28 Jul 09
Originally posted by atticus2Squelch and others have tried the approach of looking at the average difference between the score of the top move and the score of the move chosen. In my opinion it gives more accurate results. It can be thrown near the end of the game when an engine user might turn off his engine because he has an easy win and start choosing moves which are "only" +6 rather than +9.
OK, another methodological query comes to find. Take the Fischer-Spassky game cited above, In c 30% of relevant moves, players chose outside the 'top 3'. This sounds like a big deal, but of course may not be if, say, the 'top 5' are separated by small centi-pawns. Equally, a 'top 1' pick is less of a deal if it's the only sensibly strong move.
28 Jul 09
1 edit
Originally posted by atticus2Blundercheck analysis is also a very good tool - possibly even better than top 3 matchup.
OK, another methodological query comes to find. Take the Fischer-Spassky game cited above, In c 30% of relevant moves, players chose outside the 'top 3'. This sounds like a big deal, but of course may not be if, say, the 'top 5' are separated by small centi-pawns. Equally, a 'top 1' pick is less of a deal if it's the only sensibly strong move.
It's just that I can't be arsed to generate all the benchmark stats for that!
As for obvious/forced moves, well these are taken into account in the benchmarks, unless you think that all these games have unrepresentatively high or low amount of them in many hundreds of non-book moves?
Also, removing those moves which you think are obvious/forcing introduces a human (subjective) element and also meansthe benchmarks would need re-calculating again from scratch.
Top 3 matchup is a perfectly fine system for detecting the most annoying & blatant engine cheats.
28 Jul 09
1 edit
Originally posted by Fat LadyI take it you've lost to someone you suspected was using an engine. But the highest rated player on this site is rated less than 2500 which is less than is needed to be a grandmaster. So perhaps they aren't after all.
I feel proud and a little bit humble to be part of a site where the top players regularly outperform Kramnik, Topov, Alekhine, Capablanca, Fischer, Spassky and Estrin. I guess it shows how much chess has "evolved" in the last decade or so.
28 Jul 09
2 edits
Originally posted by GoshenThe ratings on this site do not have a complete correlation with official FIDE ratings. Who knows what rating a GM would achieve on this site?
I take it you've lost to someone you suspected was using an engine. But the highest rated player on this site is rated less than 2500 which is less than is needed to be a grandmaster. So perhaps they aren't after all.
And yes, I have lost to several engine users on this site (and also to lots of genuine non-engine-using players). I don't particularly care about losing the games, it's the time and effort I put into them, thinking I am playing a human and hence have a chance of winning!
28 Jul 09
Originally posted by GoshenI imagine if Anand was to come and play 100 games or whatever it took for him to establilsh a rating, it would be somewhere in the 2300s. Maybe he would even be be above weyerstass, and be number 1, but I doubt it. Anyway I think nomatter how good someone was, I think it would be virtually impossible to establish a rating very far above 2500.
I take it you've lost to someone you suspected was using an engine. But the highest rated player on this site is rated less than 2500 which is less than is needed to be a grandmaster. So perhaps they aren't after all.
29 Jul 09
1 edit
Originally posted by clandarkfireThis can't be right, surely. Anand would reach 2900 just by beating me regularly - not that he would of course 😛
I imagine if Anand was to come and play 100 games or whatever it took for him to establilsh a rating, it would be somewhere in the 2300s. Maybe he would even be be above weyerstass, and be number 1, but I doubt it. Anyway I think nomatter how good someone was, I think it would be virtually impossible to establish a rating very far above 2500.
My reasoning is as follows. I'm currently 2300. If I beat a 1700, I gain one rating point; below that, none. So I can creep ever upwards, if that is my desire 🙄 , by beating players 600 below me. By 2500, my threshold would become 1900, and so on. Technically therefore Anand and the other super-GMs could reach 3000 by beating the best in here.
What surprises me is that they haven't bothered so far. Maybe they're scared 😲
29 Jul 09
You need a much bigger pool of stronger players to inflate ratings at the extreme upper end. It's just the way the ratings formula works.
A top GM could play as many games on here as he likes & with the current players he probably wouldn't get over 2500, as someone said earlier.
29 Jul 09
Originally posted by SquelchbelchI guess so. I forgot the simple fact that, as Anand beats me and his rating creeps up, so my rating declines. Even if we played 1000 times and I lost the lot, Anand's rating would be limited
You need a much bigger pool of stronger players to inflate ratings at the extreme upper end. It's just the way the ratings formula works.
A top GM could play as many games on here as he likes & with the current players he probably wouldn't get over 2500, as someone said earlier.