High Stakes Poker - Page 73

On March 31 2009 14:09 Ket wrote: also i think baal is talking about for a computer/algorithm poker is 1000000 more complex, not for the human mind. in this case he is right

Well as ket mentions it all depends on how do we define complexity.

Poker is mathematically much more complex, therefore the algorithm would be much longer to "solve" poker than it is Chess, same reason why Chess software can be really though, but any regular poker player could beat the thoughest of poker bots.

This is because computers rely on complete information, so in the end it all becomes a bunch of massive equations, but when there is incomplete information it makes the equation grow immensely.

The human brain is way inferior than computer for quick math calculation but it totally surpasses in other things like for example "reading the table flow", which can ultimately be elaborated into an equation, but its complexity would be staggering.


So in the end, Poker is indeed more complex than Chess, regardless of its difficulty for the human brain.


Same reason why the skill difference between GrandMasters is really small and for many many years it hasnt changed much, Bobby Fisher can play againast Kasparov... but watch Durrr rape Doyle's soul hardcore, its because we are close to chess perfection, and so far from poker perfection.

On March 31 2009 07:26 CrownRoyal wrote: i dont think you'll ever see a robot that can destroy a human.

that's already the case in limit hold'em head's up.

On March 31 2009 14:07 Ket wrote: its not a stretch for there eventually be a good nl bot that decently approximates game theory optimal play, but it'll just be a bot people 'can't beat' in the long run (or at least won't be able to beat for very much). this bot wont make nearly as much money as the best human players though because the real tough problem thats completely unthinkably impossible for a bot to do close to as well as a good human is exploitive play. good players will always eat up the weaker players at the table faster than the GTO bot ever can

OK with all those thoughts.
I was thinking in terms of "learning process" to get to the higher "stakes" (if we compare 2800elo to NL($100000)).
BTW, exploitive play is still GTO-explainable. And don't forget, all poker players are still proceeding information, so i still think it is possible to rationalize it.
I think i'm gonna make sick chess-prop bet against poker player, if it's still possible to get odds such as "4 to 1" as i already read on the internet ...

On March 31 2009 14:28 Baal wrote: Show nested quote + On March 31 2009 14:09 Ket wrote: also i think baal is talking about for a computer/algorithm poker is 1000000 more complex, not for the human mind. in this case he is right Well as ket mentions it all depends on how do we define complexity. Poker is mathematically much more complex, therefore the algorithm would be much longer to "solve" poker than it is Chess, same reason why Chess software can be really though, but any regular poker player could beat the thoughest of poker bots. This is because computers rely on complete information, so in the end it all becomes a bunch of massive equations, but when there is incomplete information it makes the equation grow immensely. The human brain is way inferior than computer for quick math calculation but it totally surpasses in other things like for example "reading the table flow", which can ultimately be elaborated into an equation, but its complexity would be staggering. So in the end, Poker is indeed more complex than Chess, regardless of its difficulty for the human brain. Same reason why the skill difference between GrandMasters is really small and for many many years it hasnt changed much, Bobby Fisher can play againast Kasparov... but watch Durrr rape Doyle's soul hardcore, its because we are close to chess perfection, and so far from poker perfection.

I understand all the concepts for sure, no need for oversimplification.
But, I think you don't realize how far the 2900 grandmasters are better than the 2500elo grandmaster (for example Kasparov-Alhekine).
And I was thinking in term of learning process more than intrinsic complexity.
I still remind to all of the people here that Polaris can beat all people in this forum with probability one on, let say 100 000 hands, in limit hold'em head's up. There might be only 3 to 10 people able to breakeven+ against him. There's no rake in this game.

On March 31 2009 14:07 Ket wrote: its not a stretch for there eventually be a good nl bot that decently approximates game theory optimal play, but it'll just be a bot people 'can't beat' in the long run (or at least won't be able to beat for very much). this bot wont make nearly as much money as the best human players though because the real tough problem thats completely unthinkably impossible for a bot to do close to as well as a good human is exploitive play. good players will always eat up the weaker players at the table faster than the GTO bot ever can

But this would end it right away, wouldn't it? The rooms would become flooded with such bots, and detecting them would be expensive and technically very hard, if not impossible. For the weak players, it would become something like playing against a slot machine.

And btw, at heads-up limit game there are bots that are pretty good:
http://www.cardplayer.com/poker-news/...ine-computer-giving-players-real-test

On March 31 2009 14:09 CruiseR wrote: Show nested quote + On March 31 2009 13:46 fakeshaver wrote: how long would it take a top poker player to go pro in chess? more than top chess turn into pro in poker imo ;p

That's exactly my point, that's why I'm arguing by the way.

On March 31 2009 15:12 pluzich wrote: Show nested quote + On March 31 2009 14:07 Ket wrote: its not a stretch for there eventually be a good nl bot that decently approximates game theory optimal play, but it'll just be a bot people 'can't beat' in the long run (or at least won't be able to beat for very much). this bot wont make nearly as much money as the best human players though because the real tough problem thats completely unthinkably impossible for a bot to do close to as well as a good human is exploitive play. good players will always eat up the weaker players at the table faster than the GTO bot ever can But this would end it right away, wouldn't it? The rooms would become flooded with such bots, and detecting them would be expensive and technically very hard, if not impossible. For the weak players, it would become something like playing against a slot machine. And btw, at heads-up limit game there are bots that are pretty good: http://www.cardplayer.com/poker-news/...ine-computer-giving-players-real-test

Polaris for example is not "pretty good", he is clearly very little exploitable, and able to defeat any humans apart maybe Hawrilenko and Bryce (on their A-game, although they are losing to it). The very reason for bots not to exist, is that there is less money to win for a good poker player in developping such bots - it means a lot of work in collaboration with great scientists, informaticians, AI-specialist... I'm pretty sure that if some very good pro were helping those kind of projects the time needed would be shorten a lot. Don't forget a computer NEVER, EVER, tilt, and tilt-management is one of the key thing.

Very interesting discussion going on and I agree that until there is an AI with a much more complex alghortim that can solve incomplete problems in a similar way the human brain does it's very unlikely that a BOT will own any NL top game.

On a sidenote I really wanna watch the next episode of HSP, is Durr gonna get owned by Barry ?

That would suck hardcore after "math is idiotic" thingy i mean, getting sucked out and then owned is terrible and Durr doesn't deserve it he's the one making the whole deal interesting.

What about Ziggy ? WTH man come on you bald sexy bastard I wanna see you play not make hero calls against Durr and always loose. >_<

On March 31 2009 15:22 dejokah wrote: Very interesting discussion going on and I agree that until there is an AI with a much more complex alghortim that can solve incomplete problems in a similar way the human brain does it's very unlikely that a BOT will own any NL top game. On a sidenote I really wanna watch the next episode of HSP, is Durr gonna get owned by Barry ? That would suck hardcore after "math is idiotic" thingy i mean, getting sucked out and then owned is terrible and Durr doesn't deserve it he's the one making the whole deal interesting. What about Ziggy ? WTH man come on you bald sexy bastard I wanna see you play not make hero calls against Durr and always loose. >_<

zig is there for fun/airtime, he doesn't play NL cause you thinks he has an edge.

zig has no eyebrows

why does everyone hate negreanu?

On March 31 2009 19:18 Baal wrote: zig has no eyebrows

Take off every Zig.

On April 02 2009 04:44 Sliggy wrote: why does everyone hate negreanu?

I personnally appreciate the character.

On March 31 2009 19:18 Baal wrote: zig has no eyebrows

who needs them anyways

nice slowroll by eli with the AA, daniel keeps dissapointing.

On March 31 2009 14:28 Baal wrote: Same reason why the skill difference between GrandMasters is really small and for many many years it hasnt changed much, Bobby Fisher can play againast Kasparov... but watch Durrr rape Doyle's soul hardcore, its because we are close to chess perfection, and so far from poker perfection.

I know too little about poker game theory to have an opinion in this discussion, but this is a really weak argument. Chess has been studied for centuries and still super GMs have an enormous edge on regular GMs while at the nosebleeds edges are tiny after only a few years of poker being seriously analyzed.

It may be harder to develop a bot to play good poker than chess, I wouldn't know, but for the human brain, chess is definitely the much more complex game (and save for a few chess prodigies we are incredibly far away from playing perfect chess).

This comparision is really apples and oranges. Poker is a game of incomplete information and that makes designing AI for it fundamentally different from chess. I've mentioned it before but I think GO is a better comparision here.

Although GO, like chess is a game of complete information, the number of possibilites are so astronomical that it renders even the fastest supercomputers weak in evaluating every single possibility in the way that chess computers do. Even if you apply Moore's Law, the hope of getting a computer fast enough to computate the necessary calcutations to beat a pro in GO is unlikely in the near future. Therefore, you need to take a different approach where the computer actually thinks like a human rather than brute force. That is from a programming perspective much harder. That is the only approach that would allow a computer to beat a top pro in poker as well.

Baal, so while I agree with you that constructing AI that will beat a pro is infinitely harder for poker than chess, I think you are vastly underestimating the level of skill required in games such as chess an go. I have more respect for the "skill" of top GO pros such as Lee Sedol or Gu Li than I do for durrr. This is by no means putting down durrr in any way. Rather, it is my opinion, that if the entire world were given equal opportunies and training to be top poker players and top GO players, there would would be way more people of durrrs ability in poker than there would be of Lee Sedol's in GO. Obviously, this cannot be proven and is just my personal opinion. But I strongly believe that the skill set required to attain Lee Sedol's level in GO is rarer that the skill set required to attain durrr's level in poker.

It would be really cool if Giyom could ask Jimmy Cha this question.

the "bots" for chess were designed to beat the individual they were playing and not the masses, isn't that right???

god you guys are boring as fuck sometimes.