Some engines shows in Arena what I think is the "extended" depth reach in quiescence search using a bar"/", something like 8/14 that I guess means 8 plies completely searched and 6 more plies just for some branches.
But I do not see how to do it at least in Winboard, could be just an UCI posibilitie?
I remember old GUIs shows you the PV and at the same time actual search.
Is this possible in Arena?
Thanks!
Show thinking in Arena questions.
Moderators: hgm, Rebel, chrisw
-
- Posts: 464
- Joined: Sat Feb 28, 2015 4:37 pm
- Location: Argentina
-
- Posts: 710
- Joined: Sat Dec 06, 2014 1:53 pm
Re: Show thinking in Arena questions.
info ... seldepth = selective depthStockfish 160716 64 POPCNT by T. Romstad, M. Costalba, J. Kiiski, G. Linscott
go depth 2
info depth 1 seldepth 1 multipv 1 score cp 83 nodes 20 nps 20000 tbhits 0 time 1 pv e2e4
info depth 2 seldepth 2 multipv 1 score cp 86 nodes 52 nps 26000 tbhits 0 time 2 pv e2e4 b7b6
bestmove e2e4 ponder b7b6
discussed hereBut I do not see how to do it at least in Winboard, could be just an UCI posibilitie?
http://www.open-aurec.com/wbforum/viewt ... 1930[quote][/quote]
-
- Posts: 27809
- Joined: Fri Mar 10, 2006 10:06 am
- Location: Amsterdam
- Full name: H G Muller
Re: Show thinking in Arena questions.
I don't think Arena shows selective depth of WB engines this way. They would probably be shown in the PV field. WinBoard would also only show it in a separate column, which you can open or close at will (for UCI or WB engines alike).
It is a rather meaningless statistic anyway.
It is a rather meaningless statistic anyway.
-
- Posts: 464
- Joined: Sat Feb 28, 2015 4:37 pm
- Location: Argentina
Re: Show thinking in Arena questions.
Thanks Yuri and H.G.!
I need to see it mostly for debug.
Some time ago I made Soberango built a log for debug purpose but never solved the crashes this making of log causes.
I need to see it mostly for debug.
Some time ago I made Soberango built a log for debug purpose but never solved the crashes this making of log causes.
-
- Posts: 27809
- Joined: Fri Mar 10, 2006 10:06 am
- Location: Amsterdam
- Full name: H G Muller
Re: Show thinking in Arena questions.
In WB protocol the PV field of thinking output is free-format text, so you could print any info in there, together with the PV. I would expect a GUI to display that. WinBoard would.
WB protocol defines a standard way for printing selective depth, tablebase hits and CPU load just before the PV (separated from it with a tab). Newer winboard versions would display infoprinted that way in seperate columns in the Engine Output window.
WB protocol defines a standard way for printing selective depth, tablebase hits and CPU load just before the PV (separated from it with a tab). Newer winboard versions would display infoprinted that way in seperate columns in the Engine Output window.
-
- Posts: 193
- Joined: Wed Mar 11, 2015 3:34 am
- Location: United States
Re: Show thinking in Arena questions.
Most definitely not a meaningless value. At least not for engine developers while doing testing. I've done something stupid in quiescence search more times than I care to admit. Being able to see at a glance that selective search depth has jumped to some rediculously high values has saved me a lot of time on multiple occasions. Even being able to see anticipated seldepth changes is valuable.hgm wrote:I don't think Arena shows selective depth of WB engines this way. They would probably be shown in the PV field. WinBoard would also only show it in a separate column, which you can open or close at will (for UCI or WB engines alike).
It is a rather meaningless statistic anyway.
Re: Show thinking in Arena questions.
Hmmm indeed.zd3nik wrote:Most definitely not a meaningless value. At least not for engine developers while doing testing. I've done something stupid in quiescence search more times than I care to admit. Being able to see at a glance that selective search depth has jumped to some rediculously high values has saved me a lot of time on multiple occasions. Even being able to see anticipated seldepth changes is valuable.
Noticed that my program reaches a depth of 30:
info depth 9 seldepth 30 score cp 19 time 3404 nodes 416758 pv d2d4 e7e6 d1d3 b8c6 g1f3 g8f6 a2a4 f8d6
That can't be right, can it?
-
- Posts: 193
- Joined: Wed Mar 11, 2015 3:34 am
- Location: United States
Re: Show thinking in Arena questions.
That seems pretty normal to me. In my engines I think it's safe to say seldepth is about 2 to 3 times the regular depth on average.flok wrote:Hmmm indeed.zd3nik wrote:Most definitely not a meaningless value. At least not for engine developers while doing testing. I've done something stupid in quiescence search more times than I care to admit. Being able to see at a glance that selective search depth has jumped to some rediculously high values has saved me a lot of time on multiple occasions. Even being able to see anticipated seldepth changes is valuable.
Noticed that my program reaches a depth of 30:
info depth 9 seldepth 30 score cp 19 time 3404 nodes 416758 pv d2d4 e7e6 d1d3 b8c6 g1f3 g8f6 a2a4 f8d6
That can't be right, can it?