Please forgive me if this concept has already been discussed..
Perhaps for the top 100 players (more or less depending on processing power available) after a game has been completed, an 'automatic' analysis using the top chess engine(s), is performed on each game (i dont even know if this is possible).
Each player would then be given a percentage result ie: 15% 'move match' or in a bad case 95% 'move match' with the top engine(s).
Over a period of time patterns would become evident with the 'move match' percentages being constantly averaged over a period of time. A player with a high percentage would eventually have no opponent willing to play them. The best outcome for those that resort to having a computer play their games...
Originally posted by PKProblem is, there isn't enough computer horsepower to plow through
Please forgive me if this concept has already been discussed..
Perhaps for the top 100 players (more or less depending on processing power available) after a game has been completed, an 'automatic' analysis using the top chess engine(s), is performed on each game (i dont even know if this is possible).
Each player would then be given a percentage result ie ...[text shortened]... ng to play them. The best outcome for those that resort to having a computer play their games...
all the calculations needed for such a project. I think you would need
a supercomputer, teraflop or something. May be wrong, have been
known to be wrong on occasion🙂 You're talking thousands of games
all to be analyzed in something like real time. When I use fritz to
analyze a game the dang thing takes forever and I have a 2.4 GHZ
machine and thats only one game. I can't imagine trying to do that
for a whole site. Maybe the IBM big guns could, they are up to
135 Teraflops, 0.135 PETAflop machine. Not sure if even that would
be fast enough though. Maybe.
I don't think they'd need analysis by Fritz, just choose a move in about five seconds. If you get 90% in all your games, that would be suspicious in my mind. It can be done because some players (as we have seen) are already doing it in many, many games. Just use a program that punches out the moves just like they do.
Percentages aren't enough. Although they could be used as an alerting system for the games moderators. A crude example is if you have only one legal move - then everyone gets a 100% matchup with Fritz. The moderators will have had to have looked at the moves which weren't obvious - or forced - and determined whether there was a reasonable chance that a human could have found them unaided. They have undertaken a stunning task, but not one that can yet be automated. After all we would all like to be able to play tactically perfectly, but can't. It is not an easy matter to distinguish those who cheat from those who just work at it. David Tebb and Gatecrasher have done a lot of work and are to be applauded. They have to make an assessment as to whether a machine thought of the move or a human did, that kind of judgement is beyond a machine.
Ys, as I said, it would be suspicioius if you had a high correlation to Fritz, for example. Just having a high percentage is not proof. It's an indicator. The higher the percentage, the stronger the indicator. Once I played a blitz game and lost over internet. The game lasted about 45 moves. When I looked at it later with Fritz, I discovered my opponent had made ALL the recommended Fritz moves. So i am pretty sure he was using an engine. You can never be 100% sure of anything, but, just likein law, you have to go with the "reasonable doubt" theory. No one can be 100% sure an indicted person is guilty, but people are still found guilty. If there's a "reasonable doubt" i.e. a few moves are different or weaker then you might go with not guilty.
Originally posted by sonhouseAre you always this vague?
Problem is, there isn't enough computer horsepower to plow through
all the calculations needed for such a project. I think you would need
a supercomputer, teraflop or something. May be wrong, have been
known to be wrong on occasion🙂 You're talking thousands of games
all to be analyzed in something like real time. When I use fritz to
analyze a game t ...[text shortened]... 35 Teraflops, 0.135 PETAflop machine. Not sure if even that would
be fast enough though. Maybe.
Originally posted by sonhouseAccording to the player tables there are 8174 active players on this site. On average I think I finish about 5 games per week. That figure varies a bit so lets call it 10. Rounding up, ten thousand players all finishing 10 games per week gives you 100,000 games per week to crunch. I think this is a gross overestimate, but let's go with it. There are 604800 seconds in a week so that's 6 seconds per game. Even if I'm out by a factor of 10 that's 1 minute per game You are right there is no way it can be done without Russ splashing out on some serious hardware.
Problem is, there isn't enough computer horsepower to plow through
all the calculations needed for such a project. I think you would need
a supercomputer, teraflop or something. May be wrong, have been
known to be wrong on occasion🙂 You're talking thousands of games
all to be analyzed in something like real time. When I use fritz to
analyze a game t ...[text shortened]... 35 Teraflops, 0.135 PETAflop machine. Not sure if even that would
be fast enough though. Maybe.
The game lasted about 45 moves. When I looked at it later with Fritz, I discovered my opponent had made ALL the recommended Fritz moves. So i am pretty sure he was using an engine.Mmm...surely this means that the person was cheating.
When the Kramnik v Leko championship was going on, there was analysis provided by Fritz. OK, occasionally there were periods when the players made moves which corresponded with Fritz quite a bit, but very often they did not coincide (especially in endgames from my recollection).
Perhaps the endgame point reflects the fact that often these are quite open positions and there are often a number of options on the rank/file/diagonal that are equally good.
The point is that, if these guys aren't following the computer moves, it is even less likely that we mortal are. We will never catch cheating, however, because:
1) People could only use Fritz when the feel they really need to (i.e. once or twice a game).
2) They might deliberately sometimes choose the 4th/5th best move most of the time, when there were a variety of good moves available and in the sure knowledge that they would be avoiding a blunder.
However, that should not stop them getting booted off the site and I welcome the fact that they have been.
I wonder what is missing in these people's lives that they are happy to accept "recognition" from people that they have never met, and never will, and who only give them this recognition because they thought they were playing according to the rules?
Originally posted by CoconutTwo of the five players who had their accounts terminated were outside the top 100. If they were to introduce automatic monitoring of games then you'd have to check all the players on the site. There isn't a cut off where people suddenly stop cheating below a rating of 1,950. Domenic0987 had a peak rating of 1456.
That's nice...
maybe you should read the thread title.
We do have the processing power!! One of the most powerful resources is sitting right in front of each & every one of us with a combined network. It has been done already with data analysis on space noise looking for aliens??!!
This may be far fetched but who cares, it's only brainstorming...
If some software wiz like Fritz made available very cheap cut down checking software, then we call all do it on each and every game or whenever we so choose, or some other automatic system..
The mind boggles as to the possibilities.
OR something really simple like having a reminder permanently placed at the top of each screen. Something like 'cheaters will be...??'
Originally posted by PKI doubt that that's feasible but it would lead to the entertaining possibility that the cheats would be caught due to calculations happening on their own machines.
We do have the processing power!! One of the most powerful resources is sitting right in front of each & every one of us with a combined network. It has been done already with data analysis on space noise looking for aliens??!!
This may be far fetched but who cares, it's only brainstorming...
If some software wiz like Fritz made available very cheap cut down c ...[text shortened]... reminder permanently placed at the top of each screen. Something like 'cheaters will be...??'