Mig said :
"... That the bots aren't yet strong enough to really challenge strong players is another factor, one that won't last long ..."

Yes, if they did it in chess, they definitely can do it in poker.

Interesting...Artificial Intelligence is going to eventually spoil all activities where human intelligence is mixed with art : it's a matter of time perhaps, as the matter with chess maybe.

Bad news : I am afraid we cannot stop it.

I don't think cheating is the most annoying thing about online chess. Ironically, I think people ACCUSING the opponent of cheating as soon as there is the slightest suspicion is far worse. (I.e. "you took more than 5 seconds to make a standard theory move" or "you shouldn't have seen that tactics, you are not that good".)

The first thing that a stronger player than me (lets said 200elo pointers better) does when i beat him is accusing me of cheating.

Two additional problem bots have in poker that make it much harder to program an expert computer player than in chess:

1) It's a game of incomplete information. It's tough to program a bot to try and put the other players on a range of hands (at least, with any degree of complexity).

2) There are up to 9 other players playing at once, which creates a lot more variables than in chess. There are heads-up bots that already do play very well - the full games are a bit more tricky.

Of course, neither of these is a significant enough problem to win at the lowest limits of online poker, but they do create some barriers to making an expert poker bot. It'll happen eventually, but it might be a while.

With regard to acirc's comment, personally, nothing makes me happier than being accused of using a computer after I find a nice tactic!

There is a lot of paranoia out there. The great majority of players do not cheat, but the suspicions are rampant.

As with most ethics issues in sports, business, politics, you only need one or two cases to provoke general distrust. All politicians are crooks, all companies are like Enron, all cyclists are doped, etc. In most cases it's unlikely that the only people doing it were the ones who got caught (the "where there's smoke there's fire" argument).

In professional online chess events it says something that even the players assume everyone is cheating. This is partly sour grapes, but it's obviously rampant.

Let's me ask sth : Is there any way to judge whether one chess player has been aided by a computer : I mean analyzing the moves that he has played and finding out "inconsistences" ?
It sounds pretty tricky for a man but perhaps this analysis can be carried out by a ... computer 8-))
Is there any kind of thing ?

eg. totally agree with Joshua : though, I have an overall score of 0/10000000 against a comp ! 8-))

There are many methods for detecting cheats, often used in combination. The better the person is at it, the harder they are to catch, and if someone is competent there is no way to prove cheating 100%. This is why organizers of online events must make penalizing this at their discretion, so they can punish those they are 90% sure cheated; i.e. beyond a reasonable doubt.

The simplest way is catching people who are only using one computer. The online interface can detect the use of a computer engine, most can even tell you which engine and more. That's your 100% certainty. It can also pick up window focus, i.e. alt+tabbing between programs with each move.

Someone using two computers is obviously much harder to catch, especially if he is a strong player. A weaker player has to pretty much follow the computer's recommendations blindly, or setting it to show the two best and alternating. This leads to fairly trivial detection by heuristics analysis. Picking one of the computer's top few moves every time just isn't normal for human chess, especially at rapid time controls.

A stronger player can recognize a "computer move" that a human wouldn't consider. The trick is that if you're following computer moves, you often get led into a position where only a totally bizarre computer move gets you out. Meaning the computer saw it all along and gave its +3.13 or whatever and you followed along without seeing the whole line. Then you have the human win versus the computer win. E.g. one line leads to the win of a piece and obvious resignation, another leads to an incredibly convoluted mate in nine. Any strong human picks the first, comps pick the second. This sort of thing is tell-tale if it happens more than once.

There are other, more sophisticated methods involving things like the time used per move and detecting patterns over the course of several games. But the bottom line is that if a strong player just uses a second computer for advice at a few key moments, resisting the temptation that would lead to falling into a pattern, nothing could be proven within a reasonable doubt. Most people just aren't very good at it.

Poker and other games without 100% information and with heavy bluffing elements are a different challenge from chess, one that requires a more AI take. If a bot's betting patterns are detectable and exploitable, it will be ripped apart, unlike at chess. No matter how well you play the odds at poker, it's equally important how you play the other people at the table. So human weakness, playing the odds poorly or playing "irrationally" to bluff or fool your opponents for later hands, mean computers will never be as dominant as they are at something like chess that can be tackled directly.

But they will certainly keep getting better and better, and you don't need much more than playing the odds perfectly and some general tendency tracking to beat weak players.

Jonathan Schaeffer at U of Alberta, mentioned in the article, is better known for his champion checkers program, Chinook. He wrote a great book about it, One Jump Ahead. Cool guy. Met him when he was on the committee at the Kasparov-Junior match in NY. http://www.cs.ualberta.ca/~jonathan/

Hey, thanks for all this interesting stuff.
Did you write it yourself or any comp helped you ? 8-)))

So the whole matter is to distinguish human way of thinking and the comp's way of thinking : the comp just chooses the perfect tree of moves.

Just a comment on George's initial statement about AI. No current chess program or poker program use or is related to what a layperson would consider "AI" in the slightest.

AI is just a pipe dream for the forseable future.

Mango, it may be a paradox but chess algorithms (min-max and the like) which are heavily being used in chess programs like eg. Fritz are technically considered as AI methods for solving difficult and complex problems.
Anyway, you are right in the sense that this look rather computational to "smart". I agree.
In my opinion, however, this is the way all future robots etc. will work : making very primitive steps in an incredible speed (reminds of a CPU but will be parallel and most efficient).
Any such "brain center" will be specialized (eg. eyes center, voice recognition center etc.) and will communicate with all the others interchanging info (as in the case in brain).
Sorry for technical stuff. 9-))

Fortunately most guys that cheat online are idiots, and get caught quick. :-)

I have been captured by the poker programs on TV, and I have been ordering some poker books :) Internet poker, here I come! This talt about computers playing poker on the net is some cold water for me. I guess the future of board games for money is where you meet your opponents face to face, and not on the net. Well, I spend to much time in front of the computer anyway :)

Am I the only person to notice that the poker computer has a "team mode" in which two or more bots at the same table can collude? No doubt the developer has constructed a moral defence of that too.

I would say that a computer could never play as well as the best humans (live anyway) because of the high level of emotional and psychological content but I'd feel horribly naive

DP wrote :
.. because of the high level of emotional and psychological content ..

Sorry, can you clarify this statement a bit ?

I think I have caught your essence but not quite sure..

Nick,

You make an interesting observation about the risk of collusion. I wonder if it is really a big concern. First, if two players (or bots) work together on a 10 player table of Texas Holdem then each of the cheaters will only gain knowledge about two additional cards out of the deck of 52. This might result in small additional gains over time but the advantage would not be great (I think). A second consideration is that collusion would probably be very easy to detect because of patterns of player A always getting out when player B has a stronger hand and vice versa.

Of course as I typed that last sentence I realized that the colluding players could try to avoid detection by often betting against each other to avoid this pattern detection. But frequently betting against each other would hurt the success of the cheating because each time they lost they would lose more than they needed to (back to the first point: that knowing two additional cards is not that great an advantage.)

I agree with you Chessplayer, this is most irksome for a guy that has spent a lot of his resources just to learn how to play great chess...
Unfortunately, many especially weak players are GM in using chess s/w and are judging easily about extremely complicated (nice) positions. Many times the result is just ridiculous.
However, in tactical positions, compe may help you decide the player that made the most mistakes ! 8-))

Amstar,

I'm not a serious poker player, but am assured by those who are that collusion at the table is taken VERY seriously ( and is punished in draconian fashion ). That is generally only using subtle hints rather than a precise exchange of information, so we can assume that two comps in collusion would be considered to have a huge advantage.

By the way, a friend who is a genuine professional poker player, i.e. that's how he fills in his tax returns, assures me that he considers internet poker such an obvious candidate for cheating that he would never contemplate playing.

The basket lives!!! That is my battle cry on uchess.com -- home of amazing cheaters. The moderator - chesswizardry -is an admitted cheater using 4 processors but he now denies this. It is a good site but nearly the whole first page cheats. Can't beat em, join em I say, sooo.

My question: If you set your chess software to move randomly and not use opening book will it still be detected??