Thursday, November 18, 2004

Fear of a Black Planet

YTD: +$39399.12

I’ve been taking a cursory look at other people’s blogs again. By and large, with some noticeable exceptions, I don’t really enjoy them. This is mostly because they are about life, the universe and everything 80% of the time, and poker maybe 20% of the time. This does not mean that they aren’t well written and interesting; it’s just that I don’t have a huge desire to read about the personal lives of poker playing strangers. Sorry.

But one topic that piqued my interest on a newbie’s blog was that of poker bots destroying online poker. Now the concept of a poker bot has been around now for some time, with whispers and rumors around the edges of the poker community like old scary fairy tales to frighten children. For example, Neverlose on the 100-200 game on Stars is alleged to be a bot. These scaremongers cite the cases of Chess and Backgammon and how computers have “solved” those games, and predict poker Armageddon when the rise of the robots marches into poker. Bullshit.

The reason Chess has proven so amenable to computing is that it is a game that can be beaten if you can process through all the future permutations successfully. Kasporov wasn’t outthought; he was “ground out” by a processing engine with huge capabilities and had been programmed to understand his style. Similarly, backgammon just happened to be a game that fitted a neural net approach, whereas neural nets have not been anywhere near as successful in other games.

To my mind, one of the issues will that will prevent computers tackling high level poker play is that so many situations are very flexible – for example the compensations and differences between playing a hand against many or just one player. Another major factor is the necessary combination of lots of money, technical ability and poker excellence. As Darse Billings put it, as the inventor of the leading HU poker bot, the problem with poker versus the other types of game is that the computer has to think. This is something computers have a long history of not being very good at.

Beating low limit games on a rule-based basis, maybe. Becoming excellent at headsup play, probably. Being able to beat all-comers in a ring environment, I think not.

11 comments:

Big Dave D said...

Keith,

Ive just deleted your comment by mistake as I did a 1984-stylie clean up of my previous posts :-) To answer your question, affiliate is a general marketing term for an organisation that generates leads/customers on your behalf for a slice of the "action". As a punter you can only get now from most legit affiliates a signup bonus of some kind. And the affilaites can chose whether they get a share of the profit, or just a one off fee. The only reason I can think of Party trying to stop rakeback schemes is that they are frightened by what might happen if people really realise how much rake they paid. Cue Chaos :-)

Aksu said...

And this comes from a man who increased his blind profits with mouseclick programming =)... sorry a cheap shot.

I've been predicting this armageddon too, not very loudly though. But I do think that CPUs are going to destroy online poker at some point. Reason is not that they will be much better than experts but 1) they will be good enough and 2) volume.

Hopefully we'll get more comments from programming world (AndyW?). As I have to admit that I'm not even a novice in this area.

Cheers,
Aksu

chaos said...

Cessation Party Poker Rakeback:

Dave it maybe an awareness issue, or it may just be plain greed on their part. Probably both. Party poker is not a site I trust. Of course that doesn't mean I won't play there.

Bots. I lean towards AKsu here to a point. SH handed, headsup limit holdem they will excel at imo. When I played Erik there was one hand that made me wonder about the assistance of software. I'd seen very few people make this call from the turn to the river yet it was perfect logic all done in a second. In truth it was probably conditioning since he plays so much headsup.

It is far more easy to fit into pattern plays HU. This trait and a neural network armed with Bayesian logic could be very tough to beat.

From a purely technical point of view bots are a worry, particularly if they use on-line tells! However, bots causing the demise of poker on the internet, I don't believe so.

There is too much money being made by poker rooms, too much enjoyment shared by so many people for bots to kill the game. Since poker rooms aren't accountable to much international law, they can be draconian and keep large sums of money. Card rooms will work on solutions, anti bot devices etc etc Above anything else, what good will it do bots to kill off the game?

I think in 10 years time their will be a lot of people making money through bots or Bayesian-based software or something, but I expect they only be above average players. Not exceptional. Why would their owners want to draw attention to themselves, why would they want to make 7BBs an hour when they can livelonger earning 2 BBs?

Bots are going to be a real problem in the future, but I sense a happy medium will be struck. Hopefully they will just take the slack that rake reductions will leave.

Andy_Ward said...

I'm fairly sure that I could write a bot that would beat most online NL Sit and Goes. Writing the interface would be more difficult than the decision making.

Maybe I should !

Andy.

chaos said...

To put my point another way: I'd rather have 7 bots earning 1BB an hour than 1 Bot earning 7 BBs/hour.

Andy, I guess the interface would be tough (but I'm no coder) - perhaps as in a Borg v Enterprise battle the poker rooms could keep remodulating the shield frequencies forcing the bots to be recoded every few days. But sooner or latr I'm sure there will be a paper clip sitting on the desktop saying 'I'm not sure I'd call with that sir'.

Big Dave D said...

I think the key is volume as Aksu said. I dont think a bot could play a ring game anywhere near as good as a human, but if it could play 20-40 tables at a time, 24/7, how good does it have to be? The answer is frighteningly little. In fact to most humans it would probably not even look much better than break even. However the funds required to build a robust enough software that could handle and manage these volumes, coupled with the enormous bankroll required, could well take the best part of $1mill. But it wouldnt take long to win that back.

Andy_Ward said...

The interface is a pain in the arse, quite frankly. It involves frigging about with calls into Windows and so on. If someone could do an interface for me I'll do the logic and we can go 50/50 !

Andy.

The Fink said...

Good evening. I am a bot.

Unfortunately I've decided to quit automated online poker because I just got fed up of all the idiots outdrawing me. I lost with AA seven times in a row - whatever way I look at it it does not compute.

The only logical answer is that these sites are rigged. I'm going back to the day job as a BigTrak.

Anonymous said...

What is the deal with neverlose anyway? Has he ever chatted? Anyone have biographical information?

Anonymous said...

Recently finished a degree in AI and did the poker bot thing as my final year project. Neat way to call poker play research huh? ;)

A pretty simple bot has an easy edge over smaller stakes limit cash games. As most of you probably realise from playing these games you barely need to pay attention and so no opponent modelling is required to show a profit.

To take things to a higher level requires a bit more sophistication in the programme (and gets a bit fiddly!). As you mentioned, Billings et al. have a bunch of kick ass bots which they've been working on for years and certainly put mine to shame.

However they are all currently based on limit hold em. Grinding out a profit on these games online is all well and good but I think it will be a long time before the same can be accomplished in NL games.

As previously mentioned the most difficult part of all this is hooking the thing up to the poker rooms once it's working. There are several different ways of going about this which I won't go into but it's tough. Plus you get the whole hacker scenario where many of the sites are regularly changing things around to combat the bots.

I think this makes the whole process too tiresome for the casual programmer looking to test out their bot and I think that only someone with the serious intent of making decent money out of the venture would bother.

Therefore I can't see anyone releasing anything freeware to enable others to get involved with their hard earned cash cow - which is going to restrict the vast majority of the population from doing the same.

As with most things in life, rules exist to prohibit certain behaviour but if someone is dedicated enough they will be able to find a way around it. The number of people who are able to (AND want to) do this should be fairly small though so I don't think it is really going to impact the poker community anywhere nearly as much as the scaremongers would like to suggest.

reevio.

Big Dave D said...

Tx for the comments Reevio. Another prevention point, beside the huge amount of money, is that it is one thing for clever people to say that they can build something, it is something again to get them to build something that is very robust, bug free and works as designed. This is why the software industry is a tough one. Anyone that lived through the dotcom bubble, as I did, saw lots of people with good ideas but they just couldnt build them "good" enough.

gl

d