1 edit
Originally posted by KeplerI don't "massage" analyses like you and others apparently do. I'm perfectly willing to run my analysis at the same setting in every game and what results come in, come in.
I am not trying to down play the matter, just put some perspective on the one (or a few) game analysis malarkey. It is irrelevant. One of my analyses of a game in the other thread produced a match up of 84%. Unfortunately I can't remember if I posted that or not. However you look at it, 84% is lower than 88% which was my point. Pick the game and massage the a ...[text shortened]... bly and a person can get the match up they want. Large numbers of games are a different matter.
The games analyzed in the other thread (which, I repeat, were chosen by others to supposedly show that the player had a "non engine like" style) were not short tactical games, but long, positional ones - one went 70 moves, another 50. The games analyzed totalled more than 125 non-database moves. Not a single move was not one of Fritz's first 5 choices; close to 70% were Fritz's 1st choice. I defy anyone to find such match ups in honestly played games of those lengths.
Originally posted by no1marauderFrom 37 games I get an average (arithmetic mean) number of moves per game of 41. Some are short (15 moves) but a couple are quite long (70+ moves). Most seem to be about 40 moves. I have no information on how many games he played at a time but i suspect it was not "dozens of games at a time".
Did Rittner play dozens of games at a time? How many moves are in your sample? Are they short, tactical games? I bet they weren't the 50-70 move games like the ones analyzed in the deleted thread.
No human player can analyze variations like an engine. You try to bang out a few million possibilities a second. Your claim that CC play is "lik ...[text shortened]... ?
I do analyze forward; I don't "blunder check" when looking at suspects.
I haven't tried any of Rittner's games myself so the quoted match up is astonishing. If true, it would it would cause many to have grave doubts about the whole cheat detection process. I have a vague recollection that the old game mod system was tested against pre-computer CC masters and all would have ben found "not guilty" by the system. On that basis, Rittner has presumably already passed the test and the extraordinary match up must be an artifact of the analysis process used.
Originally posted by adam warlock"CM agreement" is pretty useless, and it by no means constitutes an engine match-up. Anything below 100% means you have blundered badly. Or it could also mean that the CM engine is just too weak to appreciate a good move.
And this is the natural language evaluation. Even though he sugests different moves on the numerical evaluation he gives me 100%. I think this has something to do with the fact that the engine does the evaluation from the end to the start... Anyone with more engine knowledge could give me a clue.
It has nothing to do with the order of evaluation.
In Chessbase, too, the built-in automated analysis tools are entirely useless for game modding. The job has to be done manually, on infinite analysis, move by tedious move.
The order of evaluation is very important. Although it will not alter an engine's preference as seen by a cheat in all cases, in many cases it will. Engines often appreciate good non-engine moves after the fact, but while running under normal conditions it might well have pruned the same move from it's evaluation tree. Reverse evaluation will almost always give higher match-ups than forward evaluation.
Originally posted by KeplerIn game modding other things are taking into account rather than just match up rates. I don't know exactly what they are but fot the method to be reliable it has to be like that.
I have a vague recollection that the old game mod system was tested against pre-computer CC masters and all would have ben found "not guilty" by the system. On that basis, Rittner has presumably already passed the test and the extraordinary match up must be an artifact of the analysis process used.
- Joined
- 26 May 02
- Moves
- 72546
1 edit
Originally posted by VarenkaInteresting post.
Like most others here, I hate cheats and want them banned. But here’s some thoughts on the latest discussions…
[b]Human Potential
In order to see what is humanly possible at correspondence chess, I’ve been analysing the games of Rittner who was CC World Champion 1968-1971. Rittner wasn’t an OTB GM. Using Rybka 2.3.2, I’m seeing matchups of 92% a ...[text shortened]... es, we should analyse the game forwards since this is what a potential cheater would have to do.[/b]
From you're statement "Using Rybka 2.3.2, I’m seeing matchups of 92% and 89% so far" I assume you've only analysed 2 of Rittner's games? You need to analyse a lot more games than that before reaching any conclusions. But even so I would be interested to see which games were chosen for analysis.
I don't agree with your theory that "Top CC play is more engine-like than top OTB".
If that's true, it's only because so many correspondence players use engines! I don't see why an honest correspondence player who doesn't cheat should play like an engine.
"A top CC player lacks the same intuition". I disagree with this statement also. Intuition is intuition, regardless of the conditions and rate at which the game is played. Intuitive players see the move they want to play almost immediately, and often then wastes time trying to justify it to themselves. In OTB chess they might spend 30 minutes or an hour thinking about and trying to analyse a tempting sacrifice before finally risking it. In CC, they might spend several days. If the intuitive player follows his instincts he will usually decide on the same move that he saw within seconds.
By the way, I wouldn’t describe Kasparov as an “intuitive player”. Kasparov’s great strengths were his calculating ability, fantastic opening knowledge and memory. In unfamiliar positions, he would often flounder. Karpov is a better example of an intuitive player. He instinctively knew which were the best squares for his pieces and could therefore play extremely quickly. If Karpov joined this site, I don’t see any reason why he would play less intuitively here than in his usual grandmaster tournaments. He would probably play quickly, win most of his games but lose to some of the engine users! His match-up rates should also be significantly lower than those players.
Originally posted by no1marauderThe massaging was done to prove a point not to deny cheating. Furthermore, I suspect that long positional games with a high match up are actually more indicative of engine use than short tactical games with similar match ups.
I don't "massage" analyses like you and others apparently do. I'm perfectly willing to run my analysis at the same setting in every game and what results come in, come in.
The games analyzed in the other thread (which, I repeat, were chosen by others to supposedly show that the player had a "non engine like" style) were not short tactic ...[text shortened]... choice. I defy anyone to find such match ups in honestly played games of those lengths.
Just in case you are in any doubt:
I AM ON YOUR SIDE HERE!
Originally posted by GatecrasherToo bad; maybe one of the Forum Mods can give you a copy.
Sadly, I did not read the original thread, so I have nothing to whitewash here.
It's nothing new; I reported this user name and provided detailed analysis to the Game Mods regarding him three years ago. He is apparently sacrosanct for some reason.
Originally posted by gambit05There is a difference between analysis of the usual chess sort and the sort of analysis we are talking about here. You can post as much analysis of the standard chess variety (13. Bc4 is an obvious blunder, so and so suggests going for a beer would have been better) but I suspect that posting engine match up rates would be taken as an insinuation, if not a direct accusation, of cheating.
Is it within the rules of this site to post game analysis of players, but without commenting them?
Would be nice to have a thread where such analysis would be shown from the top 50 or so players. Just out of interest of course and without any accusations.
Originally posted by no1marauderI’ll post the Rittner details when I get to my home PC.
Your claim that CC play is "like" engine play is false.
It was stated relative to OTB play.
what level of engine match ups in a significant number of games would you be unwilling to concede is overwhelming evidence of engine use?
My honest answer is that I don’t know. What’s your figure for this?
Should all CC sites just give up on the idea that there are engine cheats
Absolutely not. But I view it as a very difficult problem to solve effectively. I wasn’t totally convinced by previous posts, but that doesn’t mean I don’t agree with the motives behind them.
- Joined
- 26 May 02
- Moves
- 72546
Originally posted by gambit05Unfortunately, as Kepler says, if the player has a high engine match-up, it will look like a cheating accusation and almost certainly be deleted. We might also lose the entire thread!
Is it within the rules of this site to post game analysis of players, but without commenting them?
Would be nice to have a thread where such analysis would be shown from the top 50 or so players. Just out of interest of course and without any accusations.
If the player has a low match-up, the post would be probably be ok 😉
Originally posted by KeplerI mean the latter type of analysis.
There is a difference between analysis of the usual chess sort and the sort of analysis we are talking about here. You can post as much analysis of the standard chess variety (13. Bc4 is an obvious blunder, so and so suggests going for a beer would have been better) but I suspect that posting engine match up rates would be taken as an insinuation, if not a direct accusation, of cheating.
Originally posted by GatecrasherCM agreement" is pretty useless, and it by no means constitutes an engine match-up. Anything below 100% means you have blundered badly. Or it could also mean that the CM engine is just too weak to appreciate a good move.
"CM agreement" is pretty useless, and it by no means constitutes an engine match-up. Anything below 100% means you have blundered badly. Or it could also mean that the CM engine is just too weak to appreciate a good move.
It has nothing to do with the order of evaluation.
In Chessbase, too, the built-in automated analysis tools are entirely useless ...[text shortened]... n tree. Reverse evaluation will almost always give higher match-ups than forward evaluation.
Is this true because The King is a weak engine or because it is true period? I know that for game modding issues you guys couldn't just go for engine agreement percentage and take other factors into account. Is this what you're talking about here?
Engines often appreciate good non-engine moves after the fact, but while running under normal conditions it might well have pruned the same move from it's evaluation tree.
I don't think I fully understood this. Are you saying that even though sometimes an engine evaluates a move as bad and doesn't go down the line to see what's happening or that the engine just discards analysing the move from the beggining?
Sorry for the newbish questions but some aspects of how engines do their stuff are just alien to me.
Originally posted by David TebbThen the solution is to post only those with low match ups. This would give interesting gaps.
Unfortunately, as Kepler says, if the player has a high engine match-up, it will look like a cheating accusation and almost certainly be deleted. We might also lose the entire thread!
If the player has a low match-up, the post would be probably be ok 😉