Author: Uri Blass
Date: 02:03:59 12/26/02
Go up one level in this thread
On December 25, 2002 at 21:18:35, Scott Gasch wrote: >On December 25, 2002 at 18:37:48, Uri Blass wrote: > >>On December 25, 2002 at 17:57:15, Scott Gasch wrote: >> >>>On December 25, 2002 at 16:31:55, Uri Blass wrote: >>> >>>>On December 25, 2002 at 16:18:20, Scott Gasch wrote: >>>> >>>>>On December 25, 2002 at 13:09:45, Sune Larsson wrote: >>>>>> >>>>>> 5rk1/1ppq1rb1/p2p1nn1/3Pp3/2P1Ppp1/PPN3P1/3BQPB1/2R2RK1 w - - 0 25 >>>>>> >>>>>... >>>>>> >>>>>> CM 9000Utz12 25.Qd3 -0.15 Ply 14 64 Mb hash >>>>>> Shredder PB 25.Qd3 -0.28 Ply 14 128 Mb hash >>>>>> Hiarcs 8 25.Qd3 -0.55 Ply 13 128 Mb hash >>>>>> Fritz 8 25.Qd3 -0.59 Ply 14 128 Mb hash >>>>>> Ruffian 25.Qd3 -0.69 Ply 14 128 Mb hash >>>>>> Chess Tiger14 25.Bh1 -1.50 Ply 14 96 Mb hash >>>>>> GambitTiger 2 25.Qd3 -2.26 Ply 14 96 Mb hash >>>>>> >>>>> >>>>> Monsoon -1.17 for white static eval at root >>>>> Monsoon 25. Qd3 -4.18 Ply 14 512 Mb hash >>>>> >>>>>If possible, monsoon is too pessamistic about this position? It is unhappy with >>>>>white's position when I run a static evaluation of the position and when I let >>>>>it search it gets worse and worse. By ply 14, as you can see, it's much more >>>>>unhappy than any of these better engines. >>>> >>>>Can you explain how do you get -1.17 static evaluation? >>>> >>>>material is equal and no squares near the king are attacked by black pieces. >>>>I also do not see significant mobility advantage for black(white has 31 moves >>>>and after Qd3 black has 34 moves). >>>> >>>>A possible reason to see big adantage for black is mobility of the bishop after f3 Bh1 but you need 2 plies for that and you cannot see it by static evaluation. >>>> >>> >>>Monsoon thinks: >>> >>>Black has a slight edge in pawn structure, mainly because of perceived defects >>>on white's structure at c4 and f2. >> >>I do not see special defects at c4 and f2(c4 is even defended by a pawn) >> >>> >>>White's knight at c3 is scored a bit better than either of black's knights. >>> >>>White's bishops are both penalized because the position looks somewhat closed >>>and neither has very good mobility. I will not count f3, e3 or h3 as mobility >>>squares because of the black pawns on f4 and g4. I will not count f4 as a >>>mobility square because of the pawn at e5. >>> >>>Black's rooks are positioned strongly to attack white's king and support the >>>pawn storm. >> >>How do you evaluate it? >>As an human I see that black rooks are strong but computer needs exact >>definition. >> >>White's rook at f7 seems not to have mobility(only one square to go) > >Monsoon does not care about rook mobility before the endgame. It also counts >other rooks as "transparent" when it does mobility so the rook at f7 gets 2 sq. >mobility. > >>> >>>White's king safety penalty is higher than black's because black has more pieces >>>defending the king. >> >>How do you see it as a computer? >>You can claim that white has only the rook to defend the king and black has rook >>f8 and knight at f6 but I guess that you do not mean it. > >My eval counts threats to the king. When it scores a piece, it decides whether >that piece is in position to attack the enemy king, defend the friendly king, or >is just floating around. If a side has many pieces in position to defend the >friendly king I reduce the number of king safety threats the other side has. >This has the effect of not penalizing that side as much for defects to king >safety (like the open H file in the position we are examining). I do not understand from your comments how do you do it. please be more specific. Do you evaluate pieces that defend squares near the king as pieces that defend the king(in this case you can say that the rooks bishop and knights of black defend square near the king and even the queen defends f7 that is near the king). What squares do you consider as squares near the king(all the squares with distance one from the king or more squares)?. > >> >> >>> White has a pawn storm to worry about while black does not. >> >>How do you see it as a computer? > >When I score pawns I see if they are storming the other king's position. These >count as threats to king safety... I do not understand how do you do it. You can say that f4 threats to capture g3 and g3 near the king so you give f4 a bonus but I am not sure if this is what you do based on your reply. > >My eval is good at some things and terrible at others. I do not think king >safety is one of my strong points but it figures out some positions. Maybe this >one is did a good job with. > >Scott You seem to evaluate a lot and I bet that most of the top programs do not evaluate so much. I still did not evaluate king safety because I do not want to evaluate something like no pawns near the king is bad and this example proves it(I thought about evaluating no pawns near the king are bad only if the opponent threats the releavant squares but in this case there are no direct squares at distance one from the white king that black control). The problem about king safety is that before writing a code I need a plan what to evaluate and I do not know exactly how to explain the computer things about it. Your explanations do not give exact definitions in part of the cases and I can only try to guess. Uri
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.