Author: Steve Coladonato
Date: 11:37:37 09/14/02
Go up one level in this thread
On September 12, 2002 at 15:42:17, Dieter Buerssner wrote:
>On September 12, 2002 at 12:47:30, Steve Coladonato wrote:
>
>>I have played a couple of games with Yace in the Xboard environment. The SCID
>>GUI on Linux does not have the capability of sending a complete game to an
>>engine for analysis. To the best of my knowledge, Xboard does not do it either.
>>Do you know of any GUI's that run on Linux that will support analysis of
>>complete games by Yace or is it only on the Windows platform where this can be
>>done?
>
>I know, that Xboard cannot do this. I thought, Scid could. But I was obviously
>wrong. To be honest, I have not tried many features of Scid. I just investigated
>the "normal" operation of Yace under Scid. I have not the means and time, to
>test out all of the many WB-compatible GUIs. Indeed, if I would be a Linux only
>user, that would be easier. So, it seems, that there is no method to analyze
>games under Linux, with a generic Xboard or UCI engine.
>
>Perhaps, you can convince Shane, to add such a feature. You may also convince
>him to add the UCI protocol - I think some features of it, could make a GUI
>driven analysis of games worthful. (If the engine suggested move, and the game
>move differ, UCI can send to the engine a command to search to the same depth,
>at which the former analysis finished, but only to consider the game move. This
>should make this "research" having less artifacts).
>
>I will perhaps start to work on an annotate command soon. Yace would most
>probably search from back to front. The options should include:
>
>- only analyze the moves of one side
>- give start/end movenumber ("from move 10 to move 40")
>- give some limit, where to comment/show a variation. Say, if the engine
>suggested move seems 0.1 better than the game move. The 0.1 would be user
>adjustable
>- perhaps give 2 different time control methods. One would be, to search each
>move for an exact time. The other would be some sort of average time, and trying
>to finish one depth. For example the engine may stop the analysis, when one
>depth was just finished, and more of half of the suggested time was used (with
>the idea, that the next depth will not be finished with the suggested time). If
>a new depth started, the engine may use more than the suggested time
>- The outcome should be a PGN, that shows scores for each moves, and also
>variants for the moves, were the engine suggested move differs from the game
>move. The variant shown would be the PV.
>
>Are there more things, that could be interesting? How would you like to use such
>a feature?
>
>Regards,
>Dieter
Hi Dieter,
Yes, your thoughts about the "annotate" command are right on. When the
threshold for considering a move an error, the .1 (adjustable) in your comments
above, the PGN output file should show the score for the move played (I don't
believe the variation giving this score is applicable) and the score for what
the engine considered the best move followed by the variation.
The "analyze only one side feature" should be the same as Crafty where you can
choose "black", "white", "both" or by Last Name. The "Last Name" feature is
nice for analyzing pgn files of your own games, where you play both black and
white, and only showing the analysis from your side.
I'll see if I can influence Shane with an "analyze" command. I know he was
considering adding a "play against engine" feature but that one may be more
difficult than just adding the analysis feature.
Thanks,
Steve
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.