Evaluations Wanted

I believe that the cheaters will disappear automatically since there is no fun in the long run to play that way, a dying race, so no need to worry really. And sometimes it is not a good idea to follow a computers suggestion because computers are weaker than human beings when it comes to constructing meaningful long-term strategic plans. On the other hand a computer program is tactically superior, and much faster than the human brain. IBM´s chesscomputer Deep Blue was capable of evaluating 200 million positions per second. Nevertheless Kasparov beated it a couple of times, due to better strategic thinking, and of course also excellent human tactical brainpower.

Based on the evidence, I'd say the player is not playing by the rules.  Unless I'm missing something, it's pretty obvious.

And sometimes it is not a good idea to follow a computers suggestion because computers are weaker than human beings when it comes to constructing meaningful long-term strategic plans.

---

I see this said from time to time in discussions like this. I have to wonder at the people who say it. They are obviously regurgitating something they have seen Grandmasters say, which is fine...

But practically speaking for almost everyone on this site.. this is hogwash. I see people bashing Computer Program's ability to play endgames and strategize in closed positions all the time.. but the fact is computers strategic play is better than your average player. It is better than your average club player... and in my opinion it is stronger (strategically) than your average IM.

Regarding the original topic of this thread... When I used to use Chessmaster 8000... I used to analyze Grandmaster games and it was extremely rare for either player (winner or loser) to have 100% agreement with chessmaster. The human element rears its ugly head in almost every game. Those were of course OTB games from grandmasters and not Correspondence games... but even amongst very strong players like GMs... 99.9% agreement through 6 games is highly unlikely.

Incidentally, my efforts to channel Capablanca with my Ouija board (this is not specifically prohibited in the rules) have met with less than desirable results. I am either not dealing with the Capablanca I think I am, or he has really let his chess skills go in the dark yonder.

oginschile wrote:...Incidentally, my efforts to channel Capablanca with my Ouija board (this is not specifically prohibited in the rules) have met with less than desirable results. I am either not dealing with the Capablanca I think I am, or he has really let his chess skills go in the dark yonder.

---

This may be the best thing I've seen in the last month on Chess.com.

sorry to sound thick, or to add confusion, Batgirl, is your thread mainly about someone cheating ? or about the computer/computer analysis of chess engines

Just to confirm what someone has already said on this thread, computer analysis (I'm using Fritz as an example here, but I assume ChessMaster works similarly) can be adjusted to increase or decrease the number of comments by setting a 'mistake' level.  Using 100 as the value for a pawn, if the level is set at 100, then Fritz will inform you if the move played in the game is worse than the move it thinks is best by more than a pawn's value.  If it's worse, but by less than a pawn, it won't suggest it's improvement in the analysis.

 

This is the most likely explanation for the high degree of accuracy you describe.  Even computers suggest different moves in the same positions, so unless your suspect is using exactly the same version of ChessMaster, it seems rather unlikely they are cheating. 

The circumstances outlined by batgirl are indeed curious. I suspect chicanery.

My evaluation from this thread? The average person isn't too good with percentages.

 

There wree 6 games analyzed with a maximum game length of 37 moves. So at most 222 moves were analyzed (there were actually fewer, but we don't know the exact number). 221 correct moves out of 222 is 99.6%, not 99.9% or 99.99%.  Since there were somewhat less than 222 moves analyzed and we shouldn't count moves that are in the opening book, I'd estimate 150 moves were really checked, putting the percentage correct around 99.3%.

 

On a more serious note, the percentage of moves that agree with a computer analysis depends greatly on how weak the opposition is. It is not so difficult to play flawlessly when your opponent consistently makes tactical errors. I think it's difficult to make these kinds of judgements based on statistics alone. You would have to examine particular moves. For example, there are times when a human plays a winning move that leads to a simple win where a computer would play a winning move that it believes to be stronger but leads to a more complex position.

Here's an obvious mathematical item not yet brought up, with an old name of the Peter Principle.

 Everyone start with a 1200 rating, and gets matched with fairly like-rated folks. If your playing ability is quite a bit above that, you're going to win a lot of games until your rating here reaches it's approriate level. In fact, it's likely to overshoot and correct.

My own profile shows something like 9-10 straight wins and then a bunch of losses as the level of competetion went up, and now I'm something like 12-5-1.

 At the same time, the chance that my move choices match Chessmaster's 99.99% of the time is zero. I don't think ever I've had a game (nor do many with ratings below 1800) without a tactical error, whether blunder or more subtle.

RetGuvvie98 wrote: silentfilmstar13 wrote: oginschile wrote:...Incidentally, my efforts to channel Capablanca with my Ouija board (this is not specifically prohibited in the rules) have met with less than desirable results. I am either not dealing with the Capablanca I think I am, or he has really let his chess skills go in the dark yonder.

---

This may be the best thing I've seen in the last month on Chess.com.

---

Concur wholeheartedly with silentfilmstar's praise of Oginschile's last comment.

 

---

I also agree that this was the funniest comment I have read on Chess.com since it started.  I suggest that oginschile resigns his current games with me asap and becomes a comedy writer instead.

"I suggest that oginschile resigns his current games with me asap and becomes a comedy writer instead."

 

No such luck. I'm gonna force you to mate me in 25 moves like always happens.

Rael wrote:

 

I can't find "that person" anywhere on the site. Are you sure that's their username, or did erik already ban him or her?

Sorry. I tried to resist.

---

 Nice...that's like when Homer Simpson was working from home and reading directions for his job on how to work on his computer: "Next step, press any key.  Where's the 'Any' key?"

In this thread I've been exposed as a vacuous hack writer incapable of asking a direct question as well as being an incompetent computer cheater. Fortunately, I've also been tele-psychoanalzed as an exhibitionist, so being exposed may not be that big a deal. On the bright side, I've also had some pretty good responses by various individuals of sound and rational minds which have given me a little more insight into the use of computer analysis in certain situations.

 

I'd now like to look at this from a slightly different perspective with an actual direct question.

 

If you were to use a computer program(s) to identify whether the moves of a particular game were generated by the human mind or by a computer algorithm,  exactly what in the analysis would you consider indicative of computer use?

 

 

On another site this is (was) done using a statistical method. Essentially, the games are run through Fritz or some other engine using a script to automate the process. The matchup rate between the moves played and what Fritz thinks are the top three moves is noted. The details are obscure (deliberately I think) but if percentage of player moves matching Fritz reaches a certain threshold (about 85% I think) then that is cause for suspicion. On its own, this is not actually enough to condemn someone. The games are then examined by humans who determine whether it is likely that a human player would choose those moves. In some positions there will be one clearly best (or forced) move where anything up to 100% match would not be really surprising. In others there may be very little to choose between several moves and a player consistently matching engine choices could then be suspected of cheating.

 

Other factors that will affect things are the software used, the hardware it is run on, how the engine is set up (time for move, search depth etc.). It is also no use analysing opening moves, they may well come from the engine's opening book, they could equally well come from a database or book. 

Chessmaster.....is......ALIVE!!!!

I remember seeing an analysis done by people who seemed to know what they were doing. Some may recall a few years ago people thought Fischer was playing blitz games online. He destroyed the best online talent, including a few blitz games with Nigel Short who apparently asked him some questions and was convinced that it was indeed Fischer.

The long and short of it was they determined the user was actually using an engine... chess tiger if i recall... but to throw any testers off... the user was displaying not only the best move but the second best move as well... and would at times make the second best move. In a blitz game, the second best move is often good enough to still win.

This certainly doesn't answer your direct question (Excellent job by the way directizing your questination), but I thought it was interesting the depths it can take to catch a wily cheater.

I just always figure if someone beats me, they probably cheated. No investigation required.

About 10 years ago  saw a test that had been given to various chess programs as well as to humans of different strengths.  The test comprised of a position in which White, who already had a considerable material advantage, could capture Black's knight with a pawn.  However, taking the Knight would lock up the board in such a way that White could never break through, effectively dooming the game to a draw in 50 moves.  Without exception, the computers, unable to see that far, took the Knight. Even the weakest human in the test was able to understand the Knight was taboo and went on to win the game.  Of course, this test may not be as valid today but it does illustrate that there's a difference in how computers and humans see moves.  The best move determined even by a bevy of programs, may or may not be the best move for a human.  For a computer, the best move is one that satisfies all the subtleties of a situation. For a human, the best move is one that satisfies all the subtleties of a situation and furthers a plan.  So the criteria for a best move is different for computers and for humans;   different criteria means likely different choices.

Interesting.  Can you tell if someone is using a computer program to win their games?  You may not have enough games to be sure of your conclusions. Please read on... In my own games played here, I had pretty good results so far, and, I think, better than I would usually do in OTB chess.  I used my database (permitted by the rules), and so my games would show that the computer agrees with all of my moves up to a certain point.  I was very careful, I used the time I needed to make my best moves, and I made sure I didn't move when I was tired or after having a glass of wine. I used the analysis board extensively to attempt to better understand the positions which arose. A lot of my opponents were lower rated than me (my eventual rating), and they made serious errors that (I believe) had limited refutations.  Again, the computer would agree with those moves, unless I was wrong.  After playing my games (emphasis on AFTER), I put all of them thru Fritz11 to see how well I did.  But, and here is the point of my post, I noticed that Fritz 11's analysis mode did not "disagree" (this comes in the form of a yellow or red light instead of a green one) unless there was a loss of more than ".5" (half a pawn).  The increments of losses and gains were in fractional values, and it depended on a mix of factors analyzed by the program.  So that means that I could make a lot of moves in a row that individually got Fritz's "okay", but combined would result in considerable cumulative losses of positional and material points (and the game).  As a result of this realization, I gave up trying to evaluate my play based on the "agreement" of Fritz alone.  What I started working on was games without "red lights" showing up which meant a really bad move.  I found that if I worked hard and really paid attention, and if I was playing a weaker player, that was usually possible.  In fact, I found that in some (perhaps many) positions, there could be as many as 8 or 10 different moves that would be within the Fritz paramaters for a green light. However, a grandmaster might not choose all of them as acceptable moves. To assume that a grandmaster would choose any move that the computer program "agrees with" (in context of this discussion) is a flawed idea, in my opinion. So my success at these efforts to play well could also make me look like a big-fat-cheater if someone was to analyze the games looking only for computer "agreement" with my moves. By the way, one thing I discovered was that I tend to let up after I think I am winning (allowing a possible turning-of-the-tables).  Again, this is a factor in the analysis of the games that could lead one to conclude that cheating was done to get the winning positions, even if it was not the case.  Bottom line is that there could be more to this kind of analysis than meets the eye, and you may need a bigger "population" of games for your study.  Then you may need to use a different criteria than simple computer "agreement".  When I was a young man, I accused a fellow soldier of stealing $20 from my wallet, left unattended on my bunk for a while.  Later I found the $20 where I had forgotton that I stuffed it into some socks to hide it from barracks thieves.  My apology did little to undo the damage I had done by making a hasty accusation.  I have never been able to forget how I felt about that and how I made the other guy feel. I think we should make sure we know what we are talking about before we make accusations against other people.  In the final analysis, I agree with sondredanielsen when he said that in the long run, cheating is not fun, and that cheaters will die out over time on their own.  In the meantime, it is a fantastically interesting subject, and it is just as creepy as it is interesting to think that if you play well, maybe better than expected, instead of being congratulated, you may be accused.

In Game Over: Kasparov versus the Machine, Kasparov made it very clear that computer style of play opposed to humans is like apples and oranges.

"The computer played like a machine, it was lousy, well not lousy, it did not pay any attention to King saftey."

He also said some things that computers do not have deep strategic thinking like  humans and after watching the documentry it was pretty obvious that humans do not play like computers and it would be unatural to do so, so if this person plays in aggreement with Chessmaster 90% of the time you can bet your boots that he/she is cheating.

Thanks for your input, Cleptomania.

 

I never had any intention of accusing anyone.  I came upon a situation that aroused my suspicions and piqued my curiosity as to whether I could somehow make an infallible determination concerning that person - a personal challenge, not a public one.   I wasn't sure what the limited computer analysis actually revealed,  so I put the issue in a public forum and received a crash course in compter analysis.   As a rule,  I don't believe that worrying about cheaters is worth the paranoia and too much "protection" or surveillance might actually be worse than none at all.   Still,  as an academic exercise,  I would find it interesting to learn the indicative signs that differentiate your opponent's moves as human or computer-generated.

 

now, may I have my wallet back?