[jjbrr]

A trained chimpanzee can play poker, fwiw (and potentially be a winning player!)

[JLOGIC]

NickRW, on Jul 19 2010, 09:52 AM, said:

I haven't played duplicate poker, but the variance must be considerably reduced - just as it is for the duplicate versus non-duplicate forms of bridge. I do know that duplicate poker (heads up, limit) was the form used in a fairly recent man versus machine challenge.

Duplicate poker is far different from duplicate bridge. Yes it has reduced variance, but there is still a ridiculous amount of variance.

Most obviously, people will still sometimes get it in when they shouldn't (and if they didn't you have no edge), and what cards come next will still determine huge pots. Even in duplicate poker it sucks when they 2 outter you in a pot they shouldn't be in, or if you correctly slow played because they were going to barrel off but the other guy hit, and at the other table they just shoved the turn "to protect their hand" and won. Etc Etc.

For instance considering a ring game, it is a joke. Whether a guy decides to make a marginal open raise, or whether a different guy decides to make a marginal 3 bet vs call, can make a huge impact on the hand for you. This is much more true than in bridge, because there are so many decisions that are marginal. Then there is bet sizing, etc etc. The truth is you're unlikely to ever see a hand duplicated other than stuff like one guy raises and everyone folds, in duplicate poker. There is still a huge amount of variance, the only thing that is better is that in cooler situations you are protected like AA vs KK and the board is coming xxxxx.

Yes it was an ok format for heads up limit hold em but that is quite a different animal.

[JLOGIC]

qwery_hi, on Jul 19 2010, 02:55 PM, said:

JLOGIC, on Jul 18 2010, 02:42 PM, said:

Also, bridge is a much much much more complex game than poker.

Disagree. Bridge *can* be a much more complex game than poker, if the restrictions on bidding were lifted, but as the games stand, poker is much more complex than bridge.

One test that I have for complexity is who would be the winner in a computer vs human WC match. In a bridge contest, I think Zia long ago withdrew his 1M challenge. In poker, only very recently have the computers been good enough to take on the pros 1v1, let alone play at a table with 10 pros.

lol I don't even know what to say to this...

[jdonn]

jjbrr, on Jul 19 2010, 05:14 PM, said:

A trained chimpanzee can play poker, fwiw (and potentially be a winning player!)

That's why I'm so sad that I'm so bad at it.

[qwery_hi]

hotShot, on Jul 19 2010, 01:50 PM, said:

The inability of software developers to write a good poker program has nothing to do with the complexity of poker.
It has to do with the limitations of computers and with limitations that you set to allow the software to make use of all available data.

As big a LOL as I've ever read. Do you have any idea of the research efforts invested in computer poker? See, for e.g. http://www.coral-lab...marc/aaai07.pdf

[qwery_hi]

NickRW, on Jul 19 2010, 02:12 PM, said:

Zia may or may not have withdrawn his challenge, but computers can, albeit over a large number of hands, match and just about beat world class pros at heads up limit. They've had an at least competent bot to play multi player limit for years. They're working no limit and multi player versions to get up to that sort of standard. Currently the best computer bots aren't in that sort of league at bridge - given the speed of modern cpus they can play the cards well enough - but the bidding remains a chellenge.

Nick

1. Heads up limit and heads up no limit are two vastly different games.

2. "Zia may or may not have withdrawn his challenge" -
http://bbi.bridgebas...g/ginsberg.html

3. "Currently the best computer bots aren't in that sort of league at bridge - given the speed of modern cpus they can play the cards well enough - but the bidding remains a chellenge." - I'll trust Zia more than you on this for sure.

[nige1]

JLOGIC, on Jul 19 2010, 09:45 AM, said:

Poker is a game that is always played for money, and is thus a gambling game. It does not matter what the percentages are, you are still gambling every time you play for money. Sure there are such things as favorable gambles, but that is irrelevant. Some people on poker forums try to say "playing poker is not gambling because it's a game of skill." This implies you cannot gamble on a game of skill which is obviously nonsense.
Similarly, if bridge was played as rubber bridge, it would be gambling.
Bridge is not commonly played as a gambling game, and is not considered one in general. Almost all tournaments are duplicate, and are not based on a buy in with a prize pool or gambling.
There is and probably always will be a stigma attached to gambling games. Bridge does not have this. It is a legitimate concern that if bridge and poker are thought of similarly, people will think that bridge is a gambling game and that will turn some people off of it.

I agree with JLOGIC although I don't understand how a Mind Game can be a sport

SPORT, on Dictionary,Definition, said:

• noun:  an active diversion requiring physical exertion and competition
  
  
• noun:  the occupation of athletes

[JLOGIC]

qwery_hi:

I would gladly bet $100,000 on myself and a partner against any 2 computers playing in a partnership in a reasonably long format. I would submit to being monitored at all times during this match to ensure no cheating.

If there needs to be more money at stake I could try to raise funding, I doubt it would be a problem, but I cannot guarantee betting a million dollars since I don't have it. And I am not close to being the top player in the world. I'm sure if you wanted to do Meckwell vs 2 computers a lot of funding could be raised for that but I cannot speak for them or their participation.

Do you really think anyone with a computer program would touch this bet? If you do, please contact them and inform them of this offer.

If your only basis for saying that bridge played with some system regulation is a less complicated game than any form of poker, then hopefully you can see that is pretty silly since a lot of bridge players would put up a lot of money for a matchup against a computer.

[JLOGIC]

Qwery_hi, there seem to be 2 things you are implying.

1) If computer bridge players are closer to top humans than computer poker players are to top humans, bridge is a less complicated game.

How can that be true? Chess computers are MUCH better than top human chess players now. Does that mean chess is not complicated?

Of course it is absurd that you think computer bridge players are close to top humans right now, but that's a different story.

2) If Zia won't bet a $1mil vs computers, computers play better than Zia.

Are there really no other reasons Zia wouldn't do this? Perhaps if he thought he was 100 % to win he would do it, but if he's only 95 % he wouldn't. Not everyone has a ton of gamble. Also, I'd be willing to bet you that Zia would be willing to do this bet right now if offered. I would also be willing to bet that someone like Steve Weinstein or Nick Nickell would be happy to put up the money for Zia if he didn't want to.

[awm]

Most places don't put much restriction on bridge bidding after the first round of the auction. There are still a ridiculous number of bidding systems possible -- I don't see any reason to believe that disallowing a few opening bids makes bridge somehow "much less complicated" than it would otherwise be.

Poker relies to a great degree on reading other people. Computers are not good at reading people. So a single computer player entering a tournament filled with humans would not have as good an expected performance as a top human pro. But if you put a computer up head-to-head against a human pro, the human has just as much difficulty "reading" the computer as the computer does with the human, and it becomes a contest of mathematical skills. In the popular form of tournament poker there are only 1,326 possible hands (two-card hands) some of which can be eliminated based on the visible cards. Do you really think a computer program can't beat a human at this game consistently? Many people have suggested that computer programs can and do win here.

Bridge computer programs still have a long way to go. Played any hands against GIB lately? Yes, I know there are better programs (and GIB would play better if you ran it on a super-computer) but we're still pretty far from the robots challenging the human experts. There are way too many possible bridge hands to simulate all of them, even on a super-computer.

[NickRW]

qwery_hi, on Jul 19 2010, 11:25 PM, said:

1. Heads up limit and heads up no limit are two vastly different games.

For sure.

Quote

2. "Zia may or may not have withdrawn his challenge" -
http://bbi.bridgebas...g/ginsberg.html

That is a reproduction of a somewhat hyped article - I wouldn't take it as gospel. But if Zia said the bet is not on the table indefinitely - probably that is quite prudent - no doubt someone will really crack the remaining problems for computer bridge sometime soonish.

I would like to point out that Matt Ginsberg's work over a decade ago now on suitable algorithms to make DD analysis fast enough that it could be used to help computers in realtime problems at the table was brilliant. However, apart from some studies at improved point count methods, I don't really know of any further breakthroughs in that time specifically with respect to the game of bridge.

Quote

3. "Currently the best computer bots aren't in that sort of league at bridge - given the speed of modern cpus they can play the cards well enough - but the bidding remains a chellenge." - I'll trust Zia more than you on this for sure.

I'll put my faith in Justin's bet for now.

Nick

[qwery_hi]

awm, on Jul 19 2010, 06:36 PM, said:

Most places don't put much restriction on bridge bidding after the first round of the auction. There are still a ridiculous number of bidding systems possible -- I don't see any reason to believe that disallowing a few opening bids makes bridge somehow "much less complicated" than it would otherwise be.

Poker relies to a great degree on reading other people. Computers are not good at reading people. So a single computer player entering a tournament filled with humans would not have as good an expected performance as a top human pro. But if you put a computer up head-to-head against a human pro, the human has just as much difficulty "reading" the computer as the computer does with the human, and it becomes a contest of mathematical skills. In the popular form of tournament poker there are only 1,326 possible hands (two-card hands) some of which can be eliminated based on the visible cards. Do you really think a computer program can't beat a human at this game consistently? Many people have suggested that computer programs can and do win here.

Bridge computer programs still have a long way to go. Played any hands against GIB lately? Yes, I know there are better programs (and GIB would play better if you ran it on a super-computer) but we're still pretty far from the robots challenging the human experts. There are way too many possible bridge hands to simulate all of them, even on a super-computer.

Humans are no better at reading computers than computers are at reading humans. In fact, I think in a human vs computer poker game the computer has the edge since it has perfect memory. So how well have computers been doing in tournament poker vs 8 other good players? Not too well afaik, have you heard otherwise? The number of hands is one small element of the complexity.

I've made my arguments, you won't convince me and I won't convince you. Now if only the ACBL allowed 1 human + 3 computers to enter into the spingold, that would be a first step in deciding this argument.

[mike777]

I see no reason why computers could not read humans as well as the best human poker players....Give it time.....

but then I see no reason why computers could not have a billion or more times measurable higher IQ than the entire human race by 2050.

[fuburules3]

I think AWM's point was that in a poker game with 8 humans and a computer, the top poker pro would be able to read the 8 other humans whereas the computer would not and this would help the pro in many hands against other humans.

[JLOGIC]

qwery_hi,

It is illegal for computers to play on poker sites. It is impossible for computers to play in live poker. So why are you expecting to hear about bots crushing tournament poker?

Funny you mention tournament poker though, computers could easily become better than humans since the stack sizes are small. Computers can already play near perfect with 10 BB stacks in single table tournaments (sit n go's), as can humans. Sit n Go's are considered dead since they are played close to a solved level. Of course bots are playing these games, as well as the DoN sit n gos. Of course you don't hear about this since it is illegal and barred by the sites.

[mike777]

fuburules3, on Jul 19 2010, 10:39 PM, said:

I think AWM's point was that in a poker game with 8 humans and a computer, the top poker pro would be able to read the 8 other humans whereas the computer would not and this would help the pro in many hands against other humans.

no


give it time..in your lifetime?

I can certainly see computers in live poker/bridge at the table in the foreseeable future....

note even in 2010 computers in many forms are banned in live bridge/poker.

[Lobowolf]

JLOGIC, on Jul 19 2010, 08:58 PM, said:

qwery_hi:

I would gladly bet $100,000 on myself and a partner against any 2 computers playing in a partnership in a reasonably long format. I would submit to being monitored at all times during this match to ensure no cheating.

If there needs to be more money at stake I could try to raise funding, I doubt it would be a problem, but I cannot guarantee betting a million dollars since I don't have it. And I am not close to being the top player in the world. I'm sure if you wanted to do Meckwell vs 2 computers a lot of funding could be raised for that but I cannot speak for them or their participation.

Do you really think anyone with a computer program would touch this bet? If you do, please contact them and inform them of this offer.

If your only basis for saying that bridge played with some system regulation is a less complicated game than any form of poker, then hopefully you can see that is pretty silly since a lot of bridge players would put up a lot of money for a matchup against a computer.

Please cc me on any responses. I've got some crap I can sell to raise $ to back Justin and (insert anyone Justin feels comfortable playing with for this kind of money here ).

[Free]

JLOGIC, on Jul 20 2010, 02:58 AM, said:

I would gladly bet $100,000 on myself and a partner against any 2 computers playing in a partnership in a reasonably long format. I would submit to being monitored at all times during this match to ensure no cheating.

If there needs to be more money at stake I could try to raise funding, I doubt it would be a problem, but I cannot guarantee betting a million dollars since I don't have it. And I am not close to being the top player in the world. I'm sure if you wanted to do Meckwell vs 2 computers a lot of funding could be raised for that but I cannot speak for them or their participation.

Do you really think anyone with a computer program would touch this bet? If you do, please contact them and inform them of this offer.

If your only basis for saying that bridge played with some system regulation is a less complicated game than any form of poker, then hopefully you can see that is pretty silly since a lot of bridge players would put up a lot of money for a matchup against a computer.

Justin,

You'd be surprised how good computer bridge has become. Jack has played against Dutch top pairs in 2006, and beaten some of them.

Jack - Bart Nab & Gert-Jan Paulissen 26 - 90
Jack - Paul Felten & Eric van Valen 43 - 60
Jack - Erik Janssen & Jeroen Top 43 - 51
Jack - Vincent Ramondt & Berry Westra 45 - 53
Jack - Jan van Cleeff & Vincent Kroes 61 - 46
Jack - Hanneke Kreijns & Just vd Kam 74 - 53
Jack - Ton Bakkeren & Huub Bertens 67 - 32

You can read it here:
http://www.jackbridge.com/eperhum1.htm
http://www.jackbridge.com/eperhum2.htm
http://www.jackbridge.com/eperhum3.htm
http://www.jackbridge.com/eperhum4.htm

Apparently they made Jack too rigid, so human flexibility was rewarded. Still, it beat Bakkeren-Bertens who won the Cavendish that year... We're now 4 years later, I'm pretty sure Jack has become better by now.

[JLOGIC]

I own a copy of jack and use it every now and then, don't think I'll be too shocked. Of course if run on a higher powered computer than my laptop I'm sure it will play even better.

Jack plays well, as do the best no limit hold em robots, no question. And I have no doubt that at some point in life, computers will play better than humans in both games.

And of course the humans it beat are quite strong. Maybe Jack's developer will take me up on my offer. The only stipulation was a reasonably long match. No doubt this is no problem to Jack if it is the better player, since it will give it a greater chance to win.

[helene_t]

I must be missing something but poker seems a very simple game to me. By simplifying the rules a little bit I can derive the optimal strategy on the back of an envelope.

Suppose Alice draws a random number between 0 and 1, and on the basis of that decides to play or not. If she doesn't play, she pays Bob one rupee. If she does play, Bob gets to draw a number and decides to play or not. If he doesn't play, he pays Alice one rupee. If he does play, the loser pays two rupees to the winner.

Then the optimal strategy for Alice is to play if she draws a number bigger than 1/3 and for Bob to play if he draws one bigger than 1/2.

It doesn't matter that they draw five cards rather than a number, you just express the value of the hand as its quantile. Shared cards, open cards, more than two players, limited-size deck and trades would complicate the calculations a little bit but I don't think it would make the problem non-trivial, at worst I would need a computer algebra program, rather than the back of an envelope, to derive the optimal strategies. I am quite sure that there would still be no room for mixed strategies.

I suppose that what makes it non-trivial is only the fact that Alice can chose between different stakes, and Bob can decide to increase the stakes, and then Alice will have to decide again etc. Or am I missing something?