Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: How people could detect if a game was cooked?

Author: Chessfun

Date: 07:50:30 09/15/02

Go up one level in this thread


On September 15, 2002 at 10:24:26, Rolf Tueschen wrote:

>On September 15, 2002 at 08:55:57, Chessfun wrote:
>
>>On September 15, 2002 at 07:59:42, Rolf Tueschen wrote:
>>
>>>From the Macheide XP tuning games of Thorsten I found the following game. NB
>>>that I did _not_ select the game. It was the first I played. Now I will take a
>>>closer look.
>>>
>>>The game [read the article in Computerschachwelt]:
>>>
>>>Shredder 6 - XP Machëide x2cou_51
>>>40 Moves in 120 min; 20 Moves in 60 min auto232 match, 400mhz (3), 13.09.2002
>>>
>>>1.d4 book 0s 1...Nf6 1:27m 2.c4 book 0s 2...e6 5s 3.Nc3 book 0s 3...Bb4 2s 4.e3
>>>book 0s 4...0-0 3s 5.Bd3 book 0s 5...d5 3s 6.cxd5 book 0s 6...exd5 2s 7.Nge2
>>>book 0s 7...Re8 2s 8.0-0 book 0s 8...Bd6 3s 9.f3 book 0s 9...c5 2s 10.Qe1 book
>>>0s 10...Nc6 3s 11.Qh4 book 0s 11...Be7 4s 12.dxc5 -0.15/13 10:06m 12...Bxc5
>>>5:30m 13.Nd4 -0.15/13 0s 13...Bd7 3:37m (Ne5) 14.Nxd5 -0.03/13 9:32m 14...Nxd5
>>>43s 15.Qxh7+ -0.01/13 3:47m 15...Kf8 6s 16.Qh8+ -0.37/13 7:09m 16...Ke7 4s
>>>17.Qxg7 -1.08/12 5:04m 17...Rg8 8:11m 18.Qh7 -1.64/12 0s 18...Nxd4 15:04m
>>>19.exd4 -1.73/13 0s 19...Bxd4+ 2:25m 20.Kh1 -1.73/13 0s 20...Be6 5:47m 21.Re1
>>>-1.88/11 5:58m 21...Kd6 5s (Qb6) 22.Qe4 -1.70/11 6:51m 22...Qf6 4s 23.Rf1
>>>-3.51/9 12:55m 23...Rxg2 4:02m 24.Kxg2 -9.03/11 7:48m 24...Rg8+ 4s 0-1
>>>
>>>Now my questions:
>>>
>>>1. Could someone tell me where exactly the books end?
>>
>>11...Be7 is the last book move.
>>
>>
>>>2. Also could someone explain why Rebel played his book moves with different
>>>times?
>>
>>Autoplayer lag, in DOS it happens.
>>
>>
>>>3. As a general question, does someone know if besides book cooking there can be
>>>operator or tester tuning and if yes please with an example?
>>
>>
>>Naturally it's possible in theory to play through a few lines and see how any
>>program would proceed. Note; I am not saying in this case that happened.
>>
>>
>>>4. In my eyes we have a book cooking in the example above. Shredder ends with
>>>Qh4, is that right?
>>
>>Yes Qh4 is last book move.
>>
>>
>>>5. The cook is Be7, right?
>>
>>
>>Be7 is the last Rebel book move. It's also a book move for many other programs.
>>Simply look at Fritz 7.ctg and look at blacks score with this move.
>>
>>
>>>6. Look at the position after 13...Bd7. Could I give you this position to be
>>>examined with your favorite progs? Who plays Nxd5? I already checked that FRITZ
>>
>>How long did you check with Fritz for? and what version? Takes Fritz 7 a fair
>>time to get off Nxd5 on my PC AMD @1400 mhz.
>>
>>r2qr1k1/pp1b1ppp/2n2n2/2bp4/3N3Q/2NBPP2/PP4PP/R1B2RK1 w - - 0 1
>>
>>Analysis by Fritz 7:
>>
>>14.Nxd5
>>  =  (0.19)   Depth: 8/26   00:00:00  299kN
>>14.Nxd5--
>>  =  (-0.09)   Depth: 9/29   00:00:00  548kN
>>14.Nxd5 Nxd5 15.Qxh7+ Kf8 16.Qh8+ Ke7 17.Qxg7 Rg8 18.Nxc6+ Bxc6 19.Qh6
>>  =  (-0.19)   Depth: 9/29   00:00:01  791kN
>>14.Rd1!
>>  =  (0.00)   Depth: 9/33   00:00:06  4138kN
>>14.Rd1 h6 15.Kh1 Nxd4 16.exd4 Bd6 17.Nb5 Bb8
>>  =  (-0.12)   Depth: 9/33   00:00:07  5039kN
>>14.Rd1 h6 15.Na4 Be7 16.Nf5 Ne5 17.Nxe7+ Qxe7 18.Nc3 Nxd3 19.Rxd3 Qe5
>>  =  (-0.19)   Depth: 10/28   00:00:15  10129kN
>>14.Rd1 h6 15.Kh1 Rc8 16.Nf5 Ne5 17.Bc2 Nc4 18.Nxd5 Nxd5
>>  =  (-0.22)   Depth: 11/30   00:00:36  24384kN
>>14.Nxd5!
>>  =  (-0.19)   Depth: 11/30   00:00:38  25868kN
>>14.Nxd5 Nxd5 15.Qxh7+ Kf8 16.Qh8+ Ke7 17.Qxg7 Rg8 18.Qh7 Kf8 19.Qh6+ Ke8 20.Nxc6
>>Bxc6
>>  =  (-0.06)   Depth: 11/32   00:00:39  27029kN
>>14.Nxd5 Nxd5 15.Qxh7+ Kf8 16.Qh8+ Ke7 17.Qxg7 Rg8 18.Qh7 Qb6 19.Bc4 Bxd4 20.Bxd5
>>  =  (-0.09)   Depth: 12/34   00:00:55  37871kN
>>14.Nxd5--
>>  ³  (-0.37)   Depth: 13/40   00:01:42  69562kN
>>14.Nxd5
>>  ³  (-0.37)   Depth: 13/40   00:01:54  78003kN
>>14.a3!
>>  ³  (-0.34)   Depth: 13/40   00:02:19  95976kN
>>
>>(, MyTown 15.09.2002)
>
>
>Yes, you are right, here is what I get so far, analysis still running:
>
>Shredder 6 - XP Machëide x2cou_51
>[D] r2qr1k1/pp1b1ppp/2n2n2/2bp4/3N3Q/2NBPP2/PP4PP/R1B2RK1 w - - 0 1
>
>Analysis by Fritz 7:
>
>14.Nxd5 Nxd5 15.Qxh7+ Kf8 16.Qh8+ Ke7 17.Qxg7 Rg8
>  ²  (0.47)   Depth: 7/21   00:00:00  82kN
>14.Nxd5--
>  =  (0.19)   Depth: 8/24   00:00:00  145kN
>14.Nxd5
>  =  (0.19)   Depth: 8/24   00:00:01  252kN
>14.Nxd5--
>  =  (-0.09)   Depth: 9/29   00:00:04  509kN
>14.Nxd5 Nxd5 15.Qxh7+ Kf8 16.Qh8+ Ke7 17.Qxg7 Rg8 18.Nxc6+ Bxc6 19.Qh6
>  =  (-0.19)   Depth: 9/29   00:00:04  750kN
>14.Rd1
>  =  (0.00)   Depth: 9/33   00:00:26  5473kN
>14.Rd1 h6 15.Na4 Be7 16.Nf5 Ne5 17.Nxe7+ Qxe7 18.Nc3 Nxd3
>  =  (-0.19)   Depth: 10/29   00:00:43  9578kN
>14.Rd1 h6 15.Nce2 Ne5 16.Bc2 Qb6 17.b3 Rac8
>  =  (-0.19)   Depth: 11/31   00:02:09  29771kN
>14.Rd1 h6 15.Nce2 Ne5 16.b3 Nxd3 17.Rxd3 Qb6 18.Nf4 Rac8 19.Bb2
>  =  (-0.22)   Depth: 12/33   00:04:00  56324kN
>14.Nxd5
>  =  (-0.19)   Depth: 12/35   00:04:20  60874kN
>14.Nxd5 Nxd5 15.Qxh7+ Kf8 16.Qh8+ Ke7 17.Qxg7 Rg8
>  =  (-0.09)   Depth: 12/36   00:04:42  66542kN
>14.Nxd5 Nxd5 15.Qxh7+ Kf8 16.Qh8+ Ke7 17.Qxg7 Rg8 18.Qh7 Qb6 19.Qe4+
>  =  (-0.22)   Depth: 13/37   00:07:20  106087kN
>14.Nxd5--
>  ³  (-0.50)   Depth: 14/40   00:18:34  274357kN
>14.Nxd5
>  ³  (-0.50)   Depth: 14/40   00:22:55  342102kN
>14.Rf2
>  ³  (-0.47)   Depth: 14/40   00:40:14  601615kN
>14.Rd1
>  ³  (-0.44)   Depth: 14/40   00:51:18  766626kN
>
>(P 500, 16 MB hash only, engine in CB8, MyTown 15.09.2002)
>
>It's interesting that a different Hash and a different update (?) brings such
>differences. Also some changes for R etc. could be important.
>
>
>
>
>
>>
>>
>>Shredder 6 also plays Nxd5 as it did in the game.
>>
>>r2qr1k1/pp1b1ppp/2n2n2/2bp4/3N3Q/2NBPP2/PP4PP/R1B2RK1 w - - 0 1
>>
>>Analysis by Shredder 6:
>>14.Nxd5 Nxd5 15.Qxh7+ Kf8 16.Qh8+ Ke7 17.Qxg7 Qb6 18.Qg5+ Nf6 19.Nf5+ Bxf5
>>20.Qxf5
>>  ±  (1.22)   Depth: 7/14   00:00:00  202kN
>>14.Nxd5 Nxd5 15.Qxh7+ Kf8 16.Qh8+ Ke7 17.Qxg7 Qb6 18.Qg5+ Nf6 19.Nf5+ Bxf5
>>20.Qxf5 Bxe3+ 21.Bxe3 Qxe3+ 22.Kh1
>>  ±  (0.97)   Depth: 8/16   00:00:01  279kN
>>14.Nxd5 Nxd5 15.Qxh7+ Kf8 16.Qh8+ Ke7 17.Qxg7 Qb6 18.Re1 Ncb4 19.Be4 Qf6
>>20.Qxf6+ Kxf6
>>  ±  (0.77)   Depth: 8/16   00:00:01  381kN
>>14.Nxd5 Nxd5 15.Qxh7+ Kf8 16.Qh8+ Ke7 17.Qxg7 Qb6 18.Qg5+ Nf6 19.Nxc6+ bxc6
>>20.Rd1
>>  ±  (1.02)   Depth: 9/18   00:00:02  604kN
>>14.Nxd5 Nxd5 15.Qxh7+ Kf8 16.Qh8+ Ke7 17.Qxg7 Qb6 18.Qg5+ Nf6 19.Nxc6+ Qxc6
>>20.Be4 Qc7 21.Bxb7
>>  ±  (1.38)   Depth: 9/18   00:00:03  869kN
>>14.Nxd5 Nxd5 15.Qxh7+ Kf8 16.Qh8+ Ke7 17.Qxg7 Nf6 18.Bc4 Nxd4 19.Qxf7+ Kd6
>>20.Kh1 Rf8 21.exd4 Rxf7 22.Bxf7
>>  ±  (1.13)   Depth: 10/20   00:00:04  1136kN
>>14.Nxd5 Nxd5 15.Qxh7+ Kf8 16.Qh8+ Ke7 17.Qxg7 Nf6 18.Nb3 Bd6 19.Bd2
>>  ±  (1.05)   Depth: 10/20   00:00:06  1525kN
>>14.Nxd5 Nxd5 15.Qxh7+ Kf8 16.Qh8+ Ke7 17.Qxg7 Rg8 18.Nxc6+ Bxc6 19.Qe5+ Kf8
>>20.Bd2 Qb6 21.Kh1 Nxe3 22.Bxe3 Bxe3
>>  ±  (0.80)   Depth: 11/22   00:00:09  2421kN
>>14.Nxd5 Nxd5 15.Qxh7+ Kf8 16.Qh8+ Ke7 17.Qxg7 Rg8 18.Nxc6+ Bxc6 19.Qe5+ Kf8
>>20.Bd2 Qb6 21.Be4 Nxe3
>>  ²  (0.30)   Depth: 11/22   00:00:11  3026kN
>>14.Nxd5 Nxd5 15.Qxh7+ Kf8 16.Qh8+ Ke7 17.Qxg7 Rg8 18.Qh6 Nxd4 19.exd4 Bxd4+
>>20.Kh1 Qb6 21.Re1+ Kd8 22.Qh5 Be6
>>  ²  (0.28)   Depth: 11/22   00:00:16  4178kN
>>14.Nxd5 Nxd5 15.Qxh7+ Kf8 16.Qh8+ Ke7 17.Qxg7 Rg8 18.Qh6 Nxd4 19.exd4 Bxd4+
>>20.Kh1 Qb6 21.Re1+ Be6 22.Bg5+ Nf6 23.Rab1 Rae8
>>  =  (0.18)   Depth: 12/24   00:00:36  9454kN
>>14.Nxd5 Nxd5 15.Qxh7+ Kf8 16.Qh8+ Ke7 17.Qxg7 Rg8 18.Qh6 Nxd4 19.exd4 Bxd4+
>>20.Kh1 Qb6 21.Re1+ Be6 22.Bf5 Rae8 23.Bg5+ Bf6
>>  =  (-0.03)   Depth: 13/26   00:01:27  22419kN
>>
>>Note in the game Shredder went to depth 13 in 9:32 but the PC used was much
>>slower.
>
>Yes, ok. But please note that I did not assume that the moves of Shredder were a
>fake. I have a totally different hypothesis. Among others.
>
>Here comes.
>
>I. My question is what someone could do by tuning the engine. How he could find
>the idea. What I assume for this single game is that the tuning here either was
>unimportant because the game is lost after Nxd5 or the tuning was perfected to
>'win' the won game for sure. Could you or someone explain how the usual tuning
>happened?
>
>II. My opinion is that such a game where Shredder plays into a lost game out of
>the opening does not prove anything about a successful tuning of Rebel.
>
>III. Why there is no comparison in the presentation? Say Rebel (normal) plays
>suchand Macheide 51 plays such? THen it would be much easier to follow the
>process.
>
>IV. My hypothesis therefore: You can't prove the successful tuning with a deadly
>lost game by the opposing prog(here Shredder).
>
>V. Interesting note: therefore I conclude from IV. that a sort of statistics is
>unnecessary here. Here we must analyse the single games. Simply adding the
>results and making conclusions is wrong. It's interesting that Thorsten, who is
>always talking about his ability to make conclusions from a single game does not
>talk about his reflections of such a game here but he is presenting a series of
>games to let the numbers of performance speak. BTW Ed does it with his control
>games in the same wrong manner.
>
>VI. Hypothesis: You should not add up game results of "apples" games and "beans"
>games. Therefore it is important to examine the games prior any counting
>processes.
>
>
>
>
>>
>>
>>>7 does not play Nxd5. It's clear that Black gets a tempo with Rg8 and then can
>>>take on d4.
>>
>>
>>Ok agreed but Shredder plays Nxd5 so what would you have him do?
>
>You give a good question, but here it wasn't Shredder under the microscope but
>the process of tuning for Rebel...
>
>
>>
>>
>>>7. Could someone explain how K-safety could be involved in this special game?
>>>Black takes he risk of a wandering K to d6 and White is "safe" behind his pawns.
>>>Why Shredder, the multi-time Wch cannot see that the open g-file brings storm
>>>for his K?
>>
>>
>>Shredder plays the moves shown, it's a book line in which it ends playing a poor
>>move on a slow PC.
>
>All ok. I asked because I wanted to know if such a play is a well known
>(insider!) weakness. Conclusion then: such a game has no meaning at all in the
>presentation of the results of a tuning process.
>
>Thanks for the many data. For the first time I understood how to save the
>complete analysis simply by pushing it into the clipboard. Could you comment on
>the specific details of my data?


I don't think I see data, I see questions. Questions I could answer but since my
answers are not the answers of the person in question, it's easier to leave that
to Thorsten.

Sarah.







This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.