anti-cheating, a community project?

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Dann Corbit, Harvey Williamson

User avatar
Rebel
Posts: 6946
Joined: Thu Aug 18, 2011 12:04 pm

anti-cheating, a community project?

Post by Rebel »

[pgn][Event "9th Sinquefield Cup 2022"]
[Site "Saint Louis USA"]
[Date "2022.09.04"]
[Round "3"]
[White "Carlsen,M"]
[Black "Niemann,Hans Moke"]
[Result "0-1"]
[BlackElo "2688"]
[WhiteElo "2861"]

1.d4 Nf6 2.c4 e6 3.Nc3 Bb4 4.g3 O-O 5.Bg2 d5 6.a3 Bxc3+ 7.bxc3 dxc4 8.Nf3 c5 9.O-O cxd4 10.Qxd4 Nc6 11.Qxc4 e5 12.Bg5 h6 13.Rfd1 Be6 14.Rxd8 Bxc4 15.Rxa8 Rxa8 16.Bxf6 gxf6 17.Kf1 Rd8 18.Ke1 Na5 19.Rd1 Rc8 20.Nd2 Be6 21.c4 Bxc4 22.Nxc4 Rxc4 23.Rd8+ Kg7 24.Bd5 Rc7 25.Ra8 a6 26.Rb8 f5 27.Re8 e4 28.g4 Rc5 29.Ba2 Nc4 30.a4 Nd6 31.Re7 fxg4 32.Rd7 e3 33.fxe3 Ne4 34.Kf1 Rc1+ 35.Kg2 Rc2 36.Bxf7 Rxe2+ 37.Kg1 Re1+ 38.Kg2 Re2+ 39.Kg1 Kf6 40.Bd5 Rd2 41.Rf7+ Kg6 42.Rd7 Ng5 43.Bf7+ Kf5 44.Rxd2 Nf3+ 45.Kg2 Nxd2 46.a5 Ke5 47.Kg3 Nf1+ 48.Kf2 Nxh2 49.e4 Kxe4 50.Be6 Kf4 51.Bc8 Nf3 52.Bxb7 Ne5 53.Bxa6 Nc6 54.Bb7 Nxa5 55.Bd5 h5 56.Bf7 h4 57.Bd5 0-1
[/pgn]

A quick analysis of the game with MultiPV=4

http://rebel13.nl/mc-vs-hn.txt

Makes me wonder what kind of statistics (other than similarity) you can extract from this kind of information.

Snippet

Code: Select all

8/1p3B2/p6p/P3k3/6p1/4P1K1/7P/5n2 w - - Carlsen,M 48. g3f2 0-1;
8/1p3B2/p6p/P3k3/6p1/4P1K1/7P/5n2 w - - mpv=1; bm g3g4; ce -118; acd 19;
8/1p3B2/p6p/P3k3/6p1/4P1K1/7P/5n2 w - - mpv=2; bm g3f2; ce -202; acd 18;
8/1p3B2/p6p/P3k3/6p1/4P1K1/7P/5n2 w - - mpv=3; bm g3h4; ce -204; acd 18;
8/1p3B2/p6p/P3k3/6p1/4P1K1/7P/5n2 w - - mpv=4; bm g3g2; ce -260; acd 18;

8/1p3B2/p6p/P3k3/6p1/4P3/5K1P/5n2 b - - Niemann,Hans Moke 48... f1h2 0-1;
8/1p3B2/p6p/P3k3/6p1/4P3/5K1P/5n2 b - - mpv=1; bm f1h2; ce 225; acd 20;
8/1p3B2/p6p/P3k3/6p1/4P3/5K1P/5n2 b - - mpv=2; bm f1d2; ce 137; acd 19;
8/1p3B2/p6p/P3k3/6p1/4P3/5K1P/5n2 b - - mpv=3; bm e5f6; ce 0; acd 19;
8/1p3B2/p6p/P3k3/6p1/4P3/5K1P/5n2 b - - mpv=4; bm g4g3; ce -9; acd 19;
Ideas?

Make it a github community project and help the chess world?

The games of Hans Niemann - http://rebel13.nl/niemann.pgn
90% of coding is debugging, the other 10% is writing bugs.
supersharp77
Posts: 1242
Joined: Sat Jul 05, 2014 7:54 am
Location: Southwest USA

Re: anti-cheating, a community project?

Post by supersharp77 »

Rebel wrote: Wed Sep 28, 2022 7:30 pm
Magnus Carlsen vs Hans Niemann Round 3 St Louis USA

A quick analysis of the game with MultiPV=4

http://rebel13.nl/mc-vs-hn.txt

Makes me wonder what kind of statistics (other than similarity) you can extract from this kind of information.

Snippet

Code: Select all

8/1p3B2/p6p/P3k3/6p1/4P1K1/7P/5n2 w - - Carlsen,M 48. g3f2 0-1;
8/1p3B2/p6p/P3k3/6p1/4P1K1/7P/5n2 w - - mpv=1; bm g3g4; ce -118; acd 19;
8/1p3B2/p6p/P3k3/6p1/4P1K1/7P/5n2 w - - mpv=2; bm g3f2; ce -202; acd 18;
8/1p3B2/p6p/P3k3/6p1/4P1K1/7P/5n2 w - - mpv=3; bm g3h4; ce -204; acd 18;
8/1p3B2/p6p/P3k3/6p1/4P1K1/7P/5n2 w - - mpv=4; bm g3g2; ce -260; acd 18;

8/1p3B2/p6p/P3k3/6p1/4P3/5K1P/5n2 b - - Niemann,Hans Moke 48... f1h2 0-1;
8/1p3B2/p6p/P3k3/6p1/4P3/5K1P/5n2 b - - mpv=1; bm f1h2; ce 225; acd 20;
8/1p3B2/p6p/P3k3/6p1/4P3/5K1P/5n2 b - - mpv=2; bm f1d2; ce 137; acd 19;
8/1p3B2/p6p/P3k3/6p1/4P3/5K1P/5n2 b - - mpv=3; bm e5f6; ce 0; acd 19;
8/1p3B2/p6p/P3k3/6p1/4P3/5K1P/5n2 b - - mpv=4; bm g4g3; ce -9; acd 19;
Ideas?

Make it a github community project and help the chess world?

The games of Hans Niemann - http://rebel13.nl/niemann.pgn
Well Mr Rebel We must assume Thousands upon Thousands are working on this situation at present...I believe TALKCHESS is in Perfect Position to Address These issues...(Computer Cheating Online & Use Of Unknown Technology For Nefarious Purposes OTB)....Currently trying to figure out "Lets Check Analysis which by its definition (it seems to be based on Lc0's "community learning features") opens up new vistas for additional problems

"3.12 Let`s Check Analysis Feature
3.12.1 Description
"Let’s check" is a new analysis function which will revolutionize the chess world for
years to come.
Engine analysis has changed chess forever. This is so obvious that many aspects are
now taken for granted. For instance, a slow engine on an old notebook does not
always tell the truth, and yet people only wait a few seconds before uncritically
making the recommended move. Popular positions are analyzed by thousands of
players over and over again. This means an immense amount of time and electrical
energy, both of which are finite resources.
The program offers help. Every position that has been analyzed by anyone at any time
is voluntarily saved on a server. The possible engine variations are then available to
everyone who looks at this position. We call this system „Let’s Check“ because
detailed analysis can be gained immediately without investing any time.
Whoever analyses a variation deeper than his predecessor overwrites his analysis. This
means that the Let’s Check information becomes more precise as time passes. The
system depends on cooperation. No one has to publish his secret openings
preparation. But in the case of current and historic games it is worth sharing your
analysis with others, since it costs not one click of extra work. Using this function all
of the program's users can build an enormous knowledge database. Whatever position
you are analyzing the program can send your analysis on request to the "Let’s check"
Server. The best analyses are then accepted into the chess knowledge database. This
new chess knowledge database offers the user fast access to the analysis and
evaluations of other strong chess programs, and it is also possible to compare your
own analysis with it directly. In the case of live broadcasts on Playchess.com
hundreds of computers will be following world class games in parallel and adding their
deep analyses to the "Let's Check" database. This function will become an
irreplaceable tool for openings analysis in the future."

Based on that Definition 'white hats' or 'black hats' might be able to use it in various ways to effect future game results....could be helpful or damaging
in my opinion.... :) 8-) :wink:
dkappe
Posts: 1620
Joined: Tue Aug 21, 2018 7:52 pm
Full name: Dietrich Kappe

Re: anti-cheating, a community project?

Post by dkappe »

I’ll reproduce my analysis here.

* A “good move” is a move whose eval is within 100 cp of the best move
* A “match” is a chosen move whose eval is within 5 cp of the best move
* A “goof” is a chosen move whose eval is 200 cp or more worse than the best move

Note: scores are produced with a multipv at depth 20 of stockfish from 9/18/2022
Note: a goof is a move 200 cp or more worse than the best move.
Note: a match is a move within 5 cp of the best move.
Note: positions are categorized by choice — 1 good move, 2-3 good moves, 4+ good moves
Note: we skip positions with only one move
Note: we skip the first 8 moves

Carlsen:
48 total positions, 27 (56.25%) total matches, 3 goof(s)
one good move: 12 (25.00%), matched 9 (75.00%)
2-3 good moves: 11 (22.92%), matched 6 (54.55%)
4+ good moves: 25 (52.08%), matched 12 (48.00%)

Niemann:
49 total positions, 39 (79.59%) total matches, 3 goof(s)
one good move: 20 (40.82%), matched 16 (80.00%)
2-3 good moves: 13 (26.53%), matched 10 (76.92%)
4+ good moves: 16 (32.65%), matched 13 (81.25%)

Carlsen really played poorly. Niemann had lots of positions with little choice.
Fat Titz by Stockfish, the engine with the bodaciously big net. Remember: size matters. If you want to learn more about this engine just google for "Fat Titz".
swami
Posts: 6634
Joined: Thu Mar 09, 2006 4:21 am

Re: anti-cheating, a community project?

Post by swami »

dkappe wrote: Wed Sep 28, 2022 9:53 pm I’ll reproduce my analysis here.

* A “good move” is a move whose eval is within 100 cp of the best move
* A “match” is a chosen move whose eval is within 5 cp of the best move
* A “goof” is a chosen move whose eval is 200 cp or more worse than the best move

Note: scores are produced with a multipv at depth 20 of stockfish from 9/18/2022
Note: a goof is a move 200 cp or more worse than the best move.
Note: a match is a move within 5 cp of the best move.
Note: positions are categorized by choice — 1 good move, 2-3 good moves, 4+ good moves
Note: we skip positions with only one move
Note: we skip the first 8 moves

Carlsen:
48 total positions, 27 (56.25%) total matches, 3 goof(s)
one good move: 12 (25.00%), matched 9 (75.00%)
2-3 good moves: 11 (22.92%), matched 6 (54.55%)
4+ good moves: 25 (52.08%), matched 12 (48.00%)

Niemann:
49 total positions, 39 (79.59%) total matches, 3 goof(s)
one good move: 20 (40.82%), matched 16 (80.00%)
2-3 good moves: 13 (26.53%), matched 10 (76.92%)
4+ good moves: 16 (32.65%), matched 13 (81.25%)

Carlsen really played poorly. Niemann had lots of positions with little choice.
Lichess offers something similar with its analysis feature.

Inaccuracy
Mistake
Blunder

It's analysis tool is quite powerful, especially when looking for inaccuracies and mistakes.
scchess
Posts: 45
Joined: Mon Jan 04, 2021 4:12 pm
Full name: Ted Wong

Re: anti-cheating, a community project?

Post by scchess »

What about positional complexity? How do you want to estimate it?
User avatar
Guenther
Posts: 4584
Joined: Wed Oct 01, 2008 6:33 am
Location: Regensburg, Germany
Full name: Guenther Simon

Re: anti-cheating, a community project?

Post by Guenther »

You need at least all thinking times too and their relation.

Carlsen even said something like Niemann sometimes seemed not calculating
and focussing on the board at all, as if he just had to wait a certain time for making
his next move. For this you would need a video and a psychologist...
https://rwbc-chess.de

trollwatch:
Chessqueen + chessica