Author: John Merlino
Date: 13:25:13 12/12/00
Go up one level in this thread
On December 12, 2000 at 00:06:23, William Penn wrote: >On December 11, 2000 at 16:06:07, John Merlino wrote: > >>On December 11, 2000 at 14:52:28, William Penn wrote: >> >>>The signs of scores are confused in CM8000. Previously it was confirmed there >>>is a problem with omitting the minus sign when the score number is in the range >>>of -0.01 to -0.99. Now, here is an example of the signs being wrong when the >>>score number is larger than 1.00. >>> >>>BR -- -- BQ BK -- BN BR >>>-- BP BP BB -- -- -- -- >>>-- BB -- -- BP BP -- BP >>>BP WP -- BP WP -- BP -- >>>WP -- -- WP -- -- -- WN >>>-- WN WP WB -- -- -- WP >>>-- -- -- -- -- WP WP -- >>>WR -- WB WQ WR -- WK -- >>> >>>White: a custom personality Black: human Black to move >>> >>>if Chessmaster box checked in Thinking Lines window >>>Time Depth Score Positions Moves >>>0:07 7 5.25 210887 h5 Ng6 Rh7 exf6 Qxf6 Ne5 Rh6 Qc2 >>>0:27 8 5.01 716123 h5 Ng6 Rh7 exf6 Qxf6 Ne5 Rh6 Qd2 >>>1:11 9 5.46 2169809 h5 Ng6 Rh6 Bxg5 fxg5 Qf3 Bxb5 Bxb5+ ... >>>3:46 10 5.71 7208736 h5 Ng6 Rh7 exf6 Qxf6 Ne5 Rh6 h4 ... >>>21:56 11 6.15 42116395 h5 Ng6 Rh7 exf6 Qxf6 Ne5 Rh6 Nf3 ... >>> >>>if White box checked in Thinking Lines window >>>Time Depth Score Positions Moves >>>0:04 7 -5.25 210887 h5 Ng6 Rh7 exf6 Qxf6 Ne5 Rh6 Qc2 >>>0:13 8 -5.01 716123 h5 Ng6 Rh7 exf6 Qxf6 Ne5 Rh6 Qd2 >>>0:35 9 -5.46 2169809 h5 Ng6 Rh6 Bxg5 fxg5 Qf3 Bxb5 Bxb5+ ... >>>1:52 10 -5.71 7208736 h5 Ng6 Rh7 exf6 Qxf6 Ne5 Rh6 h4 ... >>>10:49 11 -6.15 42116395 h5 Ng6 Rh7 exf6 Qxf6 Ne5 Rh6 Nf3 ... >>> >>>Everything is the same in these two charts except the signs of the scores and >>>the times. That's because the custom personality is identical to Chessmaster, >>>just with a different name. The first chart takes twice as much time because >>>another engine was running in the background, unknown to me (inadvertently, i.e. >>>that's another major problem with CM8000!?). >>> >>>Although it would be possible to define a special protocol to make these results >>>acceptable, that would complicate things unnecessarily, and will cause mass >>>confusion because it departs from what almost everyone expects. Again, the >>>standard protocol is that the score is positive when white has an advantage, and >>>is negative when black has an advantage -- independent of who has the move. It >>>should also be independent of which box(es) is checked in the Thinking Lines >>>window! >>> >>>Obviously white is winning in this position, so the signs of the scores should >>>be positive in all cases. >>>WP >> >>I cannot duplicate this. How are you setting up a position with Black to move, >>and then clicking on the White checkbox and getting output? The White engine >>should not start thinking (in this instance) until it is its turn to move? Could >>you give the full move list that arrived at this position? >> >>Also, it should be noted that the bug involving negative scores less than -1.00 >>has been found and fixed. Thanks again for your assitance, >> >>jm > >The move sequence leading to the above position is: >1.Nf3 Nc6 2.d4 e6 3.e4 d5 4.e5 Bb4+ 5.c3 Ba5 6.Bd3 Bd7 7.0-0 h6 8.b4 Bb6 9.a4 a5 >10.b5 Nce7 11.Nbd2 Nf5 12.Nb3 g5 13.h3 f6 14.Re1 Nh4 15.Nxh4 >You can enter the moves as human vs. human, then change the white player to a >custom personality via Ctrl-G. I used Chessmaster#0, just a different name for >Chessmaster, but it's treated as a custom player. Then check the white box and >the analysis begins. Actually it begins as soon as you change the player to a >custom personality, but you must check the white box to see it. To initialize >things, I usually save the game to the hard drive then reload it, which seems to >kill any other engines that might be running in the background, etc. > >Checking the white or black box always shows an active evaluation in progress in >the Thinking Lines window if one of the players is a custom personality. That's >as it should be. I like it that way! When fixing this problem, I hope you don't >kill that kind of versatility!? The only problem (apparently) is that the signs >of the scores are wrong. > >I'm glad to hear that the -0.01 to -0.99 problem was fixed. Just one little baby >step at a time, as Bill Murray said... :) >WP The problem has been found and fixed. It would only occur with infinite time controls, because engines are put into analyze mode in that time control, causing their scores to report differently than if they think they are actually playing one of the sides in the game. Thanks again for the examples and the detailed descriptions, jm
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.