Author: Stuart Cracraft
Date: 07:08:17 08/05/04
Go up one level in this thread
That's it. Thanks. On August 04, 2004 at 15:10:42, José Carlos wrote: >On August 04, 2004 at 14:52:45, Cesar Contreras wrote: > >>I'm having problems figuring out how to implement an SEE, i supose you need to >>generate all moves to both sides, ordering on any way the captures, and >>calculate the captures. Any info you can add will be apreciated. >> >>Thx > > You don't need to generate moves, only attacks to a given square. The idea of >SEE (static exchange evaluator) is to predict the outcome of a capture sequence >in a given square statically. So you want to know all pieces of both colors that >attack that square. Then, your sort the pieces by increasing value (try to >capture with small pieces first) and start simulating the captures, keeping >track of the material. > Consider this position: > >[D]rnbqkb1r/pp2pppp/2p2n2/3p4/4P3/2NP4/PPP2PPP/R1BQKBNR w KQkq - 0 4 > > You want to know the SEE of the two possible captures: > 1. Nxd5: black can take with pawn, knight and queen, in this order. Black >takes with pawn (you eliminate Pc6 from the list). Now you're at -2.00 >(1.00-3.00). You can take with Pe4. exd4 and you're at -1.00; you can stop here >because you know for sure you're gonna lose material with Nxd5 > 2. exd5: again cxd5, Nxd5 and Qxd5 for black. First cxd5. Eval=0. You still >have one attacker, the knight: Nxd5 (+1.00). Black Nxd5 (-2.00) for you. You >have no more attackers, so you review the capture list. You see you can sit >after black's cxd5 with eval 0. So SEE value for exd5 is zero. > > This is the idea. The implementation is more fun if you figure out ;) > > José C.
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.