Computer cheaters

2 d4
2 Nf3


2... Nc6
2... d6

3 d4
3 Bc4

3....cxd4
3....Nf6


4 Nxd4
4 Qxd4

4....g6
4....Nc6

5 Nc3
5 Bc4

5....g6
5....Bd7

6 Bc4
6 f4

6....e6
6....Nbd7

7 f4
7 Bd3

7....Bd7
7....Be7

8 Bxf6
8 Nb3

Originally posted by no1marauder
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 priori ...[text shortened]... al players is significant. Or am I misunderstanding?

This is why no two games match unless they are thought about in the exact same way.

The computer does see each position as random where humans look for a game plan. In RHP style correspondence chess which might just as well be thought of as 1000's of find the best move problems.

In each one the human will look for his style of play moves, the computer will try to find the best move. There will be variation without a doubt, the more you go through the problems the more the variation will become apparent.

Yet its not 1000 problems because the games will come up again. Next time around the human keeps going with his plan the computer find the best two moves.

and so and so on. Even if two players have the same style of chess (ie attacking) try to get them to match up 90% of even two best moves over games is impossible.

I can see your point, but there is a cut off however where the complexities of chess make it impossible that unless the moves are thought about in the exact same way (ie engines)

If they arent thought about in the exact same way there will be variation and as you said soon to be wild variation.

The only question is how many games do you think that would take for the variation to become apparent.

Originally posted by Grayeyesofsorrow
[b]This is why no two games match unless they are thought about in the exact same way.

The computer does see each position as random where humans look for a game plan.

I received an email regarding this from a person who wishes to remain anonymous. What you say is not quite the way it works. ;-) I don't want to go into some stuff that might benefit the cheater in an open forum, but if we do select a group of people for the purpose of checking games against computer use, (and if I am in that group) I have much information to divulge to them. I can share this with the public though:

There is NO SUCH THING as RANDOM! It doesn't exist! A program has to emulate randomness... ALL PROGRAMS THAT EMULATE RANDOMNESS ARE NOT CAPABLE OF MODIFYING THEIRSELVES. Therefore, any change must be done by the programmer or user. Now then, because we are talking about a series of instructions (called a program) the program MUST repeat itself. That's why the chess programs will always play the same move for any given position UNLESS they are crippled by the user... Guess what? The "crippled program" MUST ALWAYS play the same move for the crippled version!

So, we can do a statistical analysis of any game and deduce (in all probability) if an engine was used or not.

I guess while I'm here I should say something about No1Maurader's question to me regarding if a player can improve his rating as JW did (since he keeps asking the question). My official answer is "I don't know". I suspect though, that (statistically speaking) it would be very unlikely, but not impossible.

arrakis

Originally posted by arrakis
... NO SUCH THING as RANDOM... the program MUST repeat itself.

Hearsay is a dangerous thing. And a terrible band: thank goodness they've disappeared! But back on topic, there are ways to achive behaviour that looks as if it is random. Most programs use the current date/time as a seed into a pseudo-random number generator at the very least. This level of near-randomness is sufficient for the relatively few number of chess games actually played. True randomness can be achieved by a few methods: some hardware based (using newer motherboards) and others using some coding tricks that relate to hardware or www access times. I'm afraid you were misinformed.

Originally posted by Toe
Hearsay is a dangerous thing. And a terrible band: thank goodness they've disappeared! But back on topic, there are ways to achive behaviour that looks as if it is random. Most programs use the current date/time as a seed into a pseudo-random number generator at the very least. This level of near-randomness is sufficient for the relatively few number of chess g ...[text shortened]... ome coding tricks that relate to hardware or www access times. I'm afraid you were misinformed.

Read my message again... you must've missed the point that these programs cannot modify themselves - and if there is a diversion it has to come from the user, AND if there is a diversion IT must always play the same moves!

You apparently assume the program always picks it's calculated optimum move, itself an assumption that there isn't a draw of optimum scores. I'd hope that the programmers aren't that simplistic. But maybe they are: the repeatability of chess programs that Greyeyesofsorrow has on occation mentioned does tend to suggest this simplistic nature. That doesn't mean it has to be that way. I'm not contesting the issue of chess program repeatability, just your statement that programs cannot act randomly. (PS: on the subject of code changes, see neural networks for one: adaptive self-modifying code has been round the block several times and again, if well programmed, these contain random elements that resisit consistantly repeating behaviour)

Originally posted by Toe
You apparently assume the program always picks it's calculated optimum move, itself an assumption that there isn't a draw of optimum scores. I'd hope that the programmers aren't that simplistic. But maybe they are: the repeatability of chess programs that Greyeyesofsorrow has on occation mentioned does tend to suggest this simplistic nature. That doesn't me ...[text shortened]... if well programmed, these contain random elements that resisit consistantly repeating behaviour)

Believe me chess programs are programmed to show the best variations each time every time.

Originally posted by arrakis
I received an email regarding this from a person who wishes to remain anonymous. What you say is not quite the way it works. ;-) I don't want to go into some stuff that might benefit the cheater in an open forum, but if we do select a group of people for the purpose of checking games against computer use, (and if I am in that group) I have much informatio ...[text shortened]... ugh, that (statistically speaking) it would be very unlikely, but not impossible.

arrakis

That's not really an answer; I suppose that nothing is "impossible" - it's "possible" that I'll pitch game 7 in the World Series for the Yankees next year or that I'll sleep with Selma Hayek. The odds of such things happening are as astronomically low as a player of Woodley's limited abilities all of sudden playing at a sustained high level with virtually no losses in hundreds of games.

I wasn't asking for a statistical analysis; I was asking you as an experienced player with a good knowledge of chess whether you thought such a thing was possible without cheating. Perhaps the question would be better phrased this way if you want to use statistics: what is the possibility given the fact that JW, a player of below average skills and knowledge, could suddenly raise his playing level to among the best on the site without outside assistance? I would rate it at 0%; you apparently would rate it at some higher percent. 1%? Or less? To me, the difference between "not possible" and "possible but with an infinitesimal likelihood of occurring" is unimportant in this context unless we are going to require levels of proof which will make the rules against engine use totally unenforceable. And if we're going to do that, we should just allow engine use and be done with it.

Last night I thought I'd do some engine analysis myself to get a feel for those probabilities that I have been spouted off about. The first thing that struck me was that "best move" one book moves are over IS a far more narrow concept than 1 in 4 or 1 in 3, and may well be close to 1 in 2 or even 2 in 3. This of course increases the statistical probability of a human player significantly matching an engine. The little analysis I've done, however, does seem to suggest that, in general, the "better" the player the closer the match. And of course that kind of makes sense.

Anyway here it is. I chose 4 games:

1) 1960 Moscow. Game 7 of the World Championship. Tal - Botvinnik (We know for certain they didn't cheat!)
2) RHP Game 482559. bbarr vs Ironman31. Two of the top players on RHP.
3) RHP Game 567726. Exy vs Gatecrasher. A pair of patzers with delusions of grandeur. Of course, I was keen to see how I matched up to CM10. And games with Exy are always interesting... (BTW this is my only win against him... it was so sweet. :-) )
4) RHP Game 763987. jameswoodley vs ouroboros. Had to see what all the fuss was about.

[Event "7"]
[Site "Moscow"]
[Date "1960. ??. ??"]
[Round "?"]
[White "Tal M"]
[Black "Botvinnik M"]
[Result "1-0"]
[ECO "B18/04"]

{Book Moves - CM10 uses database - no analysis}
1. e4 c6
2. d4 d5
3. Nc3 dxe4
4. Nxe4 Bf5
5. Ng3 Bg6

{Where CM10 differs, CM10's move is given in brackets}

6. N1e2 {Nf3} Nd7 {e5}
7. h4 {Nf4} h6 {f6]
8. Nf4 Bh7
9. Bc4 {Bd3} e5
10. Qe2 Qe7 {Bd6}
11. dxe5 Qxe5
12. Be3 Bc5 {Bb4+}
13. Bxc5 Qxe2+
14. Kxe2 {Bxe2} Nxc5
15. Rhe1 Nf6 {Kf8}
16. b4 {Kf3+} Ncd7
17. Kf1 Kf8
18. Bb3 {Rad1} g5 {a5}
19. hxg5 hxg5
20. Nh3 Rg8
21. Red1 {Rad1} a5
22. bxa5 Rxa5
23. Rd6 {Rd2} Ke7
24. Rad1 Re5 {Rh8}
25. Nh5 {R6d2} Bg6 {Rh8}
26. Rxd7+ Nxd7
27. Rxd7+ Kxd7
28. Nf6+ Kd6
29. Nxg8 Rc5
30. Nh6 {Nh6} f6
31. Ng4 Bxc2
32. Nxf6 Bxb3 {Bf5}
33. axb3 Rb5
34. Nxg5 {Ne4} Rxb3
35. f4 Rb1+ {Ke7}
36. Ke2 {Kf2} Rb2+
37. Kf3 Rb3+
38. Kg4 Rb2
39. g3 b5
40. Nfe4+ Kd5 {Ke7}
41. f5 b4 {Rh2}
42. f6 Ra2
43. f7 Ra8
44. Nh7 b3 {Kd4}
45. Nd2 b2
46. Kf3 Kd4 {c5}
47. Ke2 c5
48. f8Q Rxf8
49. Nxf8 c4
50. Ne6+ Kd5 {Ke5}
51. Nf4+ Kd4
52. Nb1 1-0



Tal/CM10 Matches 35/47 = 74. 5%
Bovinnik/CM10 Matches 31/46 = 67. 4%

[Event "Clan"]
[Site "http://www.redhotpawn.com"]
[Date "2004.04.20"]
[Round "?"]
[White "bbarr"]
[Black "IronMan31"]
[Result "0-1"]
[GameId "482559"]

{Book Moves}
1. e4 e5
2. Ng1f3 Nb8c6
3. Bf1b5 a6
4. Bb5a4 Ng8f6
5. O-O Bf8e7
6. Rf1e1 b5
7. Ba4b3 d6
8. c3 O-O
9. h3 Nc6b8
10. d4 Nb8d7
11. Nb1d2 Bc8b7
12. Bb3c2 Rf8e8
13. Nd2f1 Be7f8
14. Nf1g3 g6
15. a4 c5
16. d5 c4
17. Bc1g5 h6
18. Bg5e3 Nd7c5

{Where CM10 differs, CM10's move is given in brackets}

19. Qd2 {axb5} h5 {Kh7}
20. Be3g5 {Ra3} Be7
21. Bg5h6 {a5} Be7f8 {Nh7}
22. Bxf8 Re8xf8 {Kxf8}
23. Qh6 {axb5} Nf6h7 {Qe7}
24. axb5 axb5
25. Rxa8 Bxa8
26. Ra1 Ba8b7 {Qc7}
27. Ra7 Qc7 {Re8}
28. Nh4 Qd8 {Rd8}
29. Nxg6 Qf6 {fxg6}
30. Qxh7 Kxh7
31. Nxf8 Kg8 {Kh8}
32. Nd7 Nxd7
33. Rxb7 Nd7c5 {Qf4}
34. Rb8+ Kh7
35. Nf5 Qg5
36. Ne3 Qf4
37. Rxb5 Nc5d3 {Nxe4}
38. Bxd3 cxd3
39. Kf1 Qxe4
40. Rb4 Qg6
41. Rb7 Kg8 {h4}
42. Rb8+ Kg7
43. b4 {Rb6} f5
44. Rb7+ Kf8
45. Rb8+ Ke7
46. Rb8b7 {g4} Kd8
47. Rb7a7 {Nc4} f4
48. Ne3d1 {Ra8+} Qe4
49. Ra2 Qxd5
50. Rb2 {Rd2} e4 {f3}
51. b5 d2
52. Kg1 Qd3 {e3}
53. b6 Kc8
54. Kh2 {b7+} e3
55. fxe3 fxe3
0-1

bbarr/CM10 matches 27/37 = 73.0%
Ironman31/CM10 matches 23/37 = 62.2%



[Event "Upsilon Grouped Random"]
[Site "http://www.redhotpawn.com"]
[Date "2004.06.30"]
[Round "2"]
[White "Exy"]
[Black "Gatecrasher"]
[Result "0-1"]
[GameId "567726"]

{Book moves}

1. c4 e5
2. Nb1c3 Ng8f6


{Where CM10 differs, CM10's move is given in brackets}

3. b3 {e3} Nc6
4. Nf3 d5
5. cxd5 Nxd5
6. Bc1b2 {e4} Bg4 {Be7}
7. Ra1c1 {e3} Bb4 {Nxc3}
8. Nc3xd5 {Nxd4} Qxd5
9. Nf3xe5 {a3} Nxe5
10. Bxe5 Qxe5
11. h3 {Rc4 Big blunder} Be6 {O-O-O This move would have won the game much sooner, instead Be6 let him off the hook, well almost}
12. Rc2 O-O-O
13. a3 Bb4xd2 {Rxd2}
14. Rxd2 Rxd2
15. Qxd2 Rh8d8 {Bxb3}
16. Qc1 Bxb3
17. f4 Qa5+
18. Kf2 Rd8d1 {Rd2}
19. Qe3 Qxa3 {Qe1+}
20. g3 Rd1d6 {Qb4}
21. Qc3 {Bg2} a5 {Rd5}
22. Bg2 Qb4
23. Qc3xb4 {axb4} axb4
24. Rb1 Bd5 {Be6}
25. Bxd5 Rxd5
26. Rxb4 c5 {c6}
27. Rb5 Kc8c7 {f5}
28. e4 Rd2+
29. Ke3 Rc2
30. f5 {Kd3} b6 {Kc6}
31. Rb5b3 {e5} Kc6 {f6}
32. Ra3 c4 {Rg2}
33. Ra7 b5 {Rg2}
34. Ra7xf7 {e5} Rc3+
35. Kf4 Rc3d3 {b4}
36. Rxg7 c3 {Rd8}
37. Rxh7 {Rg8} c2
38. Rh6 Kc5
39. Kg4 c1=Q
40. Rg6 Qd1 {Qe3}
41. Kh4 Rd7
42. Rg6h6 {Kg5} Qf3 {b4}
43. g4 Qxe4
44. Kh5 {Rh5} b4 {Qf4}
45. Re6 Qf3 {Qd4}
46. Kg5 {Kg6} b3 {Qxh3}
47. Re5+ Kc5d4 {Rd5}
48. Re1 {Re8} b2
49. Rb1 {Kf6} Qe3+
50. Kg6 Rd7d6 {Qe8+}
51. Kg6f7 {f6} Ke5
52. Rxb2 {Re1} Qa7 {Rf6+}
53. Kf7e8 {Rb7} Qd7+
54. Kf8 Rf6+
55. Kg8 Qf7+
56. Kh8 Rh6++
0-1

Exy/CM10 Matches 34/54 = 63.0%
Gatecrasher/CM10 Match 28/54 = 51.9%



[Event "Clan"]
[Site "http://www.redhotpawn.com"]
[Date "2004.11.04"]
[Round "?"]
[White "jameswoodley"]
[Black "ouroboros"]
[TimeControl "-"]
[Result "1-0"]
[ECO "oft "]

{Book Moves}
1.Nf3 Nf6
2.Nc3 d5
3.d4 Bf5

4.h3 e6
5.g4 Bg6
6.Bg2 Bb4 {c5}
7.a3 {Bd2} Ba5 {Bxc3+}
8.b4 Bb6
9.Ne5 Nbd7 {Nc6}
10.Nxg6 hxg6
11.g5 Ng8 {Rh4}
12.e4 c6
13.exd5 {Qd3} cxd5
14.Nb5 Qe7 {Ne7}
15.Bf4 Kd8 {Rh4}
16.c4 a6
17.Nc3 {Nd6} Bc7 {Rc8}
18.Be3 Nb6 {a5}
19.cxd5 Nxd5 {Nc4}
20.Nxd5 exd5
21.Qb3 Qd7 {Ke8}
22.Bxd5 Rb8 {Ne7}
23.Bxf7 Ne7
24.Be6 Qd6
25.Rc1 {O-O-O} Nf5 {Bb6}
26.Qc3 {Bxf5} Nxe3 {Rh4}
27.fxe3 Re8 {Bb6}
28.Bg4 {d5} Qg3+
29.Ke2 {Kd2} Qg2+ {Bf4} {after the last 5 moves, white's good advantage has all but evaporarated, but the next 10 moves or so miraculously restore it :-) }
30.Kd3 Qe4+
31.Kc4 b5+ {Qxe3}
32.Kb3 Qxe3
33.Ka2 Qxc3
34.Rxc3 Bf4 {Re4}
35.Rc6 Ra8 {Bxg5}
36.Rxg6 Re7 {Ra7}
37.Rf1 Bd6 {Bh2}
38.Rf8+ Re8
39.Rxd6+ Ke7
40.Rxe8+ Kxe8
41.Bh5+ {Bf3} Ke7 {Kf8}
42.Rg6 Kf8
43.Re6 g6 {a5}
44.Bxg6 Kg7 {a5}
45.h4 a5 {Kg8}
46.bxa5 Rxa5
47.h5 1-0

jameswoodley/CM10 matches 36/44 = 81.8%
ourboros/CM10 matches 19/43 = 44.2%

I haven't used very deep analysis settings. I will do that when I have more time available. I'm not sure the numbers would change too dramatically, though.

Regardless of anything else, this is presented to get a feel for what kind of match ups one would normally expect, and at various standards of play. I found it illuminating. I said on another thread I would be shocked if had a more than 50% match on any of my games. Well, I did on the very first one I picked.

I guess it means that to prove engine use the bar has to be set pretty high. And it cannot be just an isolated instance.

I hope this will add some value to the debate.

Originally posted by Gatecrasher
Last night I thought I'd do some engine analysis myself to get a feel for those probabilities that I have been spouted off about. The first thing that struck me was that "best move" one book moves are over IS a far more narrow concept than 1 in 4 or 1 in 3, and may well be close to 1 in 2 or even 2 in 3. This of course increases the statistical prob ...[text shortened]... . And it cannot be just an isolated instance.

I hope this will add some value to the debate.

Very intresting :-)

JW at almost 80% of a 2600 engine, surprising......well not really.

No1 has opened my eyes to the idea of numbers being hard to prove as well. You can get a feel for the engine games. They match up moves that normally wouldnt match up and the game gets that feel to it. Clearly the mods have a hard hard time in front of them. Its good to see a Tal game there :-)

Id say the moves that dont match define the players style and the moves that do match are the logical conclusion of the unmatching moves. I still think 90% is a very dodgy number :-)

Originally posted by Grayeyesofsorrow
Very intresting :-)

JW at almost 80% of a 2600 engine, surprising......well not really.

No1 has opened my eyes to the idea of numbers being hard to prove as well. You can get a feel for the engine games. They match up moves that normally wouldnt match up and the game gets that feel to it. Clearly the mods have a hard hard time in front of ...[text shortened]... re the logical conclusion of the unmatching moves. I still think 90% is a very dodgy number :-)

I don't disagree that 90% of moves matching in game after game is very strong evidence - I was merely trying to point out that some posts seemed to be suggesting that anytime anybody's game matched up at 90% thuis was conclusive proof of engine use (I'm not saying you said this but others did). I'm also glad to see that Gatecrasher was willing to run his own tests and see that the percentage matchups can be quite high for the reasons I suggested. But again run any of Woodley's games in September through CM and I'd wager significant amounts of cash you'll see a very low % of matches with CM and then all of a sudden you'll see his games matching 80-90% in virtually every game. That along with what we know about his chess abilities leads to an open and shut case. I'm sure other cases will be far less clear and I'm saying we shouldn't even start an investigation into a player unless we have some kind of external evidence that suggests engine use. I think that the Chess Mods will have less work than people think, but I also think their very existence will go a long way to deter players from using engines.

Originally posted by no1marauder
I don't disagree that 90% of moves matching in game after game is very strong evidence - I was merely trying to point out that some posts seemed to be suggesting that anytime anybody's game matched up at 90% thuis was conclusive proof of engine use (I'm not saying you said this but others did). I'm also glad to see that Gatecrasher was wil ...[text shortened]... k, but I also think their very existence will go a long way to deter players from using engines.

Im not sure if JW is so stupid that he cant see how clear cut the case is against him. At least JW will be one of the very last cases of this type of cheating on the site. Iv said many times and i'll say it again, that the mods should try to prove that people dont cheat, instead of trying to prove they do. Frankly its easier to do for a start and it stops them looking for evidence to support the claim instead of seeing the game as it is. I wouldnt care less if Exy was hitting 50-80% in game after game, it would take less time and it takes to get a coffee to prove he doesnt cheat.

Originally posted by Grayeyesofsorrow
Very intresting :-)

JW at almost 80% of a 2600 engine, surprising......well not really.

No1 has opened my eyes to the idea of numbers being hard to prove as well. You can get a feel for the engine games. They match up moves that normally wouldnt match up and the game gets that feel to it. Clearly the mods have a hard hard time in front of ...[text shortened]... re the logical conclusion of the unmatching moves. I still think 90% is a very dodgy number :-)

This is a good point. My OTB games match up (for moves from the end of the opening to the middle game period at which time trouble hits) at a fairly high percentage to Fritz et al. However I find there tends to be - in my games at least - a decision point every 5-6 moves at which you basically plan, and the next 4-5 moves flow fairly obviously from this.

There's a ood example in that Conticchio (sp?) game that was analysed as matching to Shredder in one of these threads - the analysis started with the moves Bg1, h3, f5 all matching an engine - however the move before Bg1 was Kh1, which my Fritz hates (rating as even, whilst any other move was +0.3). However once it has been played the next moves all flow as being the best way to play the plan White has committed to. I make no comments on the further moves as I was too tired to go through the entire game.

Interestingly or not, in my own games it's endgame positions where I vary wildly from engine analysis - there tend to be far more candidate moves around which squares to put the pieces on, as there are fewer tactics - I'd be more suspicious of a high matchup there than in the moddlegame. Whatever, the people looking at these games have to be pretty useful chess players and analysts.

Comments above relate to one matching game btw - a series of matches becomes much less likely, as has been mentioned many times already. And people like JW are just off the scale in sheer stupidity 🙂

Dear chessfriends!!
If you want discouver an engine use there's a important point!
You must find the ! moves or !! moves.
If a player see all the !! moves in 100 games it's clearly a CHEATER!!!
But second problem!!
How decided that a move il very good and difficult for a man?
Only a GM or a strong International Master can decide, so we aren't able to understand if someone use engines.
Exemple!!
I hope that you understand my Enghish because this is the point.
When I play at the board a long time game and my opponent in a strong FIDE player I lose my game ( no all , some win or draw ) for no more than 3 or 4 inferior move ( no blunder ), but this 3 or 4 move are the difference of power between a good amateur and a professionist.
Only 3 or 4 move no more !! Understand !!
So your percent/analisy mean nothink , are like fresh-air ! Sorry for you but you must study and understand chess for a lot of year and pheraps you can see the difference between a man and an engine.
Sorry for you but this the true !!
I too have problem if I decide to enter in RHP policy.
Bye Bye

Originally posted by El Bruto
Dear chessfriends!!
If you want discouver an engine use there's a important point!
You must find the ! moves or !! moves.
If a player see all the !! moves in 100 games it's clearly a CHEATER!!!
But second problem!!
How decided that a move il very good and difficult for a man?
Only a GM or a strong International Master can decide, so we aren't able ...[text shortened]... for you but this the true !!
I too have problem if I decide to enter in RHP policy.
Bye Bye

The problem here is that the person "jameswoodley" would draw with a chess engine every single game he played it; never winning or losing. If he sat down with the program it would be like he was playing a game with himself. You could almost say he "is" a chess engine.