Search found 377 matches

by noobpwnftw
Sun May 21, 2017 5:25 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Probe EGT in quiescence?
Replies: 28
Views: 6236

Re: Probe EGT in quiescence?

Longest DTC I have so far: http://i.imgur.com/mCOHrDa.png I think we don't need to worry too much about DTx sizes, they will eventually grow too large when you add more and more pieces. My current set of corresponding WDL tablebases already reached 2TB after compression, so I'm more interested in: -...
by noobpwnftw
Sun May 21, 2017 4:40 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Probe EGT in quiescence?
Replies: 28
Views: 6236

Re: Probe EGT in quiescence?

Probing DTx tablebases during search is not really necessary, you can simply probe WDL tablebases and give it a pesudo-mate score, then probe DTx tablebases when you actually reached that position. I'm hosting a comprehensive set of DTC/DTM tablebases online, with a public query interface. Also acce...
by noobpwnftw
Sun May 21, 2017 5:29 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: How far away are we from deep learning Stockfish, Komodo,...
Replies: 48
Views: 14890

Re: How far away are we from deep learning Stockfish, Komodo

My idea, if we can use NN to filter out X% from all possible moves with Y% overall accuracy while slowing down the engine by no more than Z% and manage to achieve (1 - X) * Y > Z, that can be an improvement. I just wonder isn't that what we've always been working on, null move, probcut, history heur...
by noobpwnftw
Sat May 20, 2017 8:29 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: How far away are we from deep learning Stockfish, Komodo,...
Replies: 48
Views: 14890

Re: How far away are we from deep learning Stockfish, Komodo

Let me just re-phrase my one-line sentence into many. Why NN is good for Go, as described by HGM, is because currently there is no better modeling/scoring methods than heuristics from machine learning, plus the nature of Go: search space gradually shrinks and MCTS friendly. When it comes to the ques...
by noobpwnftw
Fri May 19, 2017 8:36 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: How far away are we from deep learning Stockfish, Komodo,...
Replies: 48
Views: 14890

Re: How far away are we from deep learning Stockfish, Komodo

After some garage experiments, I think NN-based eval for chess is pretty much a joke even compared to current PST+MAT eval alone.

At most, it might be good for move pruning and sorting, prove me wrong.
by noobpwnftw
Fri May 05, 2017 10:12 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Search-Based Opening Book Construction
Replies: 10
Views: 3115

Re: Search-Based Opening Book Construction

my advice: drop all probability factors, and never ever minimax your results, use generalization algorithm.

your engine can use it as a PV line suggester and let the search verifies/self-play against it so that while you boost your search you are also expanding the book.
by noobpwnftw
Wed Mar 22, 2017 9:32 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Cumulative building of a shared search tree
Replies: 19
Views: 6474

Re: Cumulative building of a shared search tree

Now, this project has reached a stable status. Everything seems working fine and the API gives acceptable query performance for engine usage during search. My dedicated compute cluster is running at almost 1 billion NPS, not including the nodes working on move sieving and retrograde scoring. The dat...
by noobpwnftw
Mon Mar 06, 2017 11:40 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: The Inferno thread
Replies: 34
Views: 6441

Re: The Inferno thread

Just random thoughts.
You may want to consider some kind of A* search for jumping sliders since it can do both move generation and path evaluation.
by noobpwnftw
Mon Mar 06, 2017 12:52 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Deep Pink: a chess engine using deep learning
Replies: 20
Views: 7230

Re: Deep Pink: a chess engine using deep learning

2. "Players will choose an optimal or near-optimal move" is a bad assumption for the FICS dataset. I wonder if anyone has ever tried to use tablebases to train a deep learning engine. After all the code to create all the positions and correct moves already exists; it is simply a matter of adding th...
by noobpwnftw
Mon Mar 06, 2017 12:16 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Generate Opening Book from Scratch
Replies: 10
Views: 3083

Re: Generate Opening Book from Scratch

I've been working on something similar, just not for chess. Recently I saw a couple of threads here talking about things regarding to automated opening book generation / persistent database and I can make a TODO list from my experience by taking the architecture of a standard MCTS: - Position evalua...