Carlsen withdrawal after loss to Niemann

[CornfedForever]

The metric Chessbase uses does not include opening theory or forced moves ... so the fact that Bobby Fischer did quite well does seem fascinating as it means he played lines that today are theory and so he would be at a disadvantage. I really don't know how exactly it works ... but it is a metric that compares players against each other rather than checking a single game and seeing if there is cheating information. It is IMPOSSIBLE to check one game and tell if someone is cheating ... even if it is a wild tactical game that humans are terrible at ... because that one day a person can just see all the tactical shots. But when you correlate many games together, and then use a metric that is the same for all players ... the information is certainly valid! Hans in 2022 played in the Turkish league and was scoring about 40% ... all of a sudden he is playing 90%+ games and even 100% games ???!!!! Hikaru started pulling out his best games and was curious how they would do ... then he looked up Caruana's game where he had probably the most impressive run of wins in Classical chess ... and Caruana's score was in the. Hikaru couldn't find any game of Magnus or himself or Caruana ... at close to 90% ever. Yet we have Hans Nieman that has over 23 games over 90% in just 2 years and over 10 at 100%. I mean of course you can argue that unless you catch the guy with a phone or with a gadget then he isn't cheating. But I don't agree with that. If chess is going to be viable it has to find a way to weed out cheaters. In atheletics they use blood and urine samples ... sometimes they keep the samples for years and many atheletes that were cleared intially were stripped of their medals when the testing technology caught up with their cheating systems. I think with chess it is the same ... you can't take blood samples or urine samples, but you should be able to use statistics and metrics that are updated to catch cheaters. The probability that Hans did not cheat is becoming astronomically small ... add to that the fact that he has cheated before and it becomes hard to act like there is nothing there. I think even the most sceptical people that intitially thought that Hans is not a cheater are really finding it difficult to ignore the facts. Most people tend to think that everyone is honest and give them the benefit of the doubt, but the world is full of people that are dishonest!

I edited my original post to largely be more clear about something, but it looks like you were already replying to the original.

So, if theory in...lets say a mainline Dragon today leaves theory at move 20 but in Fischers day 'theory' (totally non-computer checked lines, of course) ended in that line at move 12, that means the chessbase utility is considering more of the actual moves in the Fischer game (lets say the average game lasts to move 50) than it would in someone playing a mainline Dragon today? Or am I getting that wrong? Perhaps it doesn't matter in the end as games may last longer today due to better defensive skills?

Oh, I watched Hikaru's broadcast, I think he found (of probably less than 10 he checked) that one of his got 100%...about 1 1/2 hrs into the broadcast.

[MonteCarlo]

I already posted earlier that there's at least one Magnus win reported on reddit that has this undefined Chessbase metric at 100%.

As much as people keep making completely unresearched claims about this (either that we would find a bunch for other players or that we'd never find them for anyone but cheaters), maybe I'm getting closer to making the foolish purchase of Chessbase to start collecting this data.

It's a shame it's so expensive and otherwise useless to me, or I'd do it in a heartbeat.

Cheers!

[syzygy]

The real question has for sometime been: any evidence of OTB cheating? Frankly...it sounds like Magnus has none and is doubling down on the 'online cheating' idea. I can't say he 'wins' with that argument.

I think the real question is if Hans' rating is anywhere close to his chess ability.

If he is really around 2700, why would he cheat online (or OTB).

I think that the general opinion of those who believe Niemann cheated OTB is that his real level is in the 2500 to 2600 FIDE range. The difference in what a young player can earn with a 2700 FIDE rating vs a 2600 FIDE rating is enormous, so there is no question about the motivation to cheat, it is much greater than say for a 2300 or 2400 player wanting to add 100 to his rating. I think it is very unlikely that anyone can get away with cheating enough to raise his rating (at a high level) by more than 200 elo, it would be too obvious. There is also not much reason for someone to cheat just to go from 2650 to 2660 (for example); the risk would outweigh the benefit, and anyway it would not be normal for a cheat to stop with such a small gain. So in general, when cheating is suspected, look for signs of play 100 to 200 above the real strength of the player. Of course determining that is not easy.

That sounds all very reasonable to me.

Also, purely based on numbers, it is more likely that you can find a player with 2500-2550 potential willing to cheat than a player in the 2675-2725 range, simply because there are many more player in the 2500-2550 range. Add to that the cost-benefit analysis.

[syzygy]

I already posted earlier that there's at least one Magnus win reported on reddit that has this undefined Chessbase metric at 100%.

We don't even know how this metric is calculated?

[MonteCarlo]

I already posted earlier that there's at least one Magnus win reported on reddit that has this undefined Chessbase metric at 100%.

We don't even know how this metric is calculated?

Correct. All we have is this snippet from old Chessbase documentation online http://help.chessbase.com/Reader/12/Eng ... t_menu.htm:

This value shows the relation between the moves made in the game and those suggested by the engines. This correlation isn’t a sign of computer cheating, because strong players can reach high values in tactically simple games. There are historic games in which the correlation is above 70%. Only low values say anything, because these are sufficient to disprove the illegal use of computers in a game. Among the top 10 grandmasters it is usual to find they win their games with a correlation value of more than 50%. Even if different chess programs agree in suggesting the same variation for a position, it does not mean that these must be the best moves. The current record for the highest correlation (October 13th 2011) is 98% in the game Feller-Sethuraman, Paris Championship 2010. This precision is apparent in Feller’s other games in this tournament and results in an Elo performance of 2859 that made him the clear winner.

Exactly how it's calculated is unknown as far as I can tell; the documentation above only says that it's some "relation". There are plausible guesses, of course, but no one participating in these "analyses" across the internet seem to really know.

If the calculation is done by seeing what percentage of your moves match each engine used and then picking the maximum of those, then that's one thing (i.e., if 100% percent means there's some single engine that plays every one of your moves in the game).

If, as it's starting to seem, it's just how often the played move matches the top line of ANY engine used (i.e., that you could get 100% by matching 50% of your moves to Olithink, 25% to LC0, 10% to SF, 10% to Brutus, and 5% to Scorpio), then it's a very different thing.

Then there's the question of whether multi-PV factors in, whether moves within some cp loss of the top move also count as a match, which moves are ignored, etc. A lot of unknowns.


Some of what's in the snippet is just wrong currently, like 98% being the record, as several 100s have been reported for non-Niemann games (Carlsen-Mamedyarov, cited earlier, the Naka game from his stream, and I've seen reports of some Nepo-MVL game, but haven't found that one yet, among others).

Maybe I will just buy Chessbase. It's never seemed like it will provide much use for me beyond what my collection of free software does, but who knows, maybe it'll have some value beyond this silliness.

Cheers!

[M ANSARI]

Oh, I watched Hikaru's broadcast, I think he found (of probably less than 10 he checked) that one of his got 100%...about 1 1/2 hrs into the broadcast.

Really ... I don't remember a game Hikaru had 100%. Maybe you mean that heI found one game from GM Erigaisi Arjun? It is true there was one game from the indian prodigy that had a 100% result but Hikaru quickly discarded that game when he found out it was only 10 moves where the other side resigned. If you look at Fabiano and Hikaru's analysis, they seem dumbfounded at how someone can play so many moves without an error. There is zero doubt that they both absolutely think Hans cheated. I did see one MC vs Nepo game that apparently scored 100% ... maybe that was prep for the WCC with 2 of the best chess players in the world. It was one game in their careers. Hans had 27 over 90% games and 17 at 100% in a 2 year period where he was averaging in the 40% in the Turkish league in 2022! Say what you want but those numbers are unusual and when you use that metric to compare Hans to other players he sticks out. As for your analysis of the game that you showed analysis on... I am sorry but really your computer analysis doesn't mean as much to me as Fabi or Hikaru's analysis. The player is supposed to be a human not an engine. A player doesn't have the luxury to see that there are "only" moves according to the engine. You think that locking a Knight like that where it has no escape is ok because 17 moves down the line you get one tempo extra and can queen a pawn before the other side when all heavy pieces are still on the board!???

[MonteCarlo]

Really ... I don't remember a game Hikaru had 100%...

https://clips.twitch.tv/FaintCuteKumqua ... Bw2xTJu_q5

Nakamura-Iturrizaga, Gibraltar Masters 2017

Cheers!

[CornfedForever]

Oh, I watched Hikaru's broadcast, I think he found (of probably less than 10 he checked) that one of his got 100%...about 1 1/2 hrs into the broadcast.

e unusual and when you use that metric to compare Hans to other players he sticks out. As for your analysis of the game that you showed analysis on... I am sorry but really your computer analysis doesn't mean as much to me as Fabi or Hikaru's analysis.

A quote of mine, but you have me confused with someone else. I offered no analysis.

[dkappe]

Oh, I watched Hikaru's broadcast, I think he found (of probably less than 10 he checked) that one of his got 100%...about 1 1/2 hrs into the broadcast.

Really ... I don't remember a game Hikaru had 100%. Maybe you mean that heI found one game from GM Erigaisi Arjun? It is true there was one game from the indian prodigy that had a 100% result but Hikaru quickly discarded that game when he found out it was only 10 moves where the other side resigned. If you look at Fabiano and Hikaru's analysis, they seem dumbfounded at how someone can play so many moves without an error. There is zero doubt that they both absolutely think Hans cheated. I did see one MC vs Nepo game that apparently scored 100% ... maybe that was prep for the WCC with 2 of the best chess players in the world. It was one game in their careers. Hans had 27 over 90% games and 17 at 100% in a 2 year period where he was averaging in the 40% in the Turkish league in 2022! Say what you want but those numbers are unusual and when you use that metric to compare Hans to other players he sticks out. As for your analysis of the game that you showed analysis on... I am sorry but really your computer analysis doesn't mean as much to me as Fabi or Hikaru's analysis. The player is supposed to be a human not an engine. A player doesn't have the luxury to see that there are "only" moves according to the engine. You think that locking a Knight like that where it has no escape is ok because 17 moves down the line you get one tempo extra and can queen a pawn before the other side when all heavy pieces are still on the board!???

As it turns out, this ChessBase function uses stored analysis from users over time. So it contains SF15, Komodo, Houdini, Rybka, Fritz 5, Fritz 17, Crafty, SF2 and everything in between. If a move matches any of those saved engine analyses, it “matches.”

Obviously, the shorter the game, the more likely it is to match 100%. The longer the game, the less likely there are going to be saved analyses of the later positions.

Another ridiculous chapter in this cheating scandal.

[Uri Blass]

But when Bobby Fischer in his career cannot get in the 90% and...

People keep repeating this!

Correct me if I am wrong, but the Chessbase function only begins AFTER LEAVING CURRENT THEORY, Right? So to "judge" or "compare" Fischer's games/stats 50-60 years later with this specific function is an abomination to the word 'analysis'. To do so...only proves, one thing... and (hint): it isn't what you are looking for.

Furthermore, people play a lot stronger at 'competency levels' these days and that is not even open to question.
Also, many of those old lines were...suspect or bad and would not be played today. That would just bring down Fischer's 'level' on its own.

There is more I could say, but...let me just end by saying that one can't judge Babe Ruth by the standard of play today in Baseball.

The metric Chessbase uses does not include opening theory or forced moves ... so the fact that Bobby Fischer did quite well does seem fascinating as it means he played lines that today are theory and so he would be at a disadvantage. I really don't know how exactly it works ... but it is a metric that compares players against each other rather than checking a single game and seeing if there is cheating information. It is IMPOSSIBLE to check one game and tell if someone is cheating ... even if it is a wild tactical game that humans are terrible at ... because that one day a person can just see all the tactical shots. But when you correlate many games together, and then use a metric that is the same for all players ... the information is certainly valid! Hans in 2022 played in the Turkish league and was scoring about 40% ... all of a sudden he is playing 90%+ games and even 100% games ???!!!! Hikaru started pulling out his best games and was curious how they would do ... then he looked up Caruana's game where he had probably the most impressive run of wins in Classical chess ... and Caruana's score was in the. Hikaru couldn't find any game of Magnus or himself or Caruana ... at close to 90% ever. Yet we have Hans Nieman that has over 23 games over 90% in just 2 years and over 10 at 100%. I mean of course you can argue that unless you catch the guy with a phone or with a gadget then he isn't cheating. But I don't agree with that. If chess is going to be viable it has to find a way to weed out cheaters. In atheletics they use blood and urine samples ... sometimes they keep the samples for years and many atheletes that were cleared intially were stripped of their medals when the testing technology caught up with their cheating systems. I think with chess it is the same ... you can't take blood samples or urine samples, but you should be able to use statistics and metrics that are updated to catch cheaters. The probability that Hans did not cheat is becoming astronomically small ... add to that the fact that he has cheated before and it becomes hard to act like there is nothing there. I think even the most sceptical people that intitially thought that Hans is not a cheater are really finding it difficult to ignore the facts. Most people tend to think that everyone is honest and give them the benefit of the doubt, but the world is full of people that are dishonest!

I suspect that percentage is connected not only to playing strength but also to playing style.
I suspect that the percentage of a player who is 100 or 200 elo stronger is not always better so I fo not get conclusions from the percentage of fischer.

You need to do the analysis for many players above 2600 of today to see if bigger percentgae is higher level and if 28xx is always stronger than a stable 2600-2650 who is not cheating.

Even I have games when I have no mistake in the endgame and kept a score of 0.00 for many moves
I am not sure what is going to be the percentage because maybe it is small because maybe my drawing moves are different than the drawing move that is suggested by the engines.

If it is not the case then maybe it is possible to increase the percentage if you continue to play obviously drawn endgames without offering a draw or accepting a draw.