Carlsen withdrawal after loss to Niemann

[Alexander Schmidt]

why Carlsen did NOT accused him of cheating

I guess we answered this like 100 times...

[Ajedrecista]

Hello Larry:

Well, 14 games isn't "nothing", it has a standard deviation of something like 40 elo or so I think (depending on details of opposition) so probably 2600 means between 2560 and 2640 (almost certainly between 2520 and 2680). Yes, there is some impact of the story on results, though it could be in either direction and is probably not large; in my experience chess strength dominates over factors like fatigue, health, attitude, etc. over a series of games. The totality of the anti-cheating measures at the U.S. Championship make it very unlikely that anyone would even try to cheat and expect to get away with it, in my opinion. Maybe in the future someone could make a device that would be undetectable, but with no advance warning of the specific measures I doubt that happened.

I would say that a single standard deviation after 14 games is worth more than ± 40 Elo. Taking the formula of the sample standard deviation from the trinomial distribution:

Code: Select all

s = sqrt{[score*(1 - score) - (draw_ratio)/4]/(games - 1)}

Supposing score = 50%, draw_ratio = 50% and games = 14, I get s ~ 0.0976, then ± 400*log10[(score + s)/(score - s)] ~ ± 69 Elo and ± 400*log10[(score + 1.96*s)/(score - 1.96*s)] ~ ± 140 Elo.

Applying to Niemann's case, taking the info from 2700chess.com right now with one more game (Niemann - So, ECO C53) after the infamous Carlsen - Niemann:

Code: Select all

Niemann's latest games, reverse chronological order:

 #     White Player     Rating     Black Player     Rating     Result     Moves     ECO     Site     Year
 1     Niemann           2699      So                2774	½-½         27      C53	 Saint Louis 2022
 2     Aronian           2755      Niemann           2699	½-½         57      C58	 Saint Louis 2022
 3     Swiercz           2652      Niemann           2699	1-0         28      E00	 Saint Louis 2022
 4     Niemann           2699      Robson            2690	0-1         35      A50	 Saint Louis 2022
 5     Domínguez Pérez   2747      Niemann           2699	½-½         37      C67	 Saint Louis 2022
 6     Niemann           2699      Caruana           2763	0-1         80      E32	 Saint Louis 2022
 7     Shankland         2712      Niemann           2699	½-½         57      E32	 Saint Louis 2022
 8     Niemann           2699      Xiong             2690	½-½         43      C11	 Saint Louis 2022
 9     Yoo               2563      Niemann           2699	0-1         55      A45	 Saint Louis 2022
10     Nepomniachtchi    2792      Niemann           2688	½-½         54      A21	 Saint Louis 2022
11     Niemann           2688      Caruana           2758	0-1         69      A21	 Saint Louis 2022
12     Vachier-Lagrave   2757      Niemann           2688	½-½         55      A07	 Saint Louis 2022
13     Niemann           2688      So                2771	0-1         59      C42	 Saint Louis 2022
14     Domínguez Pérez   2745      Niemann           2688	½-½         47      C55	 Saint Louis 2022
15     Niemann           2688      Firouzja          2778	½-½         43      C55	 Saint Louis 2022

Average opposition rating = 40949/15 ~ 2729.93 Elo; WDL performance → 1 win, 9 draws and 5 loses:

Code: Select all

score = (1 + 9/2)/15 = 11/30 ~ 36.67%
draw_ratio = 9/15 = 0.6 = 60%

s = sqrt{[score*(1 - score) - (draw_ratio)/4]/(games - 1)}
s ~ 0.0766356

1 sample standard deviation:
2729.93 + 400*log10[(score - s)/(score + s)] ~ 2729.93 - 155.52 ~ 2574.42 Elo
2729.93 + 400*log10[(score + s)/(score - s)] ~ 2729.93 -  39.57 ~ 2690.37 Elo
Gap of 115.95 Elo.

1.96 sample standard deviations:
2729.93 + 400*log10[(score - 1.96*s)/(score + 1.96*s)] ~ 2729.93 - 223.47 ~ 2506.46 Elo
2729.93 + 400*log10[(score + 1.96*s)/(score - 1.96*s)] ~ 2729.93 +  11.73 ~ 2741.66 Elo
Gap of 235.20 Elo.

Roundings up to 0.01 Elo.

I hope no typos.

Rounding up to multiples of 5 Elo to get a rough estimate, Niemann's rating performance has been around [2575, 2690] Elo (gap of 115 Elo) for circa 68% confidence level and [2505, 2740] Elo (gap of 235 Elo) for circa 95% confidence level.

15 games are too few to draw any conclusions, but overall I am happy that cheating is being considered a serious threat for our wonderful game. Better anti-cheating measures will be taken over time as more knowledge on the subject will be available.

Regards from Spain.

Ajedrecista.

[dkappe]

Hello Larry:

Well, 14 games isn't "nothing", it has a standard deviation of something like 40 elo or so I think (depending on details of opposition) so probably 2600 means between 2560 and 2640 (almost certainly between 2520 and 2680). Yes, there is some impact of the story on results, though it could be in either direction and is probably not large; in my experience chess strength dominates over factors like fatigue, health, attitude, etc. over a series of games. The totality of the anti-cheating measures at the U.S. Championship make it very unlikely that anyone would even try to cheat and expect to get away with it, in my opinion. Maybe in the future someone could make a device that would be undetectable, but with no advance warning of the specific measures I doubt that happened.

I would say that a single standard deviation after 14 games is worth more than ± 40 Elo. Taking the formula of the sample standard deviation from the trinomial distribution:

Code: Select all

s = sqrt{[score*(1 - score) - (draw_ratio)/4]/(games - 1)}

Supposing score = 50%, draw_ratio = 50% and games = 14, I get s ~ 0.0976, then ± 400*log10[(score + s)/(score - s)] ~ ± 69 Elo and ± 400*log10[(score + 1.96*s)/(score - 1.96*s)] ~ ± 140 Elo.

Applying to Niemann's case, taking the info from 2700chess.com right now with one more game (Niemann - So, ECO C53) after the infamous Carlsen - Niemann:

Code: Select all

Niemann's latest games, reverse chronological order:

 #     White Player     Rating     Black Player     Rating     Result     Moves     ECO     Site     Year
 1     Niemann           2699      So                2774	½-½         27      C53	 Saint Louis 2022
 2     Aronian           2755      Niemann           2699	½-½         57      C58	 Saint Louis 2022
 3     Swiercz           2652      Niemann           2699	1-0         28      E00	 Saint Louis 2022
 4     Niemann           2699      Robson            2690	0-1         35      A50	 Saint Louis 2022
 5     Domínguez Pérez   2747      Niemann           2699	½-½         37      C67	 Saint Louis 2022
 6     Niemann           2699      Caruana           2763	0-1         80      E32	 Saint Louis 2022
 7     Shankland         2712      Niemann           2699	½-½         57      E32	 Saint Louis 2022
 8     Niemann           2699      Xiong             2690	½-½         43      C11	 Saint Louis 2022
 9     Yoo               2563      Niemann           2699	0-1         55      A45	 Saint Louis 2022
10     Nepomniachtchi    2792      Niemann           2688	½-½         54      A21	 Saint Louis 2022
11     Niemann           2688      Caruana           2758	0-1         69      A21	 Saint Louis 2022
12     Vachier-Lagrave   2757      Niemann           2688	½-½         55      A07	 Saint Louis 2022
13     Niemann           2688      So                2771	0-1         59      C42	 Saint Louis 2022
14     Domínguez Pérez   2745      Niemann           2688	½-½         47      C55	 Saint Louis 2022
15     Niemann           2688      Firouzja          2778	½-½         43      C55	 Saint Louis 2022

Average opposition rating = 40949/15 ~ 2729.93 Elo; WDL performance → 1 win, 9 draws and 5 loses:

Code: Select all

score = (1 + 9/2)/15 = 11/30 ~ 36.67%
draw_ratio = 9/15 = 0.6 = 60%

s = sqrt{[score*(1 - score) - (draw_ratio)/4]/(games - 1)}
s ~ 0.0766356

1 sample standard deviation:
2729.93 + 400*log10[(score - s)/(score + s)] ~ 2729.93 - 155.52 ~ 2574.42 Elo
2729.93 + 400*log10[(score + s)/(score - s)] ~ 2729.93 -  39.57 ~ 2690.37 Elo
Gap of 115.95 Elo.

1.96 sample standard deviations:
2729.93 + 400*log10[(score - 1.96*s)/(score + 1.96*s)] ~ 2729.93 - 223.47 ~ 2506.46 Elo
2729.93 + 400*log10[(score + 1.96*s)/(score - 1.96*s)] ~ 2729.93 +  11.73 ~ 2741.66 Elo
Gap of 235.20 Elo.

Roundings up to 0.01 Elo.

I hope no typos.

Rounding up to multiples of 5 Elo to get a rough estimate, Niemann's rating performance has been around [2575, 2690] Elo (gap of 115 Elo) for circa 68% confidence level and [2505, 2740] Elo (gap of 235 Elo) for circa 95% confidence level.

15 games are too few to draw any conclusions, but overall I am happy that cheating is being considered a serious threat for our wonderful game. Better anti-cheating measures will be taken over time as more knowledge on the subject will be available.

Regards from Spain.

Ajedrecista.

A little homework for you: repeat this for all 2450+ rated players active OTB for the last 3 years. Thanks. I look forward to your results.

[CornfedForever]

https://www.chess.com/blog/CHESScom/com ... rt=content

A new chess.com statement. Nothing really new...just doing what you would expect, being public with what they would want to be seen in the future...be it legal or otherwise.

They of course distance themselves from Magnus and any 'OTB cheating' claims. I'm sure the lawyers continue to be happy.

[Eduard]

Hans won today.

[Chessqueen]

Hans won today.

Are you insinuating that GM Hans only cheat when he wins, but not when he lose, it could also be that other players are using a better engine than Niemann when he lose ?

[pgn][Event "USA Championship 2022"]
[Site "Saint Louis"]
[Date "2022.10.15"]
[Round "10.4"]
[White "Moradiabadi, Elshan"]
[Black "Niemann, Hans Moke"]
[Result "0-1"]
[BlackElo "2699"]
[ECO "E61"]
[Opening "King's Indian"]
[Variation "Smyslov System, 5...c5"]
[WhiteElo "2534"]
[TimeControl "480+3"]
[Termination "normal"]
[PlyCount "76"]
[WhiteType "human"]
[BlackType "human"]

1. d4 Nf6 2. Nf3 g6 3. c4 Bg7 4. Nc3 O-O 5. Bg5 c5 6. d5 d6 7. e3 Nbd7
{E61: King's Indian: Early deviations for White, including Smyslov System.}
8. Be2 h6 9. Bh4 g5 {The position is equal.} 10. Bg3 Nh5 11. Qc2 {Strongly
threatening Nxg5!} Ndf6 12. Nd2 a6 13. a4 Bd7 {[#]} 14. a5 {$0 $146}
{Predecessor:} (14. O-O Ng4 15. Nde4 Bf5 16. Nf6+ Ngxf6 17. Qxf5 Qc8 18.
Qd3 Nxg3 19. hxg3 Ng4 20. Qe4 {1-0 (51) Agrest,E (2591)-Bologan,V (2663)
Istanbul 2003}) 14. .. Rb8 15. O-O Nxg3 16. hxg3 Ng4 17. Qd1 f5 18. e4 Ne5
19. exf5 Bxf5 20. Nf3 Qe8 21. Nxe5 Bxe5 22. Bh5 {Against Qg6} Bg6 23. Bxg6
Qxg6 24. Qe2 Rf7 25. Ne4 {And now Nxc5! would win.} Bd4 26. Ra3 g4 $1 27.
Rd3 (27. Rb3 {$0 $11}) 27. .. Rf5 (27. .. b6 {$0 $17} 28. axb6 Rxb6) 28.
Qd2 {$0 $11} Rbf8 29. b4 $1 Re5 30. bxc5 dxc5 31. Rxd4 cxd4 32. Nc5 {Black
must now prevent Nd7.} Ref5 (32. .. Qf5 {$1 $11}) 33. Qxd4 {$0 $16} h5
{[#]} 34. Ne6 {Weaker is} (34. Nxb7 34. .. h4 35. Qe3 (35. gxh4 $2 g3 {$0
$19}) 35. .. Qh5 {$0 $15}) ({Better is} 34. Qe4 {$1 $16}) 34. .. R8f6 (34.
.. R8f7 {$0 $11}) 35. Nf4 (35. Qb6 {$0 $16}) 35. .. Qh7 {$1 $14} 36. Qa7 h4
37. Qxb7 (37. Re1 {$0 $11}) 37. .. hxg3 {$0 $15 [#] . The board is on
fire.} 38. fxg3 $4 (38. Qc8+ {$1 $15} Kf7 39. fxg3) 38. .. e5 {$1
Discovered Attack. Weighted Error Value: (very precise)} 0-1[/pgn]

[lkaufman]

Hans won today.

Are you insinuating that GM Hans only cheat when he wins, but not when he lose, it could also be that other players are using a better engine than Niemann when he lose ?

[pgn][Event "USA Championship 2022"]
[Site "Saint Louis"]
[Date "2022.10.15"]
[Round "10.4"]
[White "Moradiabadi, Elshan"]
[Black "Niemann, Hans Moke"]
[Result "0-1"]
[BlackElo "2699"]
[ECO "E61"]
[Opening "King's Indian"]
[Variation "Smyslov System, 5...c5"]
[WhiteElo "2534"]
[TimeControl "480+3"]
[Termination "normal"]
[PlyCount "76"]
[WhiteType "human"]
[BlackType "human"]

1. d4 Nf6 2. Nf3 g6 3. c4 Bg7 4. Nc3 O-O 5. Bg5 c5 6. d5 d6 7. e3 Nbd7
{E61: King's Indian: Early deviations for White, including Smyslov System.}
8. Be2 h6 9. Bh4 g5 {The position is equal.} 10. Bg3 Nh5 11. Qc2 {Strongly
threatening Nxg5!} Ndf6 12. Nd2 a6 13. a4 Bd7 {[#]} 14. a5 {$0 $146}
{Predecessor:} (14. O-O Ng4 15. Nde4 Bf5 16. Nf6+ Ngxf6 17. Qxf5 Qc8 18.
Qd3 Nxg3 19. hxg3 Ng4 20. Qe4 {1-0 (51) Agrest,E (2591)-Bologan,V (2663)
Istanbul 2003}) 14. .. Rb8 15. O-O Nxg3 16. hxg3 Ng4 17. Qd1 f5 18. e4 Ne5
19. exf5 Bxf5 20. Nf3 Qe8 21. Nxe5 Bxe5 22. Bh5 {Against Qg6} Bg6 23. Bxg6
Qxg6 24. Qe2 Rf7 25. Ne4 {And now Nxc5! would win.} Bd4 26. Ra3 g4 $1 27.
Rd3 (27. Rb3 {$0 $11}) 27. .. Rf5 (27. .. b6 {$0 $17} 28. axb6 Rxb6) 28.
Qd2 {$0 $11} Rbf8 29. b4 $1 Re5 30. bxc5 dxc5 31. Rxd4 cxd4 32. Nc5 {Black
must now prevent Nd7.} Ref5 (32. .. Qf5 {$1 $11}) 33. Qxd4 {$0 $16} h5
{[#]} 34. Ne6 {Weaker is} (34. Nxb7 34. .. h4 35. Qe3 (35. gxh4 $2 g3 {$0
$19}) 35. .. Qh5 {$0 $15}) ({Better is} 34. Qe4 {$1 $16}) 34. .. R8f6 (34.
.. R8f7 {$0 $11}) 35. Nf4 (35. Qb6 {$0 $16}) 35. .. Qh7 {$1 $14} 36. Qa7 h4
37. Qxb7 (37. Re1 {$0 $11}) 37. .. hxg3 {$0 $15 [#] . The board is on
fire.} 38. fxg3 $4 (38. Qc8+ {$1 $15} Kf7 39. fxg3) 38. .. e5 {$1
Discovered Attack. Weighted Error Value: (very precise)} 0-1[/pgn]

This was a very simple but understandable one-move blunder by Elshan. It is due to the special feature of en-passant, that it is only good on the next move. Since White can take en-passant, he dismissed 38...e5 as bad, but unfortunately he can't play the check on c8 after 38...e5 and then take en-passant, he has to check first. I don't recall ever seeing a blunder based on this point before. So just a normal game by Hans with a lucky twist at the end. There is not the slightest indication that any player is cheating in this event, and it would be extremely unlikely with the precautions taken. Let's look on this as a fair test for all 14 players.

[ernest]

I would say that a single standard deviation after 14 games is worth more than ± 40 Elo. Taking the formula of the sample standard deviation from the trinomial distribution:

Indeed, coming from Larry, such a strange SD estimate seems awfully wrong...
Maybe he could explain !

[Chessqueen]

I would say that a single standard deviation after 14 games is worth more than ± 40 Elo. Taking the formula of the sample standard deviation from the trinomial distribution:

Indeed, coming from Larry, such a strange SD estimate seems awfully wrong...
Maybe he could explain !

I am also waiting for an explanation since I took statistics I and II accelerated about 35 years ago

[Chessqueen]

I predicted this 4 days ago GM Arjun beating Carlsen, 2 to 3 years from NOW India will have a World Chess Champion, like I stated before most chess players in the World are studying Magnus Carlsen games carefully with top engines and trying to improve with new lines and variations suggested by top engines ==>