CrowdChess Forum
Some suggestions for improvement on crowd intelligence
The wisdom of crowds work when the participants are unbiased, make independent solutions, and are not subjected to "leading" clues. There are two aspects of crowdchess that goes against these principles. 1. Vote on moves - The most popular move gets eyeballs. Instead of independent thinking, the players tend to evaluate the suggested moves. Now, when you see a move with 30% or 40% votes, there is a natural tendency to go with the "Crowd". 2. Moves with player names - As suggested moves have player names, again the popular player (with most supported moves) will tend to garner more support. It is inherent for human nature to support popular figures. My suggestions to overcome these issues are: Allow each player to make individual suggestions. No voting! Nobody gets to see what moves others have suggested. Each player independently thinks and make a move. The algorithm then takes an "average(*** - see later)", or the most popular move and makes the move. By doing this, players are not "lead" into moves. Independent moves will automatically cancel out the errors in decision making. ----------- *** "Average" is a concept where you dont make the popular move, but makes an average of all the moves suggested. My idea on averaging is too early to be detailed at this point of time. The basic idea is to have gravity for each piece, and based on the "pull" by players by proposed moves, you move the piece with the best pull. As all pieces have mutual attraction, and as pieces change orientation on the board, the strongest pull could be for a piece that may not be the most popular. However, to take care of unbalanced gravity for board edges, the algorithm might have to assume a spherical board... Well :), the thoughts are too early to be properly structured.. P.S: Join this discussion on by blog on this site.
The downside of independent moves is that it'll take the social structure or social networking aspects out of the game - like chatting, discussions, etc. This may not make commercial sense :(. </br> Looking at in another way - all social aspects are intact. Players will come back to check their moves, and what happened to their suggestions (much like what happens to your vote in an election). Ranking of players can be updated after the move. Discussions will happen after the move on why a certain move was better (in retrospect) - much like news channels - all analysis is in hindsight :)
Actually, voting on the popular move or finding the popular move from the independent votes is flawed. For chess, its quite likely that the majority of the crowd are not great players. This means that the most likely, the most popular move is not great as well. So, you are just aggregating a bunch of "ok" players and pitching them against a grand master. Thats why you need an averaging move and not a popular move. The averaging move has the potential to find the "good" move from a bunch of "ok" moves. With the current algorithm, I predict that the "crowd" will lose. And, this is not because the "wisdom of crowds" is flawed. The flaw is in "filtering" the wisdom - the algorithm is incorrect.
You must be logged in to participate in the forum