(low content) NYT: Pokerbots will finish online poker

Drakar

Article on new york times which is pretty wishy washy on the details, but suggests that as online poker is mainly about analysing stats on previous play by players (a domain that fits neatly into bots feature sets) that their prevalence in online play will be unstoppable, and will eventually result in online games between bots and human chumps. Nothing really new in it though ...

http://freakonomics.blogs.nytimes.com/2007/11/12/poker-bots-on-the-rise-a-guest-blog/

Mellor

Nothing new in the article as you said, and in fact I think its is pretty por in general.

• Comparing checkers to poker, and the fact that a computer knows every perfect move in checkers. Fundamentally wrong, as checkers is transparent, full set of info is available and of course a computer is able to get it right. The unknown element of poker makes the difference here. The computer is operating on incomplete info, and because if this it is easy to throw off.
  
• The mention of Polaris, its true that a computer was built that challanged poker pros, but we are talking of a large purpose built computer, not a .exe that will run on a home PC. Its like comparing the AI in windows vista chess to "Deep Blue". Bare in mind that chess all info in chess is available too.

Scaremongering, meh

Wreck

"it is possible for bots to randomize their strategies and even hire individual humans to run them"

lol

ocallagh

I would be surprised if a well built program using home PC processing power could not beat the average 1/2 NL player

Drakar

"Those recruitment poker bots keep calling me, don't they realise I'm happy in my current job?"

RoundTower

I'm sure it makes a good read for some people.

Drakar

When you look at people who quite happily 8 table, that kind of poker looks like it could be replicated by a home pc pretty easilly.

ocallagh

to expand:

Provided the program was well built:

Positives:
Tilt would not be a factor.
Table selection would be perfect.
Multi Tabling Multi sites would provide a huge hands/hr figure. After rake, 0.5BB/hr on 100 tables would be the same as 10BB/hr on 5 tables.

Negatives:
The negatives are obvious. Making subtle adjustments to opponents changing styles would be difficult. People are unpredictable to a degree and trying to make a decision when we don't have all the information is not easy. The program would have a serious problem computing the correct decision in these situations. However, the same information that humans would base their decision on is also available to the computer and a well built program could get quite good at processing this information. Over time it could build profiles, spot inconsistencies, patterns, tilt factors, weaknesses etc... all the things that provide us with that gut feeling. I would guess the computer would not be too far behind an 8-tabling 1/2 grinder on auto pilot. Table selection would negate this to a degree too. If the computer consistently lost against player x, it would simply ignore player x.

RoundTower

If it was really so easy there would be lots of players doing it. No question there are a lot of people trying to do it. However:

Even though no limit holdem is by far the most common poker game played online, there haven't been any big bot rings busted that I know of. Unlike in limit hold'em, SNGs, and PLO. So it's safe to say that while it might be happening in isolated spots, it's not widespread, and therefore it's not as easy as you think. (And I also think I could write a program to do it, it just seems it's harder in practice than we all intuitively think).

ocallagh

I've thought about that too.

The program would be incredibly hard to develop. It could potentially take a team of 6 top programmers months to develop. Say 3 months, €50/hr * 6 * 7.25 * 20 * 6 + hardware costs. Taking these incredibly rough figures €125k would not be too far off to develop it from scratch. If someone went to the trouble of spending €125k on something fairly illegal you can be sure they will go to great lengths to conceal it. I reckon there is every chance this has been done and unlike those Absolute idiots, their operating model is up to scratch!

ianmc38

i think it could definitely be done.

dvdfan

As far as i know one of the poker sites had a program that pops up sparidically and asks the user to confirm the letters on the screen like you see when your signing up for lots of things these days to prevent bot submissions.

I think this is a good idea as it would force them to have someone watching over the robot/s which would reduce the profitability although it would be easy for someone to employ someone to do this mediocre task from somewhere in Asia for little or nothing however it would stop the amateurs.

And it wouldnt be much of an inconvenience if this was to happen every now and then, as it is alot of sites already use pop ups to announce tournaments and some i might say are very annoying as the keep popping up every 2-5 mins for ages but if it was once every 1-3 hours i would be fine with that. Of course you wouldnt want it that when it pop's up it stops you playing if you dont get the code right as some people could have genuine problems but support should contact you straight away to see why you didnt respond.

I know theres ways around this too but with this in place and 1 or 2 other security checks it will at least make it hard for them to exist. Another option would be the cross co-operation of security poker sites to allow to cross check people with the same address/ip address/credit card to see if they are playing a large amount of tables at the same times.

Its in the interest of the poker sites to show their at least attempting to fight against these bots as credibility is important and storys like the full tilt bots etc can be damaging to these companies.

Kamaldihnio

What I wanna know is how would it know when to bluff and when not to bluff?

CaptainNemo

I can sort of picture how it would be developed but agree it would be incredibly complex. As soon as I start thinking about # of scenarios, # of players involved, player stats, possible reactions to those stats, bluffing frequencies, raise/call/fold decisions, bet sizing, etc, my head starts to hurt.

A good way to approach it, rather than to program a "perfect" poker playing program, would be a heuristic program where certain configurable options are open for the program to modify itself based on past performance (e.g. it sees that certain actions in certain situations are -EV over a large sample sizes and experiments with different actions). The program would then be let loose to "evolve" in a live environment. I expect based on their particular experiences some programs would evolve into TAGs, some into LAGs, etc...but probably eventually all would gravitate towards some sort of optimal strategy determined by interaction with the majority of players in the pool in which they play. Would be interesting to see what would happen when the bots played each other.

In practice I think it would cost so much to develop this program and buy the hardware on which to run it that it could not possibly repay its cost without swamping the networks with bots, something that would surely be easily detectable. It would be a very interesting project theoretically but not sure how useful commercially/criminally.

Imposter

I think it would be easier/cheaper and probably more profitable to build a less sophisticated bot (but still not that unsophisticated) and collude by running multiple instances on the same table. If it's thought out properly there's no reason why they should get caught either.

paul8200

It probably will be done some day. But it's incredibly difficult. Programmers are an optimistic bunch when it comes to time and cost estimation.

The complexity in the rules and the testing of all those permutations means that it will never be perfect. Far simpler IT projects have been poorly thought out by 'good' people.

Plus it's cheating, so what kind of people are you going to attract to your hypothetical venture?

White Knight

ianmc38 wrote: »

i think it could definitely be done.

From HH's you have posted, I am convinced you know it can be done :eek:

Mellor

paul8200 wrote: »

Plus it's cheating, so what kind of people are you going to attract to your hypothetical venture?

Its online cheating to use them online, a team of people (such as a universities) would build them to see if it can be done. A number of chess bots were built in the 90s for play against top human players, the poker bot could be approached from the same moral standpoint. As soon as it is proved possible, it would become a more popular pet project, details leaked etc

ianmc38

White Knight wrote: »

From HH's you have posted, I am convinced you know it can be done :eek:

Lol.

When I was in uni, we did several units on AI and the likes. One of my friends completed an insanely good chess bot within 3 months of programming. Now obviously chess is different because it uses different search algorithms that can obviously work out every possible move from a player to n number of moves in front, but I think a good poker player and a good programmer could easily build a bot to beat SSNL within 3 months.

strewelpeter

Look at it like this, a Tollbooth is a bot.
The reason that there are still people working as toll booth operators is that it remains more economically viable to put a person in the middle of the bot performing some mechanical and a little computing that could just as easily be performed by a machine. It is only a matter of time before it becomes cheaper to remove the human.
The same is true in many fields including many very sophisticated tasks in fields like medicine and engineering that people thought unimaginable even ten years ago.
I would argue that an online poker player multitabling using PT PHUD mined data etc. is not very different from a Toll booth operator - performing a pretty mundane task albeit one that requires decision making, that only exists because it is, so far, uneconomical to replace their function with a tool that performs those functions.
Whether it will ever become economical to produce such a tool is questionable in as much as once such a tool becomes generally available it will kill the goose that laid the golden egg as the humans leave and the bots lose their value proposition.Its very unlikely that that would stop people trying. People will still play their online poker on sites probably using webcams etc.where they can believe they are not playing bots.

This idea that NLHE is soooo much more sophisticated than limit and no bot could ever be produced is just silly. They said the exact thing at the start of just about everything that bots have fitted into and then changed to exclude some exceptions (like limit hold em) when the bots came along and pawned them.

Killme00

I think it would be more difficult to do for NL than Limit but not exponentially so. The variables are limited regardless of whether its Limit or No Limit. Its quite a simple programme to choose option A or B or C based on different variables. I know that a villians decision is not set but Pot Odds, Pot equity and Hand strenght are set variables preflop as well as on the Flop, Turn and River. More accruate and profitable decison can be made by using hand histories to look at each villians range and actions much in the way we do as humans. Beating the security at online poker sites would be easy enough through random timing of bets from 2 seconds to 25 seconds, randomisation of bet sizes but still related to pot equity, and its also possible to have the programme type "idiot" or "luckbox" when it loses a hand every so often. Obviously it would take time to develop and test but it is not impossible.

The problem lies in the programmes ability to learn and evolve.

Drakar

Soon afterwards they'd become self aware and then it's Terminator territory all over again.

ionapaul

Drakar wrote: »

Soon afterwards they'd become self aware and then it's Terminator territory all over again.

and then they spent your 'hard-earned' profit on bot porn or importing bot mail-order brides from Japan...damn those free-spending self-aware bots, damn them all to hell!