Originally posted by philidor positionAll I can offer is some rough Java code that I run from within an IDE. It currently doesn't run as a standalone app. If that's of use, you can have a copy, but you'll need to muck about with the code (i.e. to specifiy location of UCI engine to use; time per position; etc).
would you be willing to share it?
Originally posted by hwwThe numbers are just raw statistics and you are right to consider what they mean.
And how expound this "numer games" that a 63% top1 wins vs a 78% top1?
Suppose I play to a level of top 1 matches at 80%. That may sound really good but if the other 20% of moves are complete blunders then overall my results won't be good at all. e.g. play 4 "perfect" moves followed by losing a queen. Hence, more than just the top 1 needs consideration. Even looking at the "top 3" leaves other questions/doubts.
Originally posted by heinzkatWe're able to compare what elite CC players obtained from 30 years ago but due to possible engine assistance we can't use the most recent CC world championship games. Can we be absolutely sure that nothing else has changed during those 30 years?
What doubts?
For example, my CC play is helped significantly when I use Chessbase to keep track of all my analysis lines, especially in complex endgames. There is just no way I could do this as efficiently using pen and paper. And even with tablebases being banned on RHP, there is still so much more information readily available on the Internet, books, DVDs, etc. Aside from engines, has anything else helped raise the standard of CC in the last 30 years? I don't know. What does this mean for our "top CC" baseline that is 30 years old? And today's top OTB is a different game from CC so what can we infer from that?!
Also, some types of middlegames are more forcing than others. Compare a position with opposite castling where a storming of the enemy king is demanded by each side, to a more quiet maneuvering battle. Are we sure that some types of positions don't lend themselves to having a more obvious and narrow direction of play, and hence increase the likelihood of agreement with engines?! For a pool of GMs with similar strength, do the Sicilian players show a similar "top 3" to the French Defence players and to the King's Indian Defence players, etc. Are attacking GMs similar to those who strive for the endgame? I don't know.
The "top 3" method makes the assumption that the rate of forced moves, etc. will even itself out if enough games are considered in comparison to a baseline. So, for example, if there's only one legal move in a position, that will be marked as a top 1 choice. i.e. evidence towards cheating. Likewise if there are 10 legal moves but 7 are blunders, you either play a blunder or contribute to the cheating evidence. However, in these cases, the "top 3" theory is that the baseline will have done so too at an equal rate, and I can see the reasoning here based on statistics. But again, while I view it as worthy argument, I don't regard it as concrete proof. It remains an assumption.
The doubts raised above are only relevant for borderline cases. I wouldn't use any of them to excuse a blatant cheater.
Originally posted by ExumaLet us not forget that woodworm lost his only 3 lost games on timeout.
And so...the highest ranked unbeaten is now User 371363