Search found 2903 matches

by Evert
Thu May 31, 2018 12:30 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Move ordering based on stored values of child nodes
Replies: 4
Views: 1205

Re: Move ordering based on stored values of child nodes

With current move ordering heuristics, you get a cut-off from the first three moves in 95-99% of cases. That's the statistic you have to beat. Any effort spent on move ordering in ALL-nodes is wasted. Any effort in improving move ordering in a CUT-node is wasted in 95-99% of cases, so whatever you p...
by Evert
Tue May 22, 2018 8:52 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: 7-men Syzygy attempt
Replies: 588
Views: 120307

Re: 7-men Syzygy attempt

Ok, so I'm sure I've asked this before, but I don't remember the explanation and I can't find it using the search. Assuming I have N-piece DTM and DTZ50 tables. Can't I get DTM50 optimal play (not value) by first reducing the move list to those moves that preserve the win (according to DTZ50), and ...
by Evert
Tue May 22, 2018 6:11 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: 7-men Syzygy attempt
Replies: 588
Views: 120307

Re: 7-men Syzygy attempt

DTM50 conceptually stores a distance-to-mate value for each pair (position, rule50_counter). So that is 100x as much raw data as for DTM. The sequence of values for a particular position is far from random, but it is not immediately obvious how to find an efficient encoding that fits into a good co...
by Evert
Fri May 18, 2018 5:56 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Issue with self play testing
Replies: 5
Views: 1225

Re: Issue with self play testing

Upon reflection, I see that the Elo gain is possibly 2x that: the old Ares never made such attacks and thus the ability to defend against them went untested and unmeasured. Yes, and that's why the gain is typically less than it is in self-play: the other opponent may not have been so blind, so you ...
by Evert
Fri May 11, 2018 6:42 am
Forum: Computer Chess Club: General Topics
Topic: Regarding Talkchess update and features
Replies: 33
Views: 5274

Re: Regarding Talkchess update and features

Can you activate the option to automatically open new links in new tabs though? Currently if you forget to click with the middle button or with ctrl, it's quite annoying. Urgh, please don't! Nothing pisses me off more* than a website deciding for me that I want to open a link in a new tab. If I wan...
by Evert
Fri May 04, 2018 2:17 pm
Forum: Computer Chess Club: General Topics
Topic: LCZero Accomplishments and Goals Thus Far
Replies: 177
Views: 24009

Re: LCZero Accomplishments and Goals Thus Far

You're right of course that you cannot make a straightforward comparison. So how about this? Leela is 2900 CCRL when paired with a GTX 1060. The CPU and the rest is identical so that's the only difference. And since Leela is designed to to make use of a GPU, without which it is quite clearly crippl...
by Evert
Thu May 03, 2018 1:59 pm
Forum: Computer Chess Club: General Topics
Topic: LCZero Accomplishments and Goals Thus Far
Replies: 177
Views: 24009

Re: LCZero Accomplishments and Goals Thus Far

Remember that LCZ first began self-learning in February. It is now over 2900 CCRL. That is quite incredible if you ask me. I wonder if trying to express the strength of Leela "on CCRL scale" isn't grosly misleading. CCRL builds a rating list under specific conditions, which are scaled/tweaked in a ...
by Evert
Fri Apr 27, 2018 9:09 pm
Forum: Computer Chess Club: General Topics
Topic: A question abou LCZero for the experts
Replies: 5
Views: 1265

Re: A question abou LCZero for the experts

I would be keen to know if there is some remarkable and fresh insight contained in this nebulous pile of digits. That would, indeed, be extremely interesting. If it could be done. If we can't can someone give me the layman's explanation of why it cannot be done. The amount of data is enormous. It's...
by Evert
Fri Apr 27, 2018 5:28 am
Forum: Computer Chess Club: General Topics
Topic: LC0 on the WAC test suite, bad results
Replies: 54
Views: 10765

Re: LC0 on the WAC test suite, bad results

Once again: how do human players find candidate moves? I think it should be obvious that it's based on pattern recognition and it's very similar to policy NN of LC0 / A0. Humans definitely do not play chess by exhausting all variations within given depth. So if I would rephrase your sentence with b...
by Evert
Wed Apr 25, 2018 8:37 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Explaining heuristics with computer science, graph theory.
Replies: 19
Views: 2389

Re: Explaining heuristics with computer science, graph theor

to determine if the king can be mated in a corner or not, and which pieces/pairs of pieces have mating potential. How do you determine which pieces have mating potential? Sure it could be brute forced, but are there any tricks you do to speed things up? Sure. There are a couple of things. First, th...