Author: Jeremiah Penery
Date: 08:18:37 03/16/02
Go up one level in this thread
On March 16, 2002 at 01:21:11, Slater Wold wrote: >On March 15, 2002 at 21:45:39, Scott Gasch wrote: > >>On March 15, 2002 at 19:27:32, Slater Wold wrote: >> >>> ext-> chk=17801393 cap=870361 pp=502965 1rep=1877028 mate=95160 >>>sing=16 >>^^^^^^^^ >>... >> >>>From my (very unstable) Crafty. As you can see, it's using singular >extensions. >> >>Is that all the singular extensions it saw in the whole search? My engine only >>does singular in check and it saw way more... What is your singular margin / >>how are you checking for singular moves? >> >>Monsoon says: >> 1u -3.95 00:00:00.01 81 PV= 1. g5 Rxc2+ [Q] 2. Qxc2 [Q] Bxc2 >> > [Q] 3. Kxc2 [Q] <-3.00> >> 1u -3.75 00:00:00.01 123 PV= 1. b3 Bxb3 [Q] <-2.00> >> 1u -2.44 00:00:00.02 177 PV= 1. c3 Bxd1 [Q] 2. Rxd1 [Q] <-3.00> >> 1u -0.92 00:00:00.04 235 PV= 1. Be4 <-1.00> >> 1. -0.92 00:00:00.06 278 PV= 1. Be4 <-1.00> >> 2. -1.11 00:00:00.12 698 PV= 1. Be4 Bc6 2. Bxc6+ [Q] Rxc6 [Q] >> > <-1.00> >> 3. -1.00 00:00:00.19 2767 PV= 1. Be4 d5 2. Bd3 <-1.00> >> 4. -1.32 00:00:00.27 7968 PV= 1. Be4 b3 2. c3 Bc6 3. Bxc6+ [Q] >> > Rxc6 [Q] 4. exd6 [Q] Rxd6 [Q] <-1.00> >> 5+ -0.66 00:00:00.53 45898 exd6!! >> 5. -0.56 00:00:00.65 56871 PV= 1. exd6 Bxd4 2. d7+ Bxd7 [anti] >> > 3. Qxd4 <-1.00> >> 6. -0.90 00:00:00.97 102128 PV= 1. exd6 Bxd4 2. d7+ Bxd7 [sing] >> > 3. Qxd4 Nf6 4. Bd5 b3 <-1.00> >> 7. -0.44 00:00:01.89 253619 PV= 1. exd6 Bxd4 2. d7+ Bxd7 [sing] >> > 3. Qxd4 Nf6 4. Re1 Kf8 5. Nd2 <-1.00> >> 8. -0.97 00:00:04.00 634343 PV= 1. exd6 Bxd4 2. d7+ Bxd7 [sing] >> > 3. Qxd4 Nf6 4. Re1 b3 5. Na3 bxc2 >> > 6. Nxc2 [Q] <-1.00> >> 9. -0.97 00:00:10.67 1836312 PV= 1. exd6 Bxd4 2. d7+ Bxd7 [TT: Qxd4 >> > Nf6 Rhe1 b3 Na3 bxc2 ...] >>10. -0.49 00:00:28.18 5031155 PV= 1. exd6 Bxd4 2. d7+ Bxd7 3. Qxd4 >> > Nf6 4. Rhe1 b3 5. c4 Kf8 6. Nc3 >> > <-1.00> >>11. -0.36 00:01:11.54 13433262 PV= 1. exd6 Bxd4 2. d7+ Bxd7 [sing] >> > 3. Qxd4 Nf6 4. g5 b3 5. c4 Bf5 6. >> > Na3 Qa1+ 7. Kd2 Rd8 8. Rxa1 [Q] >> > Rxd4+ [Q] 9. Kc3 [Q] <-1.00> >>? (of 38) : exd6 0x7ffffffe >>INPUT THREAD SAW: ? >>SEARCH SUSPENDED to read input (at 25353216 nodes) >>MOVE NOW COMMAND --> stop searching now >>move e5d6 >> >>HASHING: overall efficiencies (26.86% deep + 4.98% fast = 31.84%) >> HITS exact : 96 upper: 46959 lower: 1334704 unusable: 1056761 >> STORES exact: 164 upper: 464714 lower: 5263127 null: 3067120 >> TOTAL stores: 8795125 probes: 7659365 hits: (2056954 / 381566) 2438520 >> >>PAWNHASH: overall efficiency 96.98% (77869010 / 80290520) >> >>EVALFULL: overall efficiency 12.07% (2770454 / 22952015) >> >>FLIPRATE: 0 / 0 (0.00%) >> >>TREE: nodes: 25353218 (94.18% first move betas) >> TYPE termin: 101098 front: 11082590 qnode: 6597756 std.pat: 8276426 >> CUTS nullmv: 3067120 recogn: 0 beta: 2196005 1stmv: 2068145 >> hash: 1308638 extfut: 12755848 razor: 24984 %qnode: 26.02% >> q.fut: 6808718 q.see: 8294863 lazyeval: 6113650 >> >>EXTEN: total: 1576108 >> TYPE check: 2033515 threat: 24592 pawn: 0 recap: 91972 >> !check: 791629 1reply: 191673 mult: 17823 trans: 0 >> sex: 8161 zug: 1 >> ^^^^^^^^^^^ >> >>132.4 seconds @ 191546.9 nps. > >As Jeremiah said in his post, considering the penalty for SE I could not see >doing 30+ extensions for a 60 second eval. Toning down the extensions has given >me on average 1-2 ply. And those ply come in handy sometimes. In the search you originally posted, there were *17 million* check extensions. Since it's doing the work to determine singularity anyway, why not let it extend a few more singular moves? 16 just seems way too low to me. >I *probably* have toned them down too much, but like I said, this is a work in >progress. ;)
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.