Author: scott farrell
Date: 21:10:40 01/13/03
Go up one level in this thread
IN a game chompster vs little goliath. The code didnt kick in as it didnt get directly to bishop on pawns. Ruffian thinks Rc2 is a good idea for 12 plies with a score of +0.4. - and the PV shows lots ofo swapping pawns, and white losing the h pawn. At depth 12 ruffian fails high, and sees it can march the h pawn somewhere up the board, and the score climbs to around +0.5. My chompster without blockade code sees h4 straight away, and see each successive pawn push right to h7, thanx to extensions and qsearch, it sees this quicker to ruffian (although it might be flawed). The score climbs for 1.1 to 1.2, and starts going into EGTBs at depth 12 and doesnt get a score back (not that I could be bothered waiting). With the blockade code enabled for single bishops/knight end games, it takes longer to find h4, with a few fail-lows along the way, presumably to bloackade draws. It prefer Rg7 for a bit, not quite as bad as where ruffian starts, rook swap on g8 - ouch !!! It eventually finds a 2.11 score for h4, but the PV is missing, so I dont really beleive it. With it enabled for anything less than a queen, so it includes all positions in the analysis, it sees Rc2 (then Rc4+) with a score of 0.0 immediately. It thinks really slow in these positions, with bad move ordering, but that might be correct, or it might need tuning. It might be correct, as you have to start trying dumb moves to break the blockade, like more sacs etc. It does eventually find h4, but it takes many many many more nodes to convince itself it isnt a draw. I think what happens when the code kicks in, the assumption is that it is a draw by blockade, unless it can prove it otherwise. As opposed to draw by rep, it assumes it is NOT a draw, unless it can prove it is a draw. ie. the null hypothesis is the opposite. I am only going to enable to code for single bishops/knights, as chompster can think fast enough to break out of the bloackade draw when required. [D] 5r2/8/8/8/1k3p2/p3pK1P/B5R1/8 w - -
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.