Computer cheaters

2. Nf3


1 2...d6 18,716
2 2...Nc6 13,747
3 2...e6 11,942
4 2...g6 818
5 2...a6 443
6 2...Nf6 373
7 2...b6 42
8 2...Qc7 16
9 2...Qa5 7
10 2...d5 3
11 2...Qb6 2
12 2...h6 Z Hajnal vs M Appleberry, 2001 1-0

2.....d6

move games win-draw-loss %

1 3. d4 15,475
2 3. Bb5+ 1,780
3 3. c3 635
4 3. Nc3 498
5 3. Bc4 193
6 3. d3 63
7 3. c4 30
8 3. b4 21
9 3. b3 13
10 3. Be2 12
11 3. h3 5
12 3. Bd3 5
13 3. a3 A H Wohl vs R S Jones, 2002 1-0
14 3. e5 B Eikelboom vs M Van Delft, 2000 0-1

3.d4

1 3...cxd4 14,766
2 3...Nf6 721
3 3...Nd7 3
4 3...b6 P Michel vs J Traian-Iliesco, 1943 1-0

3.....cxd4


1 4. Nxd4 14,294
2 4. Qxd4 467
3 4. c3 10
4 4. Bb5+ 3
5 4. Bc4 R Licardo vs B Kutuzovic, 1999 0-1

4. Nxd4


1 4...Nf6 14,167
2 4...Nc6 124
3 4...e6 44
4 4...a6 35
5 4...e5 13
6 4...g6 12
7 4...Bd7 2

Grrrrrr the database picks bloody good times to go down!! :-)

You get the idea however, In 5 moves iv change a database of 300,000 games to under 15, 000 following that move order.

(+ added 2000 edits to my name :-) )

Originally posted by Ragnorak
Grayeyes,

can u post the control games of Exy and bbarr? I've had a look for them, but couldn't find them.

They're a good indication on the matchup percentage of a NON-CHEATER.

D

I wish I could find the thread that they are in. I still have them on my pc somewhere if they cant be turned up :-)

Originally posted by Grayeyesofsorrow
I'll try to make this as clear as I can, not something im very good at sorry.

The chances of two 40 move chess games ever been the same is near impossible. There are databases online with 10 million+ games, out of all them no ...[text shortened]... ou have eliminated every possiblity apart from engine use.

I'll weigh in here. First of all, I know very little about computer engines: I had a CM 4000 10 years ago and there's a Fritz4 in my downloaded ChessbaseLight that I use to look over my games (after they're finished, people) - it sees tactics but isn't very good positionally and stinks in the endgames (have they improved that in the more recent models?). I do know a fair amount of chess theory and have played off and on for 40 years. When a good player looks at a middlegame position he might consider a fairly large amount of moves ( I was taught to look for piece sacs every move because if you don't look for them you'll never see them even though 99% of the time there aren't any sound ones on the board), but he's going to discard all but a handful or less in a matter of minutes. That leaves him with only a few moves that a good player would make and if you're going to say he matches with any of the top two of another strong player (in this case an engine), I would think that the percentage of agreement is often going to be rather high. 90% in a game? Sometimes; when I go over my games with other strong players very often we agree on 90% on the moves. Remember also that this isn't OTB with a set time limit; you have literally days if you want them to study a position and find the correct move and you can set up a side board and move the pieces around to yer heart's content (something that upsets OTB opponents).

What I'm saying is that I think engine analysis is evidence to be considered, but I don't necesssarily agree that just because a game of a player matches up with an engine on 90% of the moves in a game, the player cheated. You'd have to see a consistent pattern in almost all their games to make the judgment on engine analysis evidence alone. I think that what we need to do is investigate people only if there is some objective criteria met that raising suspicion of cheating: a sudden sharp jump in playing skills and winning record, a new player winning every game - rapidly making moves, a super strong player making moves at set intervals, etc. All these things have taken place here at RHP.

We'll never be able to prove beyong a "reasonable doubt" that someone used an engine without a confession, but if we adopt a standard of "clear and convincing" evidence, give the players a chance to explain the circumstances that raised the suspicion and if the people doing the evaluating are both strong experienced players and knowledable in chess engines, then I think there won't be a major problem with false accusations. I don't want the situation to exist where every time some people lose a game they start screaming "Cheater!" and an investigation must commence; that will quickly destroy the good will that primarily exists on the site. I would limit the investigations to the type of people outlined above, realize that we can never totally "solve" the problem of engine use on an internet chess site and go on believing that when lose my games I lost to a better player, not an engine. If we do these things we'll lessen the problem, get rid of blatant and obvious cheaters but still keep this the best place to play chess on the web. And that's what we all want.

From move 1. 300,000 games are whittled down to 511 in 7 moves. :-)

The average number of candidate moves comes out at 9.6 per a move.


1...c5

1 2. Nf3 46,044
2 2. Nc3 4,869
3 2. c3 3,479
4 2. d4 656
5 2. f4 545
6 2. Ne2 353
7 2. d3 317
8 2. g3 231
9 2. b3 228
10 2. c4 176
11 2. b4 109
12 2. Bc4 70
13 2. a3 20
14 2. Qe2 3
15 2. Bb5 T Christiansen vs I Sajid, 2004 0-1
16 2. Qh5 E Treger vs S Agaian, 2003 0-1
17 2. Nh3 Brian Wall vs Popovych, 2001 0-1
18 2. e5 A Mongiello vs E Yeo Min Yang, 2004 0-1
19 2. Be2 Larsen vs L Schandorff, 1997 1/2-1/2

1...c5

1 2. Nf3 46,044
2 2. Nc3 4,869
3 2. c3 3,479
4 2. d4 656
5 2. f4 545
6 2. Ne2 353
7 2. d3 317
8 2. g3 231
9 2. b3 228
10 2. c4 176
11 2. b4 109
12 2. Bc4 70
13 2. a3 20
14 2. Qe2 3
15 2. Bb5 T Christiansen vs I Sajid, 2004 0-1
16 2. Qh5 E Treger vs S Agaian, 2003 0-1
17 2. Nh3 Brian Wall vs Popovych, 2001 0-1
18 2. e5 A Mongiello vs E Yeo Min Yang, 2004 0-1
19 2. Be2 Larsen vs L Schandorff, 1997 1/2-1/2


2. Nf3


1 2...d6 18,716
2 2...Nc6 13,747
3 2...e6 11,942
4 2...g6 818
5 2...a6 443
6 2...Nf6 373
7 2...b6 42
8 2...Qc7 16
9 2...Qa5 7
10 2...d5 3
11 2...Qb6 2
12 2...h6 Z Hajnal vs M Appleberry, 2001 1-0



2.....d6

1 3. d4 15,475
2 3. Bb5+ 1,780
3 3. c3 635
4 3. Nc3 498
5 3. Bc4 193
6 3. d3 63
7 3. c4 30
8 3. b4 21
9 3. b3 13
10 3. Be2 12
11 3. h3 5
12 3. Bd3 5
13 3. a3 A H Wohl vs R S Jones, 2002 1-0
14 3. e5 B Eikelboom vs M Van Delft, 2000 0-1

3.d4

1 3...cxd4 14,766
2 3...Nf6 721
3 3...Nd7 3
4 3...b6 P Michel vs J Traian-Iliesco, 1943 1-0



3.....cxd4


1 4. Nxd4 14,294
2 4. Qxd4 467
3 4. c3 10
4 4. Bb5+ 3
5 4. Bc4 R Licardo vs B Kutuzovic, 1999 0-1


4. Nxd4


1 4...Nf6 14,167
2 4...Nc6 124
3 4...e6 44
4 4...a6 35
5 4...e5 13
6 4...g6 12
7 4...Bd7 2


4....Nf6

1 5. Nc3 14,778
2 5. f3 213
3 5. Bd3 19
4 5. Bb5+ 9
5 5. Bc4 3
6 5. c3 2
7 5. Bg5 2
8 5. Na3 C Sandipan vs D Tan, 2002 1-0


5. Nc3


1 5...a6 8,354
2 5...Nc6 6,165
3 5...e6 2,858
4 5...g6 2,302
5 5...Bd7 91
6 5...Nbd7 57
7 5...e5 28
8 5...h6 3
9 5...a5 2
10 5...Na6 G Mohr vs Mikhalchishin, 2001 1/2-1/2
11 5...Qb6 Kotronias vs Sasikiran, 2004 1-0


5.....a6

1 6. Bg5 1,958
2 6. Be2 1,815
3 6. Be3 1,658
4 6. Bc4 937
5 6. f4 840
6 6. f3 387
7 6. g3 348
8 6. a4 256
9 6. h3 100
10 6. Bd3 48
11 6. Rg1 24
12 6. Qf3 11
13 6. Nb3 5
14 6. a3 3
15 6. b3 Krum Georgiev vs Pinter, 1985 1/2-1/2


6. Bg5

6...e6 1,757
6...Nbd7 140
6...Nc6 93
6...e5 5
6...Bd7 3
6...Qb6 2
6...g6 2
6...b5 A Garrido Borja vs L Keely, 2001 1-



6....e6

1 7. f4 1,546
2 7. Qf3 79
3 7. Qd2 38
4 7. Qd3 33
5 7. Be2 24
6 7. Qe2 17
7 7. Bc4 8
8 7. f3 6
9 7. a4 5
10 7. g3 5
11 7. Bd3 4


7. f4

1 7...Qb6 511
2 7...Be7 478
3 7...Nbd7 182
4 7...Qc7 128
5 7...b5 124
6 7...h6 66
7 7...Nc6 47
8 7...Bd7 19

7...Qb6

1 8. Qd2 321
2 8. Nb3 148
3 8. a3 18
4 8. Qd3 17
5 8. Bxf6 6
6 8. f5 Murey vs Pinter, 1980 0-1

Originally posted by no1marauder
I'll weigh in here. First of all, I know very little about computer engines: I had a CM 4000 10 years ago and there's a Fritz4 in my downloaded ChessbaseLight that I use to look over my games (after they're finished, people) - it sees tactics but isn't very good positionally and stinks in the endgames (have they improved that in the more recen ...[text shortened]... aters but still keep this the best place to play chess on the web. And that's what we all want.

I think you might agree on 90% of moves, but I doubt both of you would play 90% of the same moves game after game.

Iv tried to highlight this with the Database opening post. A lot of the games are from FM,IM,GM and past chess gods. They might all agree a move is good, but they all diverse after only 8 moves.

90% in a few games means nothing, 90 in dodgy games, game after game is a strong suggestion which needs to be checked.

Originally posted by Grayeyesofsorrow
I think you might agree on 90% of moves, but I doubt both of you would play 90% of the same moves game after game.

Iv tried to highlight this with the Database opening post. A lot of the games are from FM,IM,GM and past chess gods. They might all agree a move is good, but they all diverse after only 8 moves.

90% in a few games means nothing, 90 in dodgy games, game after game is a strong suggestion which needs to be checked.

But you're also saying that either of the top two engines moves is a match; thus more than doubling the possibility of a match. I say more than doubling because the second choice move might be markedly inferior but if the player makes it he's considered to "match" the engine. Isn't that correct? When I look at a database and decide to make a move in the database, it isn't "matching" the database to play either 1 d4 or e4 or you'd be "matching" 80-90% of the time. So using either of the top two moves, plus recaptures plus forced moves considerably raises the percentage of moves that match, doesn't it?

Another thing is that one different move from a player results in a different game therefore it won't often match other player's GAMES, but you're looking at each move individually like the engine does. You know that there are many positions, perhaps most, if you show to two strong players and ask them to give the two best possible moves they are going to agree. So I think you are somewhat overstating the evidentiary value of engine analysis has you are doing it with either of the top two engine moves = a match.

Originally posted by no1marauder
But you're also saying that either of the top two engines moves is a match; thus more than doubling the possibility of a match. I say more than doubling because the second choice move might be markedly inferior but if the player makes it he's considered to "match" the engine. Isn't that correct? When I look at a database and decide to make a ...[text shortened]... value of engine analysis has you are doing it with either of the top two engine moves = a match.

Good points yes its the top two moves. Which makes the 8 moves match up closer to being a reality but not that close. Over 40 moves the variations are mind boggling.

Its best to try and not think of it as a computer picking the same moves as them. Because it is infact, the checking computer playing its own game of chess with no aid from anything else. Its a totally new game. If the protential cheaters game matchs ones played by engine game after game then yes they are cheating.


Basicly if you and me were to be given a chess position and told to find the best move, we might agree. Then someone makes the same move against us, and we are asked to pick the best move again, again we might agree but the chances are starting to decrease.

Say that started on move 8 and both of our games lasted 40 moves, our positions would be totally different by the end of it, even if we listed the two top moves.

In every game I play I try to pick the best move, yet all the games I play are different. Surely I think the same way in each game and play with the same style/rating. However chess is so complex that none of my games ever match up. If I was to take a game I played a week ago (providing I had forgot the moves played) and I played it again now, my moves still wouldnt match the ones I play a week ago. The engine moves stay the same game after game which is the difference :-)

Originally posted by Grayeyesofsorrow
In every game I play I try to pick the best move, yet all the games I play are different. Surely I think the same way in each game and play with the same style/rating. However chess is so complex that none of my games ever match ...[text shortened]... moves stay the same game after game which is the difference :-)

We can do a simple test to see how much the two moves effects the match up. From the list of opening variations above, pick the two that you would play in each position and i'll pick the two id play. We can see the match up between us after 7 moves. That might provide an indication. The test is rigged so that its more likley our moves will match up because we arent playing down the variations and the moves are already listed for us. But it will be intresting anyway.

Its 1 e4 c5 :-)

Originally posted by Grayeyesofsorrow
Good points yes its the top two moves. Which makes the 8 moves match up closer to being a reality but not that close. Over 40 moves the variations are mind boggling.

Its best to try and not think of it as a computer picking the same moves as them. Because it is infact, the checking computer playing its own game of chess with no aid from anyt ...[text shortened]... ur positions would be totally different by the end of it, even if we listed the two top moves.

Well you're making a statistical argument about how humans would look at a game; we might evolve a totally different plan - you might want to sac a pawn for an open line, I might want to occupy a weak square as my first priority. So the way we play our games is going to vary. But once we vary even on one move our games won't "match" and in all likilihood once then start diverging then will diverge sharply.

But, correct me if I'm wrong, an engine is just making what it calculates as the strongest move at that time - it is taking into account future variations but it is doing just that move. Then when we ask it to make the next move, it recalculates in toto again, it doesn't "plan" like a human. In fact often when I'm analyzing a completed game, Fritz4 will suggest as a move a move restoring a position back to where one player varied so it can make the move it thought was appropriate in that position!

I guess I feel you're somewhat comparing apples and oranges; after all if there's a 90% match on the first two moves the checking computer HASN'T played the same game as the player; they are entirely different just like a difference of one or two moves in a player's game varies it from another player's. So you seem to be saying that one or two moves differences in a checking computer from a player is insignificant, but one or two moves different from two real players is significant. Or am I misunderstanding?