Crowd Chess

Will we be black or white?

That's funny!  I know whenever I am in search of wisdom, I look for the nearest mob.

For those interested, the game has started. Moves will be combined and played against the computer every night at approximately midnight PDT. The chess engine you will be playing is GNUChess running on a 3.0 GHz Core 2 Duo processor. 

Some people have given anecdotes of crowd knowledge not working well. So I figured I'd give one that would work well.

Imagine a scenario where you have a room full of young adults. You have to answer a series of pop culture questions (topics of music, television, etc.). You get to use this group of people in one of two ways:

1) Choose a hero from the group -- one person who will answer all the questions for you.

2) Let everyone in the group answer the questions (but you can't see who answered what) and take the most popular answer.

It seems to me unlikely that there would be an individual in the group whose knowledge would be so encylopedic and eclectic that he would be able to answer all the questions correctly. But I would not be surprised to find that for each question there were a handful of people who answer it correctly.

So in this case, the crowd would perform better than an individual.

I don't think the above example is a good analogy to a chess game. But at least there is some hope for crowds. :-)

yeah that is exactly why i don't think this project will be successful. "crowd knowledge" should work in situations like that (where there is one correct answer that is static and all other answers are incorrect). that is not the case in chess and therein lies the problem with this study.

the "crowd" might play a pretty good game, but not a game that exceeds 3000+ rating of an engine. "crowd knowledge" may make a group more accurate in a static situation, but it won't make them exceed a (more or less) constant level which is in flux.

I registered, logged in and tried to move. It wouldn't let me - the piece returned to its original square when I released it. Then I tried the radio buttons and they didn't work either. Anybody else try it?

I've already put my moves in today so I don't see the instructions anymore, but I'm fairly certain they say something like "Click on the piece you want to move and then click on the square you want to move it to". So it's a click and click interface, not a click and drag.

"So it's a click and click interface, not a click and drag."

Yes, that was it. Thanks.

I think this project needs to drum up a bit more support...

Despite the saying, three is not really a crowd.

From here:

On a 2.4Ghz Pentium 4 with 512MB of RAM "nextgnu" on FICS attained standard rating of 2296

So I would say it's master strength -- it's not a 3000 rated monster.

I joined after the first two moves but the crowd agreed with my 3rd move. (Nc3 was a pretty obvious move.)

If we discuss here possible next moves, would that improve the experiment or would it destroy the experiment as they planned it?

It would utterly destroy it. The idea is to free the system from any outside influence.

At least it's not another sicilian -- the grunfeld is exciting, should be a good game.

I would advise you not to discuss the moves here - the possibility of an information cascade to occur would become more likely. An information cascade is where someone makes a suggestion, and then others agree with it because it seems reasonable, however they don't do their computations on their own because they already have a reasonable answer. This also creates a peer pressure effect - you see two people agreeing on a move and thus you agree on it as well (albeit in online forums this may be slight). In effect, this degrades the amount of information that is put into the system (and have been accredited to have dire consequences; in "The Wisdom of Crowds," Surowieki explains how an information cascade may have been the reason for the Columbia Shuttle disaster).

SocialCodeGroup, why was a Borda count method implemented in this experiment?

Get Carlsen, Kasparov, Anand, Kramnik, Nakamura, Aronian, Topalov and some other strong gm and they might... MIGHT.. have a chance to win, but its probably gonna be a draw.