Search found 175 matches

by Stan Arts
Mon Mar 31, 2014 12:33 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Thoughs on eval terms
Replies: 21
Views: 4534

Re: Thoughs on eval terms

Sad truth is, (finding this out again currently as lately I've finally regained some interest in computerchess and writing something again.) search is what wins the games and 2-3 real ply extra with a completely empty eval besides a single PSQT for all the pieces seems about enough to overcome a mat...
by Stan Arts
Sun Mar 23, 2014 1:31 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Solving Chess Kickstarter
Replies: 109
Views: 27468

Re: Solving Chess Kickstarter

I won't put the whole output on this page, but according to the program, there are: 3837824955509396206205398894933217531649406893072749958129150771163299840000000 positions in a 32 man database Yea thought that number looked vaguely familiar and then it hit me. Remember running my program overnigh...
by Stan Arts
Sun Mar 16, 2014 9:04 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: fail hard
Replies: 23
Views: 3208

Re: fail hard

I don't think the "best" move should be used for anything in case alpha is not improved upon. It's basically a random move because the search has made no effort to see how bad or not-so-bad all of the "below alpha" moves are. Sure, though pretty funny things get returned in a fail soft search and i...
by Stan Arts
Sun Mar 16, 2014 8:40 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: fail hard
Replies: 23
Views: 3208

Re: fail hard

In fail hard, the lazy approach is indeed to reuse "alpha" as "best_score" and update it accordingly. But often you also need to keep track of the original value of "alpha" that was valid when entering a new node, e.g. for correct tracking of the PV or to store correct "exact/lower bound/upper boun...
by Stan Arts
Sun Mar 16, 2014 1:37 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: fail hard
Replies: 23
Views: 3208

Re: fail hard

The main advantage, probably, of fail hard is that it's easier to implement. For example you don't need a separate eval variable set to -inf at every new node but simply an alpha variable that's allowed to rise and initially set to the value of that of 2 ply before. But that means ofcourse that you ...