saabpo: Bot or Joke?

[GrannyMae]

[ QUOTE ]
I think we should set up a 2+2 NL game and invite this Sappo

[/ QUOTE ]

if he is a monitored bot, the name will be changed soon, so you will never find him again. wait until the bot owner sees the HH's to check his 'baby' and sees all that chat. name change or search hide for sure.

[krazyace5]

Id like to watch this bot, does anyone know if he is on right now?

[GrannyMae]

[ QUOTE ]
Id like to watch this bot, does anyone know if he is on right now?

[/ QUOTE ]
not showing as on. so he is either hidden from searches, did a name change, or is being oiled and lubed.

browse the $20+2 nlhe SnG's. that is where he was yesterday.

[gomberg]

Wow - all you guys who think a poker bot is easier than a chess-bot are just plain wrong. Do you all have any experience programming AI or know the relevant modern research into the area? Chess is EASY compared to poker. It's not even close. I don't have time to explain exactly why, but if you want to do the research before making these rediculous arguments, read most of the papers from the university of alberta games poker research group. Also go to the poki-poker forums and read discussions people have on the AI forums.

[HopeydaFish]

[ QUOTE ]


It is not even a question that computer can be programmed to beat any human in the world in poker. For limit games it's even probably possible today. If all the effort that went into chess programming (which was considered one of the greatest AI challenges) went into poker, we wouldn't even have this discussion right now.

The only reason we don't see bots overrunning limit games right now is skills required to write it. Luckily, most people who possess them have better things to do.

[/ QUOTE ]

I agree totally with your statements concerning the feasibility of poker bots. I'm not an 'expert' programmer, but I have programmed automated web testing tools in the past (I do programming for web applications now). The interface in your typical poker program isn't at all different than the programs that I've created automated scripts for in the past. It would be a little more complicated, but there's nothing really in the interface itself to stop someone from creating a totally automated bot that could check, fold, and raise based on certain criteria.

It wouldn't have to play *perfect* poker, just *winning* poker for it to be profitable. Creating a "winning" bot would be simply a matter of trial and error. I'd start with Sklansky's SSHE system and then continuously tweak it from there every day to adjust my winrate. Once I'd found a system that consistently produced winning results, it would simply be a matter of letting the bot run. You'd only want it to run for a reasonable amount of time every day on any particular site (not like the idiot with saabpo, who let it run 24 hours straight). You could get fancy by having it say "nh" or "arg" or "lol" or what have you ocassionally (but not always) when it gets beaten on the river. Keep in mind that I'd probably be shooting for the 1/2 or 2/4 tables -- higher stakes players would be more suspicious of the bot and notice its betting patterns much quicker.

What stops me from doing this (aside from ethics), is lack of free time to engage in such a project, as well as a suspicion that by the time I had something working, Party would have stepped up their detection methods and I would see it banned before making any real money -- making the whole thing a total waste of time.

[HopeydaFish]

[ QUOTE ]
Wow - all you guys who think a poker bot is easier than a chess-bot are just plain wrong. Do you all have any experience programming AI or know the relevant modern research into the area? Chess is EASY compared to poker. It's not even close. I don't have time to explain exactly why, but if you want to do the research before making these rediculous arguments, read most of the papers from the university of alberta games poker research group. Also go to the poki-poker forums and read discussions people have on the AI forums.

[/ QUOTE ]

I think where some people get sidetracked is the difference between programming a "winning" bot, and one that plays perfectly and can beat anybody. I'm a mediocre chessplayer but can beat the average human opponent because I know more about chess strategy than someone who knows just the basic rules. However, I can't beat any of the chess programs out there. However, an expert chess player can always beat me and most chess programs quite easily.

What does this mean poker-wise? Well, think of the players playing 1/2 as being equivalent to the chess players that *I* can beat as a mediocre chess player. They know a little strategy, but basically all they *really* know is the basic rules of the game (that's why you see 50-70% of them in on the flop consistently). These are the players I'd want my bot to play against. My bot isn't capable of beating the "Grand Masters" -- so I keep it out of those games and have it playing against players it can beat.

It would just be a matter of tweaking my bot to fine tune it to grind out winning sessions. Once I have it grinding out 1BB/100 I would have it multi-tabling and playing on multiple sites to turn my 1BB/100 into decent money.

[NiceCatch]

Ok, I briefly read through the AI/Bot Chess/Poker posts, and wanted to throw in my .02.

Chess is a game of complete information. The REASON a computer can defeat a human being is the number of moves in advance that a computer can compute; i.e. there is a finite number of moves that a chess player can make on a given turn, and another finite set given any of the possible moves the player makes. What makes a computer so perfectly suited to play chess is that it can look 15, 17, 20 moves in the future and look at ALL the possible combinations of moves. This is literally BILLIONS of possible scenarios. (I don't know the exact numbers, but that's the main idea.) Knowing all of these different scenarios, it can choose the most optimal move. And it can do this analysis on every move.

Clearly, no human being can do this.

Poker is entirely different in that as a poker player, you pretty much have to learn the styles of how everyone else is playing. Learning, as opposed to looking at different scenarios, is in the realm of artificial intelligence, NOT supercomputing. This is an ENTIRELY different ball game.

Here's what I believe about poker-playing bots.
A bot could be created that plays optimal strategy at Limit, and might just be able to pull down a couple of big bets and hour. This is assuming that it could adjust to the style of play at the table. I do believe this is possible.

A bot could be created to do well at small-stakes No Limit. It's strategy would be similar to a nut peddler's. I think this type of bot would be reasonably successful, ASSUMING the other players DIDN'T know it was a bot. If they did, and were smart enough to take advantage of this, the bot would probably be unprofitable, unless it adjusted it's strategy somehow.

At high-stakes No Limit, I don't believe the AI will be available for a very long time to allow a computer to compete with the top pros. You can even run an experiment. Play by an exact algorithm, regardless of who you're playing with. Practice that method of play at a small-stakes game. Optimize it. Then move up to high-stakes NL games and try it out. I believe any method you choose that does not adjust over time (large amounts of time or small, such as during the course of a few hands) will get owned.

I think a huge point is, if you know you're playing against a bot, that knowledge should easily be enough to defeat the bot. If I ever find a player that consistently plays a style in no-limit, it's generally very easy to find a method that will crack them. You guys have pretty much proved it with your analysis of Saabpo.

One thing I will say is, the most likely people to be doing research on this right now are criminals. There is obviously money to be made, and if this technology does come out, criminals will take full advantage of it. You all might want to work on your live-play skills!

[gomberg]

good post - creating a good poker bot is good for science as well because solving a game of imperfect information is much more realistic of problems occuring in real life - so methods used to "solve" poker could be greatly used in industry as well. These bots are also being created to help teach poker as well...

[skoal2k4]

Ever since I started reading these forums, the thought of bots playing online brought instant laughs from most people, usually with a remark along the lines of "bring on the bots! There's no way they can beat me!"... It seems from some of the posts now concerning saabpo, the tone is a little different from some (not everyone). Some people actually seem scared. I'm just curious about what brought about the change?

[FishBurger]

Saapbo looked like he was an overall winner and placed third (ITM) in at least one table where there were 4 other brown trout. He looked like proof that bots can win at the lower buyin tourneys.