To keep it fresh, the full report was released by chess.com:
I guess I should stop reading the Dope while I’m waiting for my opponent to move.
(To be clear, my win-loss record should be ample enough evidence that I’m not cheating. I think most people would watch me play and suggest I give cheating a try.)
This could make a huge difference. Carlsen has the chess.com report to back up his statements. He’s no longer making an isolated claim. I have my doubts about the report methodology but I’m sure there is more to come.
You’re fine as long as you always read the Dope between moves. I mean, you should probably concentrate on the game, but it won’t be detected as cheating.
If you switch tabs after every opponent’s move in a blitz game it would be very suspicious.
I imagine that they also have statistics that show that his better moves tend to be ones right after tab-switching.
Exactly.
In fact I think the auxiliary evidence is really just a form of “parallel construction.” They already know the cheat moves by how they compare against the general strength of the player and if they are the same ones an engine would suggest. But any one anomalous move could be a fluke. If those identified moves also correlate with tab-switching and other behavior (and the non-anomalous moves have no such extra behavior), though, that would be very strong evidence that the player indeed cheated on those.
At this level of play, no one is using an engine for all of their moves. They cheat when they recognize critical points. Maybe just once a game, or even less. Still, playing beyond their means for those critical points shows up in the stats.
It’s probably also possible to play exactly by an engine, but to program the engine itself to play suboptimally (but still above the level of one’s human opponent), in such a way as to disguise the fact that it’s an engine playing.
That would again require some degree of subtlety or cleverness on the part of the cheater… which so far has not been observed.
How do the engines work? Do they provide the best move only, or is there a scoring of several potential moves? If the latter, the cheater wouldn’t have to be particularly clever to pick the occasional 2nd-best move.
There could be players doing this and we haven’t “observed” it because the cheat filter is looking for someone who makes the best possible moves.
You can get a range of moves and their scores (not fixed numbers…they tend to wobble around a bit as they go through more potential moves).
Engines can give multiple lines with a relative strength for each move. Clearly only the dumbest of cheaters would pick the top engine line every move.
That’s not what they are looking for.
One way to think about it is that in chess (at least from a computer’s perspective) you can’t actually make “good” moves. You can only make “less bad” moves. Every move you personally make can only keep your odds of winning where they are or reduce them. There is a “best move”, but all it does is keep your odds of a win where they are. Every other move reduces those odds.
So what these algorithms do is try to determine how “bad” you play. How often do you reduce your winning chances by your mistakes, and by how much. Playing the “second best” move would not dramatically reduce your winning chances (unless it was much worse than the best move, in which case a cheater wouldn’t play it), so it would still be flagged as “abnormally strong play”.
Players at Niemann’s level can make nearly the best move most of the time. But making a mistake one or two fewer times a game than you “should” will show up in the numbers (at least over enough games). Because even the very best players (Magnus, for example) play much more poorly than the computers.
Your comment got me interested in just when this transition happened as I only had a vague memory of Gary Kasparov vs. Deep Blue. Came across this tidbit on wikipedia - The Ponomariov vs Fritz game on 21 November 2005 is the last known win by a human against a top performing computer under normal chess tournament conditions.
- Man…
For a sense of scale, the top chess program (Stockfish) has an Elo of 3534. Magnus Carlsen is at 2882. Elo works on differences, so Stockfish is as much better than Carlsen as Carlsen is to an Elo 2230 player.
Now, 2230 is a pretty good player, but Carlsen would essentially never lose to such a player. A win calculator says that he’d win 97.8% of the time and draw 2.2% (and lose 0.02% of the time). The same is true of Stockfish compared to Carlsen.
At least, in principle. The chess programs are so good that the Elo scores aren’t really even comparable. And that’s without even using a supercomputer; it’s with a fairly conventional quad-core PC. Since the chess programs are only competing against each other at this point, they just normalize to given CPU usage.
@glee here had a 2390 Elo at one point. It’s a good rating, but someone like Carlsen is at a whole other level. And Stockfish vs. Carlsen puts even that difference to shame.
Yes indeed - in my (amateur) career over many years, I’ve never beaten a Grandmaster (drawn with a few though. 
)
For a couple of years I was rated in the top 4,000 in the World…
Personally I think cheating is despicable whether or not money is at stake.
In one British Championship I had made my move and was walking around the tournament hall. I encountered the top rated GM Speelman (who I knew.) I muttered “Hello” - he ignored me. 
After both our games had finished, Speelman came up and explained “Sorry - I couldn’t talk to you in case other player thought you were giving me advice!”
This shows:
- Speelman is a gentleman (and completely honest)
- how players should behave during an event
- Speelman had a slightly exaggerated idea of my ability
Impressive! That’s a higher ranking than I’d intuitively expect. Assuming the scores scale smoothly, it sounds like there are roughly 1/5 as many players as you go up 100 Elo points. Does that sound approximately true?
Speelman does sound like a gentleman. Cheating should be taken very seriously. If Niemann’s original claims had been true–that he had only cheated a couple of times online, years ago, then it is perhaps forgivable, given his age. But it appears the cheating is far more widespread and recent than he admitted. I’m somewhat skeptical that there was any OTB cheating due to practical concerns, but to my mind Niemann doesn’t deserve any benefit of the doubt at this point.
Thanks for chiming in!
My peak was back in the 1970s, so there has been a massive rise in the number of players since then (especially in countries like China and India.)
Also the overall standard of chess has risen, due to players starting younger and computers providing analysis (especially in endings) and practice opponents.
IIRC, there is potential for inflation or deflation of Elo ratings over time because the nature of the pool of points (there’s a “floor”, below which players aren’t part of the pool, so the distribution can become a little wonky over time) and FIDE “injects” points into the rating system to keep it balanced, but how many and at what rate is tricky.
I vaguely remember a colleague analyzed this phenomenon and telling me that without corrections, the system wouldn’t be stable over time. He was a tournament pool player, one day we were talking about competitive pool, and he wondered if there was a way to do a better job of seeding players in a pool tournament. I had just gotten my TD card from USCF, so I told him how we did it in chess, since both sports are one on one matches. Among other things (I tried to persuade him that adopting Swiss System would be more efficient) I mentioned Elo ratings as a method for assessing relative strength. Subsequently, I gave him a book I had on the Elo system, and he went off and actually adapted it for pool, which involved analysis and simulation on his part. That’s where his assessment of inflation/deflation came from.
And, after a little googling, I realize I was partially responsible for the current method of rating pool players!
Do you think it’s possible that there was cheating in the live match? If so, how could it have been achieved?
Well it’s certainly possible.
- Computers are much stronger than humans at chess.
- Chess moves are easy to encode (4 characters will do e.g. e2e4.)
- So you only need an accomplice running a suitable program, plus a method to transmit information to the player.
- This method could be concealed in a shoe (or even apparently anal beads
), or line of sight. - Having said that, the tournament referee does know what to look out for.
- One way to spot cheating is if a player chooses the top computer move throughout the game (in any position, computers give you two or three best moves - ranked in order), since it’s unheard of for a human to be able to do this.
- Accomplices can slightly avoid a succession of top computer moves by offering the second best occasionally. Also moves that are forced (the player is in check) or that involve recapturing are not evidence of computer assistance.
From what I’ve read, my opinion is:
- since Niemann has admitted cheating a couple of times and there is a strong claim (by chess.com based on top computer moves in Niemanns’ online games) that he has done so many other times, I think he did cheat
- it would be jolly helpful to find out what method he used