Not to go into too much detail Matthew, but that simply is not the case, at least when batches of OTB games are analysed.Matthew Turner wrote:...
2. Historical analysis of games is a deeply flawed comparison, because players increasingly use engines to prepare/analyse their games, so a higher incidence of 'computer moves' is to be expected.
I've been helping online sites find & boot cheats for many years, and it's quite remarkable how little has changed in terms of engine-like non database play from the days of Capablanca, Alekhine, Steinitz, Fischer etc.
The analysis needs to be done under certain criteria though, and one game proves very little.
If I take the games from a modern great like Carlsen or Kramnik, what I'll do is select the 20 most recently played games vs 2200+ FIDE opponents. These games should be checked to make sure that they have at least 20 non-database moves. You want longish games, because these should be more balanced - thus engine choice frequency becomes more important, rather than a game where a blunder is made early on & then there's a clear winning line. This process avoids cherry-picking high match rate games.
Next I analyse the batch for engine match rate correlation for all the non-database moves for the top 3 or 4 engine choice moves.
Then you do likewise for your suspected cheat.
To cut a long story short, if I analyse a batch of games from any modern Super GM OTB player, the match rates for top 1, 2, 3 engine choice moves will all hover around 60/75/85%.
I know because I've spent several hundreds of hours running batches, and so have about 20 or so other people, all finding remarkably similar results when using Deep Rybka, Houdini, Deep Fritz....
