Author: Hans Havermann
Date: 11:31:56 04/06/99
Go up one level in this thread
On April 05, 1999 at 20:30:22, Dann Corbit wrote: >>After some 30 billion nodes, here is MacChess' winning line for white (after >>1.Re1 fxg4) at the same depth: >> 2. Rxe6 gxh3 >> 3. Rg6 Qxd4 >> 4. Rxd6 h2+ >> 5. Kh1 Qxf2 >> 6. Bxh6 Qxb2 >> 7. Bxg7+ Qxg7 >> 8. Rf1+ Ke7 >> 9. Qxg7+ Kxd6 >>10. Qg6+ Kc5 >>11. Kxh2 >What is the centipawn eval at that point? r1r2k2/1p4pQ/1n1bp2p/p2n1p2/P2P2Nq/1B5P/1P3PP1/R1B1R1K1 b - -, depth 15 MacChess' eval *number* at that level is -82: obviously *not* centipawns. To put that number into some context, here (again) is MacChess' analysis of the previous position in this study: r1r2k2/1p4pQ/1n1bp2p/p2n1p2/P2P2Nq/1B5P/1P3PP1/R1BR2K1 w - - 07/01|00:00:05| 1192024| +3 | Ne3 Qf6 Bd2 Nf4 Kf1 Bc7 Bc3 08/01|00:00:17| 3571273| -3 | Ne3 Qf6 Nxd5 Nxd5 Bxd5 exd5 Bd2 b6 Rdc1 Bc7 09/01|00:00:55| 11100158| -4 | Ne3 Qf6 Ra2 Ke7 Nxd5+ Nxd5 Bxd5 exd5 Re1+ Kd7 Be3 b6 09/22|00:01:57| 27029401| +27 | Nxh6 gxh6 Bxh6+ Ke8 Qg6+ Kd7 Bg5 Qe4 Re1 Qxd4 Qxe6+ Kc6 Qxf5 Qxb2 10/01|00:03:18| 51062439| +99 | Nxh6 Qf6 Bxd5 Nxd5 Qg8+ Ke7 Nxf5+ Qxf5 Qxg7+ Qf7 Bg5+ Ke8 Qxf7+ Kxf7 g4 Be7 Bxe7 Kxe7 11/01|00:06:39| 101499073| +99 | Nxh6 Qf6 Bxd5 Nxd5 Qg8+ Ke7 Nxf5+ Qxf5 Qxg7+ Qf7 Bg5+ Ke8 Qxf7+ Kxf7 g4 Bf4 Bxf4 Nxf4 Ra3 12/01|00:13:45| 220445858| +96 | Nxh6 Qf6 Bxd5 Nxd5 Qg8+ Ke7 Nxf5+ Qxf5 Qxg7+ Qf7 Bg5+ Ke8 Qxf7+ Kxf7 g4 Bf4 Bxf4 Nxf4 Kh2 b6 13/01|00:38:17| 632391553| +98 | Nxh6 Qf6 Bxd5 Nxd5 Qg8+ Ke7 Nxf5+ Qxf5 Qxg7+ Qf7 Bg5+ Ke8 Qxf7+ Kxf7 g4 Bf4 Bxf4 Nxf4 Ra3 Ne2+ Kf1 Nf4 As Will Singleton pointed out, MacChess probably "over-values it's king-side pawns" to give white the +high-90's for Nxh6. Also, with Re1 doing no better than +82 (at a much deeper level), we see why MacChess prefers Nxh6 to Re1. I'd be interested to know how *your* chess programs evaluate Nxh6.
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.