his headphones much to everyones amusement

[/ QUOTE ]

for one not capping is horrendous. of course in tournaments there are a lot of reasons why you might deviate from standard play due to the fact that chipEV != cashEV, so on the face of things its hard to make fun of phil without any other contextual information.

in a cash game i think you see people cap and lead the flop more often than not.

!#$
 

[ QUOTE ]
Does the computer have the ability to adapt to the players playing style? Or will it always play out a hand exactly the same against both players if they both make the same moves for the hand?

[/ QUOTE ]

It definately is using adaptive play. You can read articles on the UAB website over the recent incorpartion of opponent modeling into the algorithm. Salami or whatver the @!$% is name is, even commented on the fact that he could feel the computer adjusting to his play during the first match.

FYI, the match a few years ago did not use any adaptive modeling.

Re: Live Updates of the Laak vs Computer Match HERE
 

[ QUOTE ]
[ QUOTE ]
Does the computer have the ability to adapt to the players playing style? Or will it always play out a hand exactly the same against both players if they both make the same moves for the hand?

[/ QUOTE ]

It adapts.

[/ QUOTE ]

IIRC this bot has different "modes" where it plays with different styles, and it is the humans which pick which "version" of the program goes into each match, so its adaptation isn't 100% automated.

Re: Live Updates of the Laak vs Computer Match HERE
 

From the live updates:

[ QUOTE ]
03:36 pm Phil says he felt that he was getting punched down. He said he got knocked down to -$1500. But he clawed his way back to $465. So the human team won by $70 -- 7 small bets which means the first match is a draw!

[/ QUOTE ]

They either have the total or the result backwards, but I guess either way it's a draw.

Re: Live Updates of the Laak vs Computer Match HERE
 

Somebody explain this to me...
Laak: +$465
Ali: +$395
Combined: +$70 and it's a draw? wtf?

Re: !#$
 

[ QUOTE ]
03:36 pm Phil says he felt that he was getting punched down. He said he got knocked down to -$1500. But he clawed his way back to $465. So the human team won by $70 -- 7 small bets which means the first match is a draw!

[/ QUOTE ]

well it was pretty close, statistically insignificant certainly. hopefully something more exciting will happen in the next set.

pretty amazing that phil went down 1500$ while the computer was never down more than 600ish it seems. maybe tilt/psychological factors that affected phil can account for this. if this is true (pretty big if), it also implies that the humans are mostly likely superior players when playing their A game.

Re: Live Updates of the Laak vs Computer Match HERE
 

[ QUOTE ]
Somebody explain this to me...
Laak: +$465
Ali: +$395
Combined: +$70 and it's a draw? wtf?

[/ QUOTE ]

http://en.wikipedia.org/wiki/Statistical_significance

EDIT: oops missed the +/- sign discrepancy posted. if you read the post they make it apparent that polaris won against phil though.

Re: Live Updates of the Laak vs Computer Match HERE
 

*deleted*

Re: !#$
 

[ QUOTE ]
[ QUOTE ]
03:36 pm Phil says he felt that he was getting punched down. He said he got knocked down to -$1500. But he clawed his way back to $465. So the human team won by $70 -- 7 small bets which means the first match is a draw!

[/ QUOTE ]

well it was pretty close, statistically insignificant certainly. hopefully something more exciting will happen in the next set.

pretty amazing that phil went down 1500$ while the computer was never down more than 600ish it seems. maybe tilt/psychological factors that affected phil can account for this.

[/ QUOTE ]

Doesn't make any sense though. Something in the reported results has to be incorrect.

Re: Live Updates of the Laak vs Computer Match HERE
 

[ QUOTE ]
[ QUOTE ]
Somebody explain this to me...
Laak: +$465
Ali: +$395
Combined: +$70 and it's a draw? wtf?

[/ QUOTE ]

http://en.wikipedia.org/wiki/Statistical_significance

[/ QUOTE ]
I'm sorry, no amount of Wikipedia searching will make 465+395=70.

Re: !#$
 

[ QUOTE ]
[ QUOTE ]
[ QUOTE ]
03:36 pm Phil says he felt that he was getting punched down. He said he got knocked down to -$1500. But he clawed his way back to $465. So the human team won by $70 -- 7 small bets which means the first match is a draw!

[/ QUOTE ]

well it was pretty close, statistically insignificant certainly. hopefully something more exciting will happen in the next set.

pretty amazing that phil went down 1500$ while the computer was never down more than 600ish it seems. maybe tilt/psychological factors that affected phil can account for this.

[/ QUOTE ]

Doesn't make any sense though. Something in the reported results has to be incorrect.

[/ QUOTE ]

doesn't it seem to be implied that phil clawed his way back to being DOWN only 465?

Re: !#$
 

If thats the case then the humans LOST by $70, not won.

Re: !#$
 

[ QUOTE ]
doesn't it seem to be implied that phil clawed his way back to being DOWN only 465?

[/ QUOTE ]

Ah! But that would mean the computer "team" won by $70.

Re: !#$
 

Obv their updates are being provided by the same monkeys that do the Cardplayer updates.