BOTs

[Evan]

[ QUOTE ]
[ QUOTE ]

This is almost certainly untrue.

[/ QUOTE ]

I wouldn't doubt that you are right. Though I admire Chanos's "courage" for trying to make money as a short, I think I've read more than one Chanos analysis and thought he was misinterpreting some facts to make his argument. I wouldn't doubt he heard a juicy rumour and decided to run with it.

I myself would never pay people to play online when it's so easy to program a computer to do so for free.

[/ QUOTE ]
I have no idea how easy it is to program a computer to win at poker. If we're talking about games for any meaningful amount of money I would guess it's hard, but I'm no expert.

My reasoning was that whatever you're paying these people to play for you is not worth it for them if they're good enought hat you'd be willing to pay them. It's not that tough to scrape together a grand to play 2/4. If you're decent at multitabling you could probably make $40+/hour and move up pretty quick. No way playing for someone else is hte besst deal.

[Sniper]

[ QUOTE ]
I have no idea how easy it is to program a computer to win at poker. If we're talking about games for any meaningful amount of money I would guess it's hard, but I'm no expert.


[/ QUOTE ]

For a competent programmer, it's really not that difficult to program a bot to play ABC poker... add on datamining and an opponent modeling module, and you have a strong player, at least at low limits.

[Evan]

[ QUOTE ]
[ QUOTE ]
I have no idea how easy it is to program a computer to win at poker. If we're talking about games for any meaningful amount of money I would guess it's hard, but I'm no expert.


[/ QUOTE ]

For a competent programmer, it's really not that difficult to program a bot to play ABC poker... add on datamining and an opponent modeling module, and you have a strong player, at least at low limits.

[/ QUOTE ]
What kind of low limits are we talking about? And what is the theoretical win rate of these things? I don't know the answers, but I have some conjectures just based on what I know about poker.

I have a hard time believing it's "not that difficult". Have you (or anyone) made one of these bots? I'll play it heads up if you want to test it. I know there was some software available that had various bots. I used it for a bit and the bots definitely weren't great. There was one that was supposed to be nearly unbeatable heads up, which is I guess how bots do best. Anyway, I doubt that whoever made that would call it easy, and I really doubt that those bots could beat a 10 handed game above .5/1 (if that).

[DesertCat]

[ QUOTE ]

I have a hard time believing it's "not that difficult". Have you (or anyone) made one of these bots? I'll play it heads up if you want to test it. I know there was some software available that had various bots. I used it for a bit and the bots definitely weren't great. There was one that was supposed to be nearly unbeatable heads up, which is I guess how bots do best. Anyway, I doubt that whoever made that would call it easy, and I really doubt that those bots could beat a 10 handed game above .5/1 (if that).

[/ QUOTE ]

I haven't written a bot but I've developed a lot of software and played a lot of poker. And I think building a bot that can beat games up to 5-10 online would not be very hard. Remember, the bot doesn't have to beat you, it just has to beat the plethora of bad players online.

Consider a bot that is a "rock", it only plays premium hands and plays them aggressively. It keeps a database on every player it's ever played with, and can quickly calculate a hand range for that player in any situation (hands he raises with preflop, on the flop, checkraises the turn with, etc). In many cases it would be better able to predict the relative likelyhood of being ahead, or behind to an opponent, than you could, due to it's perfect memory and instantaneous calculations.

I don't doubt it would be very difficult to make the bot play well at high limits, or to make subtle moves, but beat low limits online? Doesn't seem very hard to me. After all, I do it. And I have huge holes in my game.

[Evan]

[ QUOTE ]
[ QUOTE ]

I have a hard time believing it's "not that difficult". Have you (or anyone) made one of these bots? I'll play it heads up if you want to test it. I know there was some software available that had various bots. I used it for a bit and the bots definitely weren't great. There was one that was supposed to be nearly unbeatable heads up, which is I guess how bots do best. Anyway, I doubt that whoever made that would call it easy, and I really doubt that those bots could beat a 10 handed game above .5/1 (if that).

[/ QUOTE ]

I haven't written a bot but I've developed a lot of software and played a lot of poker. And I think building a bot that can beat games up to 5-10 online would not be very hard. Remember, the bot doesn't have to beat you, it just has to beat the plethora of bad players online.

Consider a bot that is a "rock", it only plays premium hands and plays them aggressively. It keeps a database on every player it's ever played with, and can quickly calculate a hand range for that player in any situation (hands he raises with preflop, on the flop, checkraises the turn with, etc). In many cases it would be better able to predict the relative likelyhood of being ahead, or behind to an opponent, than you could, due to it's perfect memory and instantaneous calculations.

I don't doubt it would be very difficult to make the bot play well at high limits, or to make subtle moves, but beat low limits online? Doesn't seem very hard to me. After all, I do it. And I have huge holes in my game.

[/ QUOTE ]
If you had to ballpark it, how many hours would it take to build it?

Note: I have neither interest in building a bot nor the ability to do it. I'm just curious.

[krishanleong]

[ QUOTE ]
[ QUOTE ]
[ QUOTE ]

I have a hard time believing it's "not that difficult". Have you (or anyone) made one of these bots? I'll play it heads up if you want to test it. I know there was some software available that had various bots. I used it for a bit and the bots definitely weren't great. There was one that was supposed to be nearly unbeatable heads up, which is I guess how bots do best. Anyway, I doubt that whoever made that would call it easy, and I really doubt that those bots could beat a 10 handed game above .5/1 (if that).

[/ QUOTE ]

I haven't written a bot but I've developed a lot of software and played a lot of poker. And I think building a bot that can beat games up to 5-10 online would not be very hard. Remember, the bot doesn't have to beat you, it just has to beat the plethora of bad players online.

Consider a bot that is a "rock", it only plays premium hands and plays them aggressively. It keeps a database on every player it's ever played with, and can quickly calculate a hand range for that player in any situation (hands he raises with preflop, on the flop, checkraises the turn with, etc). In many cases it would be better able to predict the relative likelyhood of being ahead, or behind to an opponent, than you could, due to it's perfect memory and instantaneous calculations.

I don't doubt it would be very difficult to make the bot play well at high limits, or to make subtle moves, but beat low limits online? Doesn't seem very hard to me. After all, I do it. And I have huge holes in my game.

[/ QUOTE ]
If you had to ballpark it, how many hours would it take to build it?

Note: I have neither interest in building a bot nor the ability to do it. I'm just curious.

[/ QUOTE ]

I'm a programmer but don't have the programming expertise to make a bot (windows side of things). That said, I think I have enough poker knowledge to build one that could beat 3/6 short for 1 BB / 100.

Krishan

[DesertCat]

[ QUOTE ]

If you had to ballpark it, how many hours would it take to build it?

Note: I have neither interest in building a bot nor the ability to do it. I'm just curious.

[/ QUOTE ]

It's hard to estimate accurately without doing any specific research and writing a schedule. But I'd think it's months, not years, to get something going. I'd think 3 months would be more than enough. As Krishan noted, the windows side offers some challenges, but I've written windows application control programs before.

[Sniper]

[ QUOTE ]
I have a hard time believing it's "not that difficult".

[/ QUOTE ]

A good programmer can get a computer to "think" quite a bit... and ABC poker isn't difficult.

There is no reason why (with the right programming skill) you couldn't program a bot to play almost the same as you would, were you sitting at the table!

Being the Stock Market forum, its worth noting that trading, like poker, is a game of incomplete information... and computer programs are responsible for the vast majority of trading volume every day!

[ QUOTE ]
Have you (or anyone) made one of these bots? I'll play it heads up if you want to test it. I know there was some software available that had various bots. I used it for a bit and the bots definitely weren't great. There was one that was supposed to be nearly unbeatable heads up, which is I guess how bots do best.

[/ QUOTE ]

If you want to test the capability of bots, pick up a copy of PAP2, and play around with the bot settings! They specifically are working on creating a bot that will play HU "optimally". It's worth noting that to make a profit at low limits (up to 5/10 Full), it's not necessary to play optimally!

A successful bot, would have 3 main components: Automation, Decision/Analysis, Opponent modeling. The first two components are readily available, and significant information is available on design work that has gone on with regard to opponent modeling.

Its worth noting that the poker rooms have a problem when a bot is automated, but some have gone so far as to say that Decision/Analysis programs are "approved". So if you wanted, you could download one of them and play around with the decision matrix and test for yourself!

I also agree with everything Cat said!

[PickyTooth]

It would be a whole lot easier to make a bot beat sngs instead of ring games.

[Evan]

[ QUOTE ]
It would be a whole lot easier to make a bot beat sngs instead of ring games.

[/ QUOTE ]
This I can believe.

[Sniper]

[ QUOTE ]
It would be a whole lot easier to make a bot beat sngs instead of ring games.

[/ QUOTE ]

I don't play enough SNGs to really have an opinion... although the consensus seems to be that programming a bot to play NL is harder than Limit.

[Evan]

[ QUOTE ]
[ QUOTE ]
It would be a whole lot easier to make a bot beat sngs instead of ring games.

[/ QUOTE ]

I don't play enough SNGs to really have an opinion... although the consensus seems to be that programming a bot to play NL is harder than Limit.

[/ QUOTE ]
Learning to play one or two streets is much easier than learning to play four.