Myracle GUI

[tapio]

the range in is %, 0 = no randomization (weights as specified in the book) to 100% = all nonzero-weight moves have the same weight

Yes, with >0 you start unifying the weights. I meant the opposite direction. With <0 ("less randomization") weights are made more extreme, i.e. a set of probabilities (20 20 20 20 10 5 5 0) will deviate towards only the highest values, towards (25 25 25 25 0 0 0 0) "play only the best move(s)". I hope that makes sense. So you can adjust towards a tendency to preferably play the highest weighted moves.

[mar]

Yes, with >0 you start unifying the weights. I meant the opposite direction. With <0 ("less randomization") weights are made more extreme, i.e. a set of probabilities (20 20 20 20 10 5 5 0) will deviate towards only the highest values, towards (25 25 25 25 0 0 0 0) "play only the best move(s)". I hope that makes sense. So you can adjust towards a tendency to preferably play the highest weighted moves.

ah yes, ok, I could do that

[mar]

build 79 is up:
- randomness for polyglot books, per set
- virtual own book override in engine settings (per engine, path/polyglot only)
- 16 analysis engine sets instead of 8
- fix low contrast text (gray on blue) when hovering over active game in schedule (light mode)

[Elorejano]

Thanks, It was what I was asking for.

[tapio]

Thanks Mar, I think it's working well so far... according to my testing, are my assumptions correct?

a) If a polyglot book which is set in engine options is used, it uses default randomness as set in the book?
b) If I select a book in the "New game" dialog, this will be used instead, by engines?
c) If a polyglot book set in engine options has been used, the UI cannot know? Because AFAICS then in the annotation there is no "book" annotation.

I was thinking, but maybe it's worth another thought and may be due to my preference, the book randomization would be possible in engine options. Because this way you can create a complete engine setting and be done with it. And if you want comparability, override book for all engines via the tournament or game setup.

For some reason my currently favorite piece set Kosal is shown like this:

[vencenda]

Openings don't repeat in sprt tournaments even though the checkbox is checked. This seems like a bug. Sorry if I'm missing something.

[mar]

Openings don't repeat in sprt tournaments even though the checkbox is checked. This seems like a bug. Sorry if I'm missing something.

for sprt, since I only do trinomial and not pentaniomial, repeat reverse makes no sense and it's grayed out. repeat openings is not grayed out for some reason, but it doesn't make sense either (because 1 game = 1 round) so I'll gray it out too
that's ok, that's what you want. sprt is for engine devs to compare perf of two engine versions, not for testers

[mar]

Thanks Mar, I think it's working well so far... according to my testing, are my assumptions correct?

a) If a polyglot book which is set in engine options is used, it uses default randomness as set in the book?
b) If I select a book in the "New game" dialog, this will be used instead, by engines?
c) If a polyglot book set in engine options has been used, the UI cannot know? Because AFAICS then in the annotation there is no "book" annotation.

I was thinking, but maybe it's worth another thought and may be due to my preference, the book randomization would be possible in engine options. Because this way you can create a complete engine setting and be done with it. And if you want comparability, override book for all engines via the tournament or game setup.

For some reason my currently favorite piece set Kosal is shown like this:

a) yes
b) yes, but it behaves like own book - if engine has OwnBook option, it has to be enabled. also, if override engine options is enabled in new tour and "disable own book" is checked, it will be disabled too
c) yes, because it's injected at engine parser level when a new search is started, just exactly the way the GUI doesn't know if the engine has a real own book enabled

that's the correct behavior and I don't intend to change that

of course I could add randomness in eng options too, but that would waste 1 extra column of vertical space. I don't think randomness is such a great idea in general, certainly not for serious eng-eng tests but rather for fun - if you can convince one other person who thinks
this is a great idea it'd be an easy job for me for the next build

I'll look at the piece set, the GUI uses my custom svg parser and renderer so not everything is fully supported, for example there
is no support for custom line caps and line joins (always round only) and no outline dashes, no embedded base64-encoded images
and no effects like blur and no text.
some piece sets like pixel from lila look incorrect, because it relies on full line cap support
maybe I'll implement some more svg-related things in the future but unlikely at the moment

I have to say the SVG is probably the single most overengineered format I've encountered so far,
and I think this goes in line with basically all web technologies that are just a giant pile of crap...
part of the SVG is really cool though, such as the draw commands themselves
I'd vastly prefer the origin to be lower left rather that upper left for a vector format, but it's easy enough to flip in the root transform matrix

[mar]

build 80 is up:
- randomness for per-engine own book override (do not ask for more book-related features, you've spent your 'tokens')
- fix parsing/rendering kosal SVG piece set
- fix: repeat openings is grayed out now for sprt tours as it should be

[vencenda]

Openings don't repeat in sprt tournaments even though the checkbox is checked. This seems like a bug. Sorry if I'm missing something.

for sprt, since I only do trinomial and not pentaniomial, repeat reverse makes no sense and it's grayed out. repeat openings is not grayed out for some reason, but it doesn't make sense either (because 1 game = 1 round) so I'll gray it out too
that's ok, that's what you want. sprt is for engine devs to compare perf of two engine versions, not for testers

Actually I meant the "repeat reverse" option, which was already grayed out, not the "repeat openings" one. I was confused because it was checked even if grayed out. I thought that "checked and grayed out" meant "the openings always repeat and you can't turn it off", which is the behavior I was expecting after I saw it. I now realize that it was checked because I had checked it before.