A new online poker art?

[eastbay]

Bot exploitation. Spotting bots, diagnosing their strategic flaws, and raping them for their roll before the author can figure out what happened.

It might come to this. It might even be more fun (for those who enjoy the puzzle solving aspect of the game) than real poker.

In its most potent form this would involve collusive strategizing. If the bot writers want to summarily declare that the rules don't apply to them, they may get more than they bargained for.

May you live in interesting times...

eastbay

[se2schul]

I think it would be fun to have Bot competitions. Have a few people write bots, then stick them in a room with small blinds and deep stacks for a few days and see who wins.

[ilya]

http://www.robotarena.com/

I'm not sure why the website isn't clear on this, but the game's all about designing your robot's AI.

[The Yugoslavian]

[ QUOTE ]
Bot exploitation. Spotting bots, diagnosing their strategic flaws, and raping them for their roll before the author can figure out what happened.

It might come to this. It might even be more fun (for those who enjoy the puzzle solving aspect of the game) than real poker.

In its most potent form this would involve collusive strategizing. If the bot writers want to summarily declare that the rules don't apply to them, they may get more than they bargained for.

May you live in interesting times...

eastbay

[/ QUOTE ]

Meh.

Yugoslav

[ilya]

Anyone know the game I'm trying to remember? You build robots and design their AI, then watch them fight it out? Lots of multiplayer competition online?

[spentrent]

AI Wars. I had one called "HALitose."

[dfscott]

There was also a thing way back when called CoreWars where you wrote programs in an psuedo-assembly language. They got loaded into the same memory space and each program tried to crash the other one.

[lastchance]

I don't think this is bot-specific. And it is very hard to really break someone in a SNG format.

[eastbay]

[ QUOTE ]
I don't think this is bot-specific. And it is very hard to really break someone in a SNG format.

[/ QUOTE ]

It's somewhat bot-specific because while a bot may consider a number of variables in its play, the strategy within those parameters is almost certainly fixed. It won't change up or close a hole when it starts losing unless the operator takes it down, diagnoses the problems (after almost certainly having not watched the games) reprograms it, and puts it back up. It can't adapt or stop playing quickly like a human player would.

With two players I think I could break saabpo. With only one it would be harder.

eastbay

[lastchance]

Eh.... The biggest problem, I believe, which adanthar touched upon is Party's structure.

A decent push/fold strategy combined with reasonable L1-L3 play can probably create a strategy that can't be exploited for much, unless you get major collusion, and maybe even then...

There's only so much you can do if the guy is mathematically correct.

And if the guy is running PT too, he could bust you for collusion after reviewing a few hand histories.

What I am saying here is, Adanthar's right. Many 2+2'ers have gotten greedy off Party Poker trying to create a huge $/hour.

Simply moving to PS or UB or any other site that has deeper stacks will easily combat bots.

PP is much too formulaic and programmable right now, IMHO.

[eastbay]

[ QUOTE ]

There's only so much you can do if the guy is mathematically correct.


[/ QUOTE ]

Right. But that's the whole point. He's not.

However, I agree with you that a preflop game is far more exploitable than a deep money game.


eastbay

[lastchance]

How long will it take for saabpo to become truly good?

Let's assume instead of creating SNG power tools, you were creating a bot right now. How good would that bot be (and remember, Party low-limits only have 800 chips, not 1,000)?

I don't think it's unreasonable to assume that given enough time (2-3 years), you, or someone like you could create a top tier SNG bot that could hold it's own against the best $33 has to offer, especially after L4.

[eastbay]

[ QUOTE ]
How long will it take for saabpo to become truly good?

Let's assume instead of creating SNG power tools, you were creating a bot right now. How good would that bot be (and remember, Party low-limits only have 800 chips, not 1,000)?

[/ QUOTE ]

Good point. /images/graemlins/tongue.gif

It's kind of unfortunate I have scruples about it. /images/graemlins/laugh.gif

eastbay

[lastchance]

So, how good are bots going to be? What do you think is the maximum exploitation rate you can expect out any reasonably good bot?

Given all this, do you think what you proposed in this thread is reasonable?

Right now, it's probably easy and reasonable to follow saabpo around, for as bad as Party players are, there are obvious flaws to it that you can exploit. But I don't expect those holes to be around for long.

[eastbay]

[ QUOTE ]
So, how good are bots going to be? What do you think is the maximum exploitation rate you can expect out any reasonably good bot?

Given all this, do you think what you proposed in this thread is reasonable?


[/ QUOTE ]

Yes. Most bot writers will suck, just like most poker players suck.

Given the logic of "you can't do that because others are capable of what you're capable of", I wouldn't play poker at all.

eastbay

[lastchance]

Well, if bots are going to suck, then knowing the bots that suck the most and exploiting them could be reasonably profitable. I do expect a bot writer to be able to write something that has +ROI though, so it's not too exploitable.

I'm just saying that before someone creates a bot, it's far likelier that a person would read up on how their bot should play as compared to someone who watched WPT and wants to gambool.

[eastbay]

[ QUOTE ]

I'm just saying that before someone creates a bot, it's far likelier that a person would read up on how their bot should play as compared to someone who watched WPT and wants to gambool.

[/ QUOTE ]

I certainly agree with that.

eastbay

[lastchance]

So you expect it to be something like this then?

2+2er>bot>Party fish? Because I still can't believe saabpo, with his numerous holes, is a winning player.

[eastbay]

[ QUOTE ]
So you expect it to be something like this then?

2+2er>bot>Party fish?

[/ QUOTE ]

Yep.

eastbay

[adanthar]

Disjointed stream of thought follows:

I don't think that's applicable because the average bot will last about a week or two before the account is shut down and all funds confiscated.

There might be two or three that will be winning and undetectable. That's fine, since they're going to be well kept secrets and almost indistinguishable from a decent multitabler anyhow - no real threat and they'll still have flaws/be exploitable. The rest will be your typical saabpo drones with a few improvements, like the pot odds call or 2 pair raise on the rare occasions they see a flop. Again, fine.

The far greater concern is what happens when the poker boom slacks off just a little and Party/Stars are infested with a few bots, a lot of players playing exactly like the bots, and not enough real fish. I have decided I need to get my postflop play up to speed and am working on sites where I can do that. Specifically, I'm thinking about the AJo hand; I advocated open pushing, which is probably best at Party and is exactly what a decent bot does - but this is a pretty damn basic hand that any halfway decent MTT player has got to know how to play!

Thinking about that, it occurs to me that I'd rather change sites right now and learn to deal with what happens when I suboptimally just raise it PF, rather than open pushing a 13 BB stack because the structure's forcing me to. I've just gotten an $1800 multitable win, mostly because of a lot of postflop theories I've gotten off here and the MTT forum; I'm confident I can do it again, but not if I spend the rest of the poker boom pushing that hand PF into other short stacks because it's 2% more profitable.

I think learning what to do with a 13 BB stack postflop in the long run is more +EV than this is.

[Big Limpin']

Yeah, giving up a tiny bit of profit, to get fee lessons in postflop play. How long before you're improved play wins you more, long run? I guess no that long.