Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Rybka 2.3.2a evaluation
- - By RVisitor (*) [us] Date 2008-05-03 20:18
Consider this position and evaluation:

FEN:
5k2/6RR/r7/p2p3p/Pn2pP1P/2p1P3/5K2/8 b - - 1 49


23  26:54   159.365.908  101.056  +2.82  49. ... Rh6 50.Rf7+ Kg8 51.Rhg7+ Kh8 52.Rb7 c2 53.Rgc7 Nd3+ 54.Ke2 c1Q 55.Rb8+

  22  26:54   140.160.280  102.408  +3.24  49. ... Nd3+ 50.Ke2 c2 51.Rc7 c1Q 52.Rh8+

  22  26:54   142.094.485  102.420  +3.24  49. ... c2 50.Rb7 Nd3+ 51.Ke2 Kg8 52.Rbg7+ Kf8 53.Rc7 c1Q 54.Rh8+

Rybka2.3.2a evaluates the position as winning for white, but the PV's end in checkmate while the numbers indicate something else.
 
Parent - - By Permanent Brain (*****) Date 2008-05-03 21:16
This is unusual. But it has been explained that the PVs which are displayed, are constructed from the hash tables, "somehow" (it would require a programmer to explain what that means, in detail). What I have understood from it, which is not much :-) is, that such strange looking differences between evaluation and variation can therefore happen and are more or less normal. But it should happen very rarely.
Parent - By Vasik Rajlich (Silver) [hu] Date 2008-05-05 21:29

> What I have understood from it, which is not much  is, that such strange looking differences between evaluation and variation can therefore happen and are more or less normal.


This is true, but it won't cause PVs which end in mate.

The problem here is that Rybka 2.3.2a has a bug in displaying underpromotions in the PV. The c1/Q should actually be c1/N+.

Vas
Up Topic Rybka Support & Discussion / Rybka Discussion / Rybka 2.3.2a evaluation

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill