Bots and the future of online poker

[shinigami]

<= not worried.

internet poker room patriot act, ldo. If bots become a real problem they'll make you install and run a program that acts like a key logger / screen hook / connexion (ethernet, bluetooth, wifi, younameit...) sniffer. good luck running a bot with that....at least you'll still be able to jerk off while playing, as some ppl seems concerned about that...

[MaxCaliber]

As an IT professional I think it is possible to program really good bots (even NL) if they use their hand database well. However it requires quite a long time to do it.
And as it was mentioned earlier there are some easy solutions to prevent bots playing. For example in every 5 minutes a small dialog pops up and you have to type in a code that is shown in a picture (as it is on a lot of registration forms).

So I would think it is risky to invest in such a project.

However this is based on the assumption that the sites are willing to use such bot-prevention solutions.

[Victor]

"
For instance, you steal-raise from the button, and the BB calls. The flop comes, he checkraises. What does he have? Well maybe that depends on whether the flop is K62 rainbow or 987 single-suited. If the former, then maybe he is using a different rule for checkraising. Maybe his range depends on the fact that he is estimating what your range is, and he thinks he can blow you off of a hand on the dry flop. So it's not enough to merely enumerate hands, say that he'll checkraise on any pair or draw, and estimate your equity. You need to do something more."

im not sure lhe works this way. u usually have enough outs to peel with overs. flop bets are pretty small in relation to pot size so one over and bd draw is enough to peel with a lot.

"I've told the story before of playing a bot at the AAAI competition, "

how good was the guy who wrote the bot? did he have access and incorporate to pokerstove?

[disjunction]

[ QUOTE ]


im not sure lhe works this way. u usually have enough outs to peel with overs. flop bets are pretty small in relation to pot size so one over and bd draw is enough to peel with a lot.

[/ QUOTE ]

Ok, but there are two possible ways to program the bot. We can make rules, such as "We have two overs, call". But you run into problems based on rules alone, because overs are worth less when they complete someone else's draw, and on two-tone boards, etc. Another example of this is an straight draw. A 1-cards straight draw is worse than two cards, but if your card is a 4, that is rarer and therefore better.

So you can take the second approach, estimate your opponents range and your equity. But I was describing the flaws of that in the last approach. That approach doesn't say anything like "You have overs". It says "You have equity to continue against his range". But overs getting 7.5:1 on a two-tone flop often don't.

[ QUOTE ]

how good was the guy who wrote the bot? did he have access and incorporate to pokerstove?

[/ QUOTE ]

I don't know, I didn't talk to him. When I talk to AI people, I generally hide the fact that I know anything about poker. But the impression I get is that nobody who works on this knows a thing about poker. Certainly the professors don't, I'm sure maybe a few grad students do, but I've never met one that would pass for a twoplustwo'er. Like I said, I haven't paid much attention to the poker domain (I'm hoping to read some papers this summer), but the impression that I get is that they just throw away information before they even start which would just make you cringe.

[wpr101]

[ QUOTE ]
It wont develop reads. So no, it won't. Despite the fact that it won't tilt.

[/ QUOTE ]

You are so wrong; yet you have no idea why.

[wpr101]

Also, it's amazing how ignorant some people in this thread are about computer programming.

[RobAtticus]

[ QUOTE ]

For instance, you steal-raise from the button, and the BB calls. The flop comes, he checkraises. What does he have? Well maybe that depends on whether the flop is K62 rainbow or 987 single-suited. If the former, then maybe he is using a different rule for checkraising. Maybe his range depends on the fact that he is estimating what your range is, and he thinks he can blow you off of a hand on the dry flop. So it's not enough to merely enumerate hands, say that he'll checkraise on any pair or draw, and estimate your equity. You need to do something more. It's easy to screw up, and I don't know if it's possible to do a great job. I've told the story before of playing a bot at the AAAI competition, and I could bluff-raise it with air on any dry board. I changed my range so that its computations were useless. If the programmer fixed this, he might introduce a new problem.


[/ QUOTE ]

The problem is you're assuming that bots will run on hard and fast rules, which is not necessarily the case. A bot will have access to all sorts of databases that helps it evaluate the decision. It can look at past hands with this player and find similar situations, and it can look at players with similar playing styles and look at similar situations. It can find out how the hands played out, figure out a range, and run the same EV calculations that humans run (but usually AFTER the hand is over). If it finds that its odds are that its usually behind and doesn't have a good drawing percentage, it can fold; similarly, it can figure conditions on which to raise and call.

Bots can eventually do the same analysis that you do -- defining a range and figuring out how to act against it. Bots can learn and adapt, and they can do it better and faster than most players.

[disjunction]

[ QUOTE ]

The problem is you're assuming that bots will run on hard and fast rules, which is not necessarily the case. A bot will have access to all sorts of databases that helps it evaluate the decision. It can look at past hands with this player and find similar situations, and it can look at players with similar playing styles and look at similar situations. It can find out how the hands played out, figure out a range, and run the same EV calculations that humans run (but usually AFTER the hand is over). If it finds that its odds are that its usually behind and doesn't have a good drawing percentage, it can fold; similarly, it can figure conditions on which to raise and call.

Bots can eventually do the same analysis that you do -- defining a range and figuring out how to act against it. Bots can learn and adapt, and they can do it better and faster than most players.

[/ QUOTE ]

But the use of these databases is just more hard and fast rules.

I know what you are saying, but I guess my claim is that what you are saying is hand-wavy. What does "similar situation" mean? Is a button steal versus blind a similar situation to a cold-call of a preflop raise? Is a checkraise on a K82 rainbow flop a similar situation to a K82 2-tone flop? Should the bot adjust to my general stats, or how I play against it specifically? Either way, I think I can defeat it. My claim is that the everyday leveling that goes through your head is more than can be encapsulated in a static algorithm.

[vmacosta]

Disjunction,

How much experience do you really have with a specifically k82r board. My quick (possibly wrong) calc shows that flop coming up 171 times every 10^6 hands.

There's just no way you will have more experience than a bot with a gigantic database.

That said, its not really an easy task because popular lines change with times so the bot's sample size is never gonna be that huge either (though obv bigger than a human's).

Victor,
I don't know anything about you and perhaps you are a freak computer genius, but my guess is you are overestimating your ability to program a bot good enough to beat you. Even with access to every hand played on the internet the code will be very extensive and will be easy to f up.

[Victor]

"perhaps you are a freak computer genius,"

im not, but i have access to a few.