Page 5 of 7

Re: UCI Win/Draw/Loss reporting

Posted: Sat Oct 26, 2019 12:38 am
by Daniel Shawul
I missed this discussion.
So what is agreed up on: info wdl 200 600 200 format for uci, and what for xboard?
Thanks.

Re: UCI Win/Draw/Loss reporting

Posted: Sat Oct 26, 2019 8:49 am
by Ras
Daniel Shawul wrote:
Sat Oct 26, 2019 12:38 am
So what is agreed up on: info wdl 200 600 200 format for uci
I'd add in to make the latter two numbers optional. So "info wdl X" with 0<=X<=1000 would mean an expected score, but no information on the distribution of wins, draws and losses as e.g. LC0 only gives a score expectation. Always permill because other fractions in UCI are also permill.

If used on a line of its own for now, the GUI won't even need to configure anything via an activation boolean since it won't be confused if it doesn't know that tag.

Re: UCI Win/Draw/Loss reporting

Posted: Tue Oct 29, 2019 8:08 am
by Gian-Carlo Pascutto
Daniel Shawul wrote:
Sat Oct 26, 2019 12:38 am
I missed this discussion.
So what is agreed up on: info wdl 200 600 200 format for uci, and what for xboard?
Thanks.
I don't think anything has been agreed upon, but Fritz 17 is now confirmed to ship with a modified lc0 that has the patch pointed out in the original post: https://github.com/DanielUranga/lc0/com ... 0df2f06db2

So I'll just use that, defaulted on, and provide a (non UCI_ prefixed) option to disable it in case anyone ever runs into an "UCI" interface that breaks on it (which it shouldn't).

Re: UCI Win/Draw/Loss reporting

Posted: Tue Oct 29, 2019 3:09 pm
by hgm
What is the usual policy of interfaces for UCI_ prefixed options they do not know? Are these completely ignored, or are they treated as any other engine-defined option, shown in the Engine Settings dialog?

If the latter (which is also what UCI2WB does) is common, it might be better if you used the proposed UCI_pvinfos option for switching the behavior, except that you give it "wdl" as default value.

Re: UCI Win/Draw/Loss reporting

Posted: Tue Oct 29, 2019 8:44 pm
by Ras
hgm wrote:
Tue Oct 29, 2019 3:09 pm
What is the usual policy of interfaces for UCI_ prefixed options they do not know?
The UCI spec says:
If the GUI gets an unknown Option with the prefix "UCI_", it should just ignore it and not display it in the engine's options dialog.
Given how the whole spec misuses "should", this is meant as requirement, not recommendation.

Re: UCI Win/Draw/Loss reporting

Posted: Thu Oct 31, 2019 10:18 am
by crem
Gian-Carlo Pascutto wrote:
Tue Oct 29, 2019 8:08 am
Daniel Shawul wrote:
Sat Oct 26, 2019 12:38 am
I missed this discussion.
So what is agreed up on: info wdl 200 600 200 format for uci, and what for xboard?
Thanks.
I don't think anything has been agreed upon, but Fritz 17 is now confirmed to ship with a modified lc0 that has the patch pointed out in the original post: https://github.com/DanielUranga/lc0/com ... 0df2f06db2

So I'll just use that, defaulted on, and provide a (non UCI_ prefixed) option to disable it in case anyone ever runs into an "UCI" interface that breaks on it (which it shouldn't).
I'll probably do UCI_ShowWDL (default off), but let's see what voting on discord days.

Re: UCI Win/Draw/Loss reporting

Posted: Thu Oct 31, 2019 2:17 pm
by hgm
I think that is a bad idea to have to introduce a separate standard option for every future extension of the UCI info command. Much better to solve it once and for all with a single option.

For CECP I will probably define a new string feature:

feature infos="name1,name2,..."

to indicate the engine will send some extra numbers in front of the PV that it wants displayed in the Engine Output window in columns with the given names as header.

Re: UCI Win/Draw/Loss reporting

Posted: Thu Oct 31, 2019 3:33 pm
by Fulvio
Gian-Carlo Pascutto wrote:
Tue Oct 29, 2019 8:08 am
So I'll just use that, defaulted on, and provide a (non UCI_ prefixed) option to disable it in case anyone ever runs into an "UCI" interface that breaks on it (which it shouldn't).
The drawback of this approach is that the user will see an option which does nothing.
Let's take a GUI which is not updated to support the WDL:
https://github.com/benini/scid/blob/git ... i.tcl#L147
The extra wdl info will simply be ignored, but in the config dialog a non UCI_ option will appear.
And users will be confused by a WDL option that can be set to false or true without any visible difference.

That's the rationale why UCI_ options default to false and why it is recommended to show those options only if they are supported.

Re: UCI Win/Draw/Loss reporting

Posted: Thu Oct 31, 2019 3:38 pm
by Fulvio
hgm wrote:
Tue Oct 29, 2019 3:09 pm
if you used the proposed UCI_pvinfos option
Please, consider at least a minimum consistency.
All the options start with a capital letter and use the camel case:
http://wbec-ridderkerk.nl/html/UCIProtocol.html

Re: UCI Win/Draw/Loss reporting

Posted: Thu Oct 31, 2019 3:44 pm
by hgm
I thought UCI options were not case sensitive.