You dont have to turn to me for that, theres plenty of websites that have detailed results of engines running on different specs, providing a much wider range of data than my personal experience can offer.
I really hope people dont ever trust anyone when they make a chess statement, even something like Na3 is a bad opening move could be turned on its head if it was found that was the starting move of the perfect game. I havent made a public accusation since the chess mods have started (with the exception of JW) iv said a bit in the threads and confirmed other things.
The 90% fritz is pretty easy, just by looking at a game you can get a feel if its been played by an engine or not. Basicly the way id go about seeing if someone uses an engine or not is to try and prove they dont, a good start would be to find key moves, a move with a whole load of ideas and variations all of them good and then check with one Fritz likes. I dont want to go into lots of detail in the open forum because it will just mean giving enough information for people to cheat and not be caught. After you have found the key moves, found an engine that picks those moves, you can start to run games. Its then just a case of %. Im going to stand by my statement that no human players like a engine game after game, unless someone can prove otherwise. You shouldnt run down the list of moves on an engine looking for match ups, because you would find a match to everyone on the site then, but top two moves over 90% then yes its an engine. As for the time you run it, basicly till you get a match or the move clearly isnt fritz, however you shouldnt jump about wildy with time either, If you arent getting matchs for the same time between moves then its not fritz, since the top fritz moves dont really jump about that much, for 20 seconds they do go wild but then its only position that changes, as fritz picks the one it likes best. I imagine most cheaters in the top block use about 5 mins at least. Someone pointed out it was odd that when Fritz does analysis its starts with the last move and works to the first. There is a reason for that in that Fritz stores the variations it has looked at, so if its going backwards though a game it can see ahead better. The people who cheat with Fritz use it going forward. Quickly testing for cheating games can be done going backwards, but you need to do it forwards to check. Since you need to recreate the game they played with no help from the game itself or you.
I'll try and make it clear, iv posted the number of variations in chess a few times I wont do it again but imagine this.
A player on the north pole playing someone and a player on the south pole playing someone else. The chances of these players moves matching up to each other for even 1 game is tiny id think you would agree? However change those players for two engines and the moves would match up. The 10% inaccurace between the pcs will be speed. The players could play game after game and you wouldnt find much corrolation between them, however the engines would keep running the same varations, the same ideas because that is the way they are programed and it is set in stone, they are the same, the games will be the same.
Now a player in the north pole and an engine in the south pole, do think that time after time the player will match 90% of the engines thoughts and ideas?
For all the chess games ever played im not sure how many will match up to each other, but for games 40 moves long I cant think there will be that many at all. Maybe less than 1000 games, iv no idea really but im sure the number is small, since we'd all have seen the GM games that would match up to each other. But some peoples games on this site match fritz when you just let fritz pick its own moves with no human aid at all, you have to admit that is dodgy.
Really its not that hard to tell a computer cheater and a player apart.