Author: Joseph Ciarrochi
Date: 12:58:07 01/05/06
Go up one level in this thread
Yes, your method sounds efficient. I think the next part would be time consuming. I think you would need to have a massive data set of games of players from all levels who are known to not be using computers. Then you would have to generate some stats for each playing level. E.g., how many times does a human pick a move that is less than .1 difference from the best move picked by the computer. Lets say that you find players between 1600 to 1700 on average have only 15% of their moves being in this top level. Lets say this varies at the 99% confidence interval between 12% and 18%. This means that if a 1650 player consistently has more than 18% of their moves in the top level, they may be cheating. Now, if someone uses a computer for every move and pretends to be a gm, we may not be able to detect this.. (although I wonder if even GM's don't pick as many top moves as computers :)). To detect cheating this way, you need say a 1700 player to be occasionally using the computer. best Joseph O n January 05, 2006 at 10:36:37, Cap Ricardo SantAna wrote: >On January 04, 2006 at 15:38:44, Joseph Ciarrochi wrote: >Hello > >ok. your ideai is very close to mine :) First I analyse games with a good chess >engine (said fritz or hiarcs or junior - I do not have schredder :( ) with >itś own full analysis tool. I choose a value of 10 (0.1 pawn value) >parameter. So, if engine finds a variation which is better than 0.1 pawn the >move played, then it adds a variation. > >Ok. this is easy, because all "dirty" job is a computer one. Itś >interesting that this analyed game gives me a good clue if one is playing or not >with computer help (ok, I am considering that cheater which uses computer almost >all time). > > After this, I use chessbase and analyse the game with one more two engines >(fritz and hiarcs is my common choice). So, I begin selecting the moves played: >1 - for first engine choice, 2- for second engine choice, and 3 - for trhird >choice. > >My problems begin now. What can be consider a cheater ? whow much moves ? >thanks in advance >Sant Ana > > >>I am speculating here but..... >> >> >>I think one way is to look at the moves the human makes and see how they are >>ranked by a good engine. e.g., how often does the person choose the number 1 or >>2 move suggested by the computer. >> >>I have heard that it is statistically unlikely for someone who is not a GM or >>master to consistently choose a number 1 or 2 move. >> >>You could presumably analyze a massive number of games you know to be human and >>find the average rank of the move by rating level (e.g., an 1800 player may may >>on average pick a move that is rated 4th by computers, or tends to make x >>percentage of rank 1, 2, 3 and 4 moves). Then you could analyze any new human >>game and determine if the human is making moves that are statistically much >>stronger than you would expect based on their rating. You could also catch >>humans who are using the comptuer sometimes but not others. Their distribution >>of ranked moves would be very different from other humans (e.g., they would make >>alot of rank one and rank 4 or greater moves, but too few rank two and three >>moves) >> >>This is just speculation mind you. I wonder how else it could be done? >> >>best >>Joseph
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.