Cheater or not - what do you think?

[dejan_lekic]

Let me first provide some context. - There was an interesting online tournament in the recent past with an open section and an U1700 section. Time control is G/60m with 10s/move increment.

The winner of the U1700 section, let call him PlayerB, has ~1560 ELO and had 7/7 performance. I passed all 7 games of his through lichess.org analysis and here is the summary of all 7 games: 10 inaccuracies 0 mistakes 2 blunders. Let me tell you something - if you do the same for Magnus Carlsen (pick any tournament) he would probably end up with worst performance than PlayerB.

Let's for the sake of argument compare this performance with the #1 of the open section, let's call him PlayerA... PlayerA has 2271 ELO rating. And had 5.5/7 performance. When I passed his games through lichess.org I was able to make the following summary: 16 inaccuracies 8 mistakes 3 blunders . My experience tells me that this is a realistic performance from 7 classic games of someone who is ~2200 ELO.

I humbly believe PlayerB cheated. He made one blunder when he was in time-trouble, and the other one he made could be deliberate...

What strikes me is that according to the arbiter, an anticheating software have been used and PlayerB passed the test(s) and it seems have been awarded the prize.

Anyway, if I am wrong (I hope I am), PlayerB will soon be a super GM if he continues like this.

[Matthew Turner]

Dejan,
You cannot tell anything from the details you have supplied. Adam Raoof has suggested that you put games through PGNspy, if you do this and present the evidence then people may be able to offer an opinion.

[Joseph Conlon]

Without the games, it isn't possible for us to say though. Low numbers of inaccuracies occur in games where one side gets a big advantage early on.
I've seen games by my <70 ECF son with 'no' inaccuracies or mistakes and very low acpl, because he was a queen up from move 6 and then that entirely distorts how these numbers are computed.

If all the games were strategically complex, then the figures you mention begin to sound implausible, but in and of themselves they don't definitively mean something either way.

[dejan_lekic]

Well, I did not want to offend anyone by giving names, etc. What I can tell you is that in one 42move (84ply) game PlayerB did not make a single mistake (inaccuracy, mistake or blunder). In another one, also 42moves he made just two tiny inaccuracies. I understand Joseph's point that when one blunders terribly the game ends up quickly and possibly without inaccuracies - this has not happened in this tournament. Statistically, PlayerB played like a super GM. If the games were played on lichess.com or chess.com I am almost completely sure he would be banned.

Finally, how many ~42move games you have in your databases without a single mistake (no matter how big) or just two inaccuracies? I have none out of ~300. (I am talking about my own games)

I did not participate in this tournament, but I know few players who did as I have played club games against them in the past...

[Ian Thompson]

Without the games, it isn't possible for us to say though. Low numbers of inaccuracies occur in games where one side gets a big advantage early on.
I've seen games by my <70 ECF son with 'no' inaccuracies or mistakes and very low acpl, because he was a queen up from move 6 and then that entirely distorts how these numbers are computed.

That sort of thing could apply to the player in question. Three of his games from the tournament were posted on this forum yesterday, before being deleted. I ran those games through PGNSpy. It said there were only 2 undecided positions in total:

PGNSpy Results.jpg

[JustinHorton]

Well, I did not want to offend anyone by giving names, etc. What I can tell you is that in one 42move (84ply) game PlayerB did not make a single mistake (inaccuracy, mistake or blunder).

I don't have a record of my games where I've managed this, but I do know that even playing games at 7+5, I have managed it a fair few times, and not all in very short games. But my Elo is a bit higher than player B's.

EDIT: And coincidentally here's one from a few hours after I posted, 46 moves played.

[Joseph Conlon]

Finally, how many ~42move games you have in your databases without a single mistake (no matter how big) or just two inaccuracies? I have none out of ~300. (I am talking about my own games)

A quick look gives an answer of several, as I found three out of the first 20 or so I looked at from my lichess account, restricting to games (a) I won (b) I had previously run the analysis tool on. These are either at blitz (3+2) or rapid (10+0) time controls; I don't know whether the analysis tool adjusts dynamically to the time control in terms of what it deems an inaccuracy or a mistake.

This isn't because, while stronger than 1700, I am particularly good in the grand scheme of things; it's more a reflection of the fact that, if you define an inaccuracy as a move which is more than one 'pawn' worse than the best move (or whatever the definition is), certain positions just lend themselves to there being fewer inaccuracies than others, and so one can play games with no mistakes and few inaccuracies - but the sharper the position, the harder it is to do this.

As has been commented on many times, in such positions there's still a huge cumulative difference between consistently playing the best move and playing the -0.4 pawn move, even though the latter is not an 'inaccuracy' and for any one position both seem completely reasonable choices for a human player to pick.

[JustinHorton]

if you define an inaccuracy as a move which is more than one 'pawn' worse than the best move (or whatever the definition is)

An estimation

[Alan Walton]

Agree it depends on the position and style of the game

I had one game, one of those where pieces come off and quickly into a pretty equal drawn endgame, my opponent made one slight mistake and converted it; my acpl was lower than 10 with no inaccuracies, mistakes or blunders; even my opponents acpl was very low because he only made one mistake out of 50 moves

[Joey Stewart]

Regardless of whether this player was using an engine or not they are still cheating by rating sandbagging as the performance clearly shows they are NOT u1700 strength.

[JustinHorton]

You obviously can't claim sandbagging without some knowledge of their previous history.

[Ian Thompson]

Regardless of whether this player was using an engine or not they are still cheating by rating sandbagging as the performance clearly shows they are NOT u1700 strength.

You obviously can't claim sandbagging without some knowledge of their previous history.

... which is:

canvas.png

He would have been 7 or 8 years old when he first got a FIDE rating.

Sandbagging seems unlikely to me.

[Nick Grey]

I would need a name.

Lots of games U1700.-

[Roger de Coverly]

He would have been 7 or 8 years old when he first got a FIDE rating.

Sandbagging seems unlikely to me.

So now 11 or 12. A breakthrough in chess understanding to a higher level isn't unknown at that age. For someone who suddenly "got good", players in an under 1700 tournament aren't the most testing opposition. A critical test may be to maintain that new level of play in other competitions.

[dejan_lekic]

EDIT: And coincidentally here's one from a few hours after I posted, 46 moves played.

I would expect a 2050+ ELO player to have such performance more frequently than 1560 ELO player. Being one I know how rare it is and how highly improbable it is to have two games without even a tiny mistake in 7 rounds... Also having NO games with mistakes (in lichess.org terms) in 7 rounds borders with impossible. But as I said, it is possible that the guy improved tremendously, which I doubt as he played in March and had under 1500 ELO performance...