Article Request: Analysis of Bots

[jjshabado]

[ QUOTE ]
The largest problem with bots, as i see it, is card sharing.

Probably its the bot-programmers own bots who gets to share the cards of the bots ppl downloaded from his site. This will give his bot an unfair dissadvantage over both other players and other bots. The cardsharing ability is of cause why the bot-programmers wants their bots to be downloaded - the more bots that share cards with the programmers bots - the more he will make.

The main dissadvantage with bots are that they tend to be predictable. this will make the better players win in the long run (if not playing agains several cardsharing bots). And many of the low limit players are in it for the thrill. They dont care if they win or loose, and if they win or loose to a semi-pro or a bot.

This may even be of advantage for onlinepoker. Bots, ensuring that games are always available for recreational players on low limits.

[/ QUOTE ]

Bots will get better. Its only a matter of time before they're good enough to beat the vast majority of players. I suspect its also only a matter of time before they're unbeatable by all players.

Anybody that assumes bots will stay simple and predictable is ignoring the past 5 decades of computer history. There was also a good thread in Computer Technical Help about bots:

Bot Thread

[dwf76]

Can't say I know much about bots but how about taking any possible article a bit further for us less knowledgable players, for example.

1. What poker rooms are doing to protect us.
2. How can we detect them?
3. Is there a strategy to defeat them?

[Grasshopp3r]

There are counter-measures for bots. Strategies to defeat them should be handled very carefully so that the bot makers can not anticipate them.

Bots are consistent, which is how they are exploited and is also how they make money. The standard play can be +EV long term with a bot and they can calculate those odds. Does that mean that they will never hit a downswing and go busto? Obviously, the closer that they play to +EV, the more risk of ruin, unless they adjust for that, somehow.

[proud_botter]

I'd be happy to write one if anyone is interested.

I can go into details on:

The underground subculture of botters
How bots are programmed
Detection vs anti-detection
Limit vs No Limit
Creating Multiple accounts
Toughest sites to bot on
Commercial vs non-Commercial bot software

[dwf76]

proud botter,

I'm assuming by your name and post you have either built a bot or have the know how to do so, and from a detecting and strategy against point of view I would love to read anything your willing to share.

There clearly is a lot of useful information within these few pages and it seems a shame to waste it debating whether to have an article or not, so I have my fingers crossed the Magazine decides not only to do the article but to have the common sense to consult posters in this thread and make use of their knowledge.

[RedManPlus]

I'm both a software engineer and Pro Trader...
And am successfully running Automated Tradings Systems on the NYSE.

The spread of Poker Bots...
Is somewhat analogous to explosion in "algorithmic" or "black box" trading on stock exchanges...
Where Trading Bots now account for > 50% of all volume.

So what has been the effect?

Not much...
Algorithmic trading infrastructure is VERY expensive...
And, imo, the "black boxes" are just playing a Zero Sum Game among themselves...
Where 10% of the "black boxes" win and 90% of the "black boxes" lose...
But the ULTRA HIGH COST has created a sort of benign equilibrium.

A key problem is the state of leading edge AI.

Current AI cannot come anywhere close to replicating the high level, multi-faceted decision making of a Pro Trader. So the most successful "black boxes" are actually manned and "overidden" by Pro Traders that cost, say, 100K/year. And the guys coding the software make > 100K/year on Wall Street.

Poker is different...

But it's KEY to understand...
That Bots can be severely limited by applying simple "prove you are human" techiques we are all familiar with. This inconveniences customers and scares them... so Poker Sites have chosen the correct financial strategy: ignore Bots and enjoy the rake they produce.

But once things evolve and there is an actual clamp down on Bots... Good-bye to simple, amateur Bots. Only the Top Bots designed to be manned by a Poker Pro will be viable... and will be coded by guys that would be making > 100K/year on Wall Street.

Is there really a big difference between:

(a) 4 Pros versus 6 Donks at a table

OR

(b) 2 Very Expensive Pro-manned Bots + 2 Pros vesus 6 Donks at a table ???

Not really. Everybody relax. Nothing really changes.

[Sniper]

[ QUOTE ]

Is there really a big difference between:

(a) 4 Pros versus 6 Donks at a table

OR

(b) 2 Very Expensive Pro-manned Bots + 2 Pros vesus 6 Donks at a table ???

Not really. Everybody relax. Nothing really changes.


[/ QUOTE ]

This analysis is incorrect, of course things change, or no one would be spending the time to develop complex bots.

Same with the Stock Market example...

[jjshabado]

One other difference is that the simple interface that poker clients have to their user will make bots extremely hard to detect.

I'm with Sniper, its extremely stupid to assume they're not going to improve. People use to say the same thing about chess software. While poker is a lot harder to develop AI for, its still not impossible. To repeat a line I use a lot when talking about bots: to say they'll never improve to the point of being a danger to online 'pros' is ignoring the complete history of computer science.

[RedManPlus]

I think my post was misunderstood...
"Nothing really changes" was just a general point about the world.

Cliff Notes version:

(1) The very high cost of Top Bots will moderate their impact.

(2) Top Bots MUST be manned by Poker Pros... because today's AI cannot handle the 5-10% most subtle poker decisions.

(3) Poker sites have exactly 2 choices:

(a) scare the sh*t out of customers by admitting heavy Bot penetration and instituting intrusive tests to "prove humanity"

OR

(b) The obvious choice: lie and count the rake

(4) Poker Bots are here to stay. The Poker World is like the stock market in 1965... when I'm sure computer trading was viewed as akin to "cheating". Today 50-70% computer trading.


And in reply to jjshabado...
As a Computer Science grad I have been following AI trends since the late 1970s. No field have ever overrated their potential advancement more than the AI field. Ever.

Virtually none of the great things promised have become part of everyday life. In 2007, AI cannot even replace a telephone operator.

Like that Full Tilt Bot operation that was discussed here in May. What a ridiculously primitive setup. Lay people actually think this is sophisticated. My trading operation is more sophisticated... and I'm nowhere near leading edge where < 1 ms latency is the standard.

In truth, for the truly talented software engineers that chose fraud and their sponsors... beating poker sites is penny ante stuff... and way down the list of lucrative targets.

[jjshabado]

As a Computer Science grad I know that this "No field have ever overrated their potential advancement more than the AI field. Ever." is a stupid statement.

[ QUOTE ]
Virtually none of the great things promised have become part of everyday life.

[/ QUOTE ]

I'm not talking about the flying-car type promises made 30 years ago. I'm talking about the fact that there has been steady and significant progress in the field of AI game playing for the past 2 or 3 decades. There is a lot of very promising and interesting research being done in AI now. To think that this progress is just going to stop is naive. I'm not talking about the next 2 or 3 years. I'm talking about the next 5-20 years. Maybe it'll be on the near end, maybe the far, but either way Poker AI will be advanced enough to beat the vast majority of online players.

[ QUOTE ]
. In 2007, AI cannot even replace a telephone operator.

[/ QUOTE ]

Actually, its pretty close. And the fact that you think this is a relevant example shows that you really haven't thought too much about bots. Why is replacing a telephone operator with a machine hard? Because of the interface. The actual job being performed is trivial for a computer. Its making machines process/decode/replicate the human voice and language that is hard. A Poker bot needs a very narrow interface. It just needs to be able to get its cards, the board cards, and the action. All of this needs to be presented to the human users in a reasonable way ( who's going to play poker at a site that uses squiggles everywhere to obscure suit values? ) and that means its not that hard for a bot to get that information for processing.