Cutechess-cli Time Control: Me or Bug

Discussion of chess software programming and technical issues.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
User avatar
ilari
Posts: 750
Joined: Mon Mar 27, 2006 5:45 pm
Location: Finland
Contact:

Re: Cutechess-cli Time Control: Me or Bug

Post by ilari » Mon Sep 04, 2017 1:45 pm

Ferdy wrote:
hgm wrote:That seems a bug.
Care to elaborate why it seems a bug?
I can see both sides of the issue and my original intention was to make the behaviour configurable. But then I received a pull request that implemented the change and as a result Cute Chess no longer adds time increment for book moves. It seems to be the behaviour that most users want. But then again I understand the argument that it shouldn't matter how the move was made.

User avatar
Evert
Posts: 2929
Joined: Fri Jan 21, 2011 11:42 pm
Location: NL
Contact:

Re: Cutechess-cli Time Control: Me or Bug

Post by Evert » Mon Sep 04, 2017 2:51 pm

ilari wrote: I can see both sides of the issue and my original intention was to make the behaviour configurable. But then I received a pull request that implemented the change and as a result Cute Chess no longer adds time increment for book moves. It seems to be the behaviour that most users want. But then again I understand the argument that it shouldn't matter how the move was made.
Making it optional is probably the way to go.

For testing (during development), it's perhaps useful to disable because you want to test at a specific time control, but this gives "free" time to the engine. If you're using a book where not all lines end at the same depth, then not all games end up using the same time. Presumably the tester knows what he's doing though.

For actual matches, not adding an increment seems undesirable. Consider that an engine would get the increment if it were using its own book instead of the GUI handling the book.

User avatar
ilari
Posts: 750
Joined: Mon Mar 27, 2006 5:45 pm
Location: Finland
Contact:

Re: Cutechess-cli Time Control: Me or Bug

Post by ilari » Tue Sep 05, 2017 6:52 am

Evert wrote:
ilari wrote: I can see both sides of the issue and my original intention was to make the behaviour configurable. But then I received a pull request that implemented the change and as a result Cute Chess no longer adds time increment for book moves. It seems to be the behaviour that most users want. But then again I understand the argument that it shouldn't matter how the move was made.
Making it optional is probably the way to go.

For testing (during development), it's perhaps useful to disable because you want to test at a specific time control, but this gives "free" time to the engine. If you're using a book where not all lines end at the same depth, then not all games end up using the same time. Presumably the tester knows what he's doing though.

For actual matches, not adding an increment seems undesirable. Consider that an engine would get the increment if it were using its own book instead of the GUI handling the book.
We had some discussion about this at Github:
* https://github.com/cutechess/cutechess/issues/9
* https://github.com/cutechess/cutechess/pull/174

I remember a lot of conversations here about whether or not engines should use the books instead of the GUI and should those moves be treated like any chess move. The result was always that people agreed to disagree. Personally, I think if the GUI operates the opening book it's the same thing as setting a custom starting position from an EPD file etc., in which case it usually makes sense that both players get the same amount of time.

User avatar
Evert
Posts: 2929
Joined: Fri Jan 21, 2011 11:42 pm
Location: NL
Contact:

Re: Cutechess-cli Time Control: Me or Bug

Post by Evert » Tue Sep 05, 2017 7:07 am

ilari wrote: We had some discussion about this at Github:
* https://github.com/cutechess/cutechess/issues/9
* https://github.com/cutechess/cutechess/pull/174

I remember a lot of conversations here about whether or not engines should use the books instead of the GUI and should those moves be treated like any chess move. The result was always that people agreed to disagree.
Which is probably why an option is the best of both worlds. ;)
Personally, I think if the GUI operates the opening book it's the same thing as setting a custom starting position from an EPD file etc., in which case it usually makes sense that both players get the same amount of time.
Well, that's fine if you want to say that you started from an opening position based on some opening book. Seems wrong to me if you want to claim you ran a match between engines X and Y and used some opening book.
Under XBoard, there is a common defect of engines not properly counting moves made in force mode against the number of moves in each epoch for repeating time control. A similar case could be made that the engines did not think about such moves, therefore they should not be counted against them in terms of time control (and they should not get the increment). In this case you can argue until the cows come home, but XBoard (the GUI) doesn't implement it that way (and the protocol doesn't say to interpret it that way), so doing it like that leads to problems.

So yeah, there are arguments for both sides to be made, so put me up in favour of making this an option (of course that's easier said than done, given that I'm not offering to do the actual work necessary). :)

User avatar
hgm
Posts: 24435
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: Cutechess-cli Time Control: Me or Bug

Post by hgm » Tue Sep 05, 2017 7:23 am

So how are you going to treat classical time controls? Will the book moves be ignored in the move count for the session?

Post Reply