Author: Vincent Diepeveen
Date: 05:40:45 08/14/02
Go up one level in this thread
On August 13, 2002 at 17:29:54, Uri Blass wrote: >On August 13, 2002 at 15:18:04, Vincent Diepeveen wrote: > >>On August 13, 2002 at 07:23:38, Uri Blass wrote: >> >>>My definition for a sacrifice or blunder >>>is a move that lose material based on >>>the depth that programs can see. >>> >>>The definition of losing material is based on >>>the material values 1,3,3,5,9. >> >>with all respect but your table is outdated in advance. >> >> a) 2 rooks are weaker than a queen in 99.9% of all cases >> the computer sees 2 rooks for a queen >> b) this table indicates that giving away 2 pieces for a rook+pawn >> is great >> c) giving away a piece for 3 pawns is great according to this >> table, especially if we know that one of the 3 pawns gets >> a passer >> >>So the definition is not correct! >> >>In diep queen = 11.5 pawns >> rook = 5.5 pawns >> piece = 3.6 pawns > >It means that Diep probably does not play the following move >of Quark against Crafty from leo's tournament >(note that movei also does not play >the move inspite of evaluating queen as less than 2 rooks because queen >is still evaluated as slightly more than rook piece and a pawn) >r3kb1r/pp1b1ppp/2B1pn2/8/Q1pq4/6P1/PP2PP1P/RNBR2K1 b kq - 0 9 >Quark played Qxd1+ and won the game. Qxd1 actually gives a lot of mobility to black. For the material white is obviously won there. DIEP in fact doubts between Qxd1 and bxc6. I consider personally Bxc6 better there. A material advantage when not proven to be incorrect must be considered correct for white. Bxc6 is a clear case of a draw. Qxd1 you are in fact lost with black but if white plays a few inaccurate moves the mobility of black will decide the game. I assume that is what happened in the blitz games Leo plays at his old PII duals. Qxd1 in this position is a matter of taste whether you speculate or do not speculate on mobility. It has nothing to do with material consideration. the material consideration is clear. put the material at the openings position at the openings squares and then go play. you will see the queen wins directly. The only consideration here: that black gets great mobility is the reason why qxd1 is interesting. I happen to know that Quark is built after some considerations i publicly mentionned it is nice to see that this has positive results! Latest experimental versions (no SE, and parallel speedup i fear 1.0 out of 2, as scaling at big computers isn't what scaling at 2 cpu pc is yet): the (2) means nowadays that it is on 2 processors: 00:00 200 0k 0 0 8 (2) 1 4.476 Bd7xc6 Rd1xd4 Bc6xa4 Rd4xc4 00:00 952 1k 0 0 162 (2) 2 -0.071 Bd7xc6 Qa4xc6 b7xc6 Rd1xd4 ++ d4-d1 00:00 1023 1k 0 0 174 (2) 2 0.057 Qd4xd1 Qa4xd1 Bd7xc6 00:00 2634 3k 0 0 685 (2) 3 0.057 Qd4xd1 Qa4xd1 Bd7xc6 00:00 6935 7k 0 0 2150 (2) 4 -0.077 Qd4xd1 Qa4xd1 Bd7xc6 Nb1-c3 ++ d7-c6 00:00 7924 8k 0 0 2615 (2) 4 -0.071 Bd7xc6 Qa4xc6 b7xc6 Rd1xd4 00:00 16256 16k 0 0 7153 (2) 5 -0.289 Bd7xc6 Qa4xc6 b7xc6 Rd1xd4 Bf8-c5 Rd4xc4 ++ d4-d1 00:00 17240 17k 0 0 8103 (2) 5 0.498 Qd4xd1 Qa4xd1 Bd7xc6 Nb1-c3 Ra8-d8 00:00 24545 25k 0 0 14973 (2) 6 0.001 Qd4xd1 Qa4xd1 Bd7xc6 Bc1-g5 Ra8-d8 Nb1-d 2 00:00 38379 38k 0 0 26482 (2) 7 0.385 Qd4xd1 Qa4xd1 Bd7xc6 Bc1-e3 Ra8-d8 Nb1-d 2 Bf8-b4 00:00 57561 58k 0 0 54683 (2) 8 0.036 Qd4xd1 Qa4xd1 Bd7xc6 Bc1-e3 Nf6-g4 Be3-g 5 Bf8-c5 e2-e3 00:01 88526 89k 0 0 154922 (2) 9 0.108 Qd4xd1 Qa4xd1 Bd7xc6 Nb1-d2 b7-b5 a2-a4 a7-a6 a4xb5 a6xb5 Ra1xa8 Bc6xa8 00:02 103354 103k 0 0 295593 (2) 10 0.011 Qd4xd1 Qa4xd1 Bd7xc6 Nb1-d2 b7-b5 a2-a 4 a7-a6 b2-b3 c4-c3 Nd2-b1 ++ d7-c6 00:07 119168 119k 0 0 897342 (2) 10 0.056 Bd7xc6 Qa4xc6 b7xc6 Rd1xd4 Ra8-d8 Bc1- e3 Nf6-d5 Be3-d2 Rd8-b8 Bd2-c1 00:08 129354 129k 0 0 1131855 (2) 11 -0.167 Bd7xc6 Qa4xc6 b7xc6 Rd1xd4 c6-c5 Rd4 xc4 O-O-O Nb1-c3 Rd8-d4 b2-b3 Kc8-d7 ++ d4-d1 00:10 137495 137k 0 0 1443698 (2) 11 0.108 Qd4xd1 Qa4xd1 Bd7xc6 Nb1-d2 b7-b5 a2- a4 a7-a6 a4xb5 a6xb5 Ra1xa8 Bc6xa8 00:16 148776 149k 0 0 2408686 (2) 12 -0.290 Qd4xd1 Qa4xd1 Bd7xc6 Nb1-d2 c4-c3 b2 xc3 O-O-O Qd1-b3 Kc8-b8 Nd2-f3 Nf6-e4 Bc1-g5 Ne4xg5 Nf3xg5 ++ d7-c6 00:22 155259 155k 0 0 3549226 (2) 12 -0.027 Bd7xc6 Qa4xc6 b7xc6 Rd1xd4 Ra8-d8 Rd 4xc4 Rd8-d1 Kg1-g2 Ke8-d7 Nb1-c3 Rd1-e1 Ra1-b1 Bf8-d6 00:30 164392 164k 0 0 5078096 (2) 13 -0.155 Bd7xc6 Qa4xc6 b7xc6 Rd1xd4 Bf8-b4 Rd 4xc4 O-O-O Kg1-g2 Rd8-d1 a2-a3 Bb4-e1 b2-b4 Kc8-c7 Ra1-a2 00:49 179992 180k 0 0 8970846 (2) 14 0.015 Bd7xc6 Qa4xc6 b7xc6 Rd1xd4 Ra8-d8 Rd4 xc4 Rd8-d1 Kg1-g2 Rd1-e1 Rc4xc6 Ke8-d7 Rc6-c2 Bf8-d6 Kg2-f3 Re1-h1 01:40 185493 185k 0 0 18643941 (2) 15 -0.166 Bd7xc6 Qa4xc6 b7xc6 Rd1xd4 Bf8-e7 K g1-g2 e6-e5 Rd4xc4 Ke8-d7 Nb1-d2 Nf6-d5 Nd2-f3 f7-f6 Rc4-a4 Be7-c5 04:38 179845 180k 0 0 50153446 (2) 16 -0.049 Bd7xc6 Qa4xc6 b7xc6 Rd1xd4 Ra8-d8 R d4xc4 Rd8-d1 Kg1-g2 Rd1-e1 Rc4xc6 Ke8-d7 Rc6-c2 Bf8-b4 Kg2-f3 Bb4-d6 h2-h3 Re1-h 1 Now the world champs version (same evaluation other search with SE): it goes clearly for bxc6: 00:00 0k 0 0 4 (0) 1 -8.101 b7xc6 Rd1xd4 00:00 0k 0 0 8 (0) 1 4.476 Bd7xc6 Rd1xd4 Bc6xa4 Rd4xc4 00:00 0k 0 0 137 (0) 2 -0.071 Bd7xc6 Qa4xc6 b7xc6 Rd1xd4 ++ d4-d1 00:00 0k 0 0 149 (0) 2 0.057 Qd4xd1 Qa4xd1 Bd7xc6 00:00 1k 0 0 660 (0) 3 0.057 Qd4xd1 Qa4xd1 Bd7xc6 00:00 2k 0 0 2125 (0) 4 -0.077 Qd4xd1 Qa4xd1 Bd7xc6 Nb1-c3 ++ d7-c6 00:00 3k 0 0 2590 (0) 4 -0.071 Bd7xc6 Qa4xc6 b7xc6 Rd1xd4 00:00 8k 0 0 8127 (0) 5 -0.289 Bd7xc6 Qa4xc6 b7xc6 Rd1xd4 Bf8-c5 Rd4xc4 ++ d4-d1 00:00 9k 0 0 9190 (75) 5 0.498 Qd4xd1 Qa4xd1 Bd7xc6 Nb1-c3 Ra8-d8 00:01 20k 0 0 20639 (1803) 6 0.385 Qd4xd1 Qa4xd1 Bd7xc6 Bc1-e3 Ra8-d8 Nb1-d2 Bf 8-b4 00:01 18k 0 0 31641 (1841) 7 0.385 Qd4xd1 Qa4xd1 Bd7xc6 Bc1-e3 Ra8-d8 Nb1-d2 Bf 8-b4 00:02 26k 0 0 70942 (3719) 8 0.036 Qd4xd1 Qa4xd1 Bd7xc6 Bc1-e3 a7-a5 Be3-b6 Nf6 -d5 Bb6-d4 ++ d7-c6 00:02 32k 0 0 94682 (4505) 8 0.056 Bd7xc6 Qa4xc6 b7xc6 Rd1xd4 Ra8-d8 Bc1-e3 Rd8 xd4 Be3xd4 c6-c5 Bd4-e5 00:04 44k 0 0 195204 (7626) 9 0.056 Bd7xc6 Qa4xc6 b7xc6 Rd1xd4 Ra8-d8 Bc1-e3 Rd 8xd4 Be3xd4 c6-c5 Bd4-e5 00:07 63k 0 0 486073 (37545) 10 0.056 Bd7xc6 Qa4xc6 b7xc6 Rd1xd4 Ra8-d8 Bc1-e3 Rd8xd4 Be3xd4 c6-c5 Bd4-e5 00:13 94k 0 0 1295005 (80575) 11 -0.167 Bd7xc6 Qa4xc6 b7xc6 Rd1xd4 c6-c5 Rd4xc4 O-O-O Nb1-c3 Rd8-d4 b2-b3 Kc8-d7 00:29 111k 0 0 3285329 (232136) 12 -0.027 Bd7xc6 Qa4xc6 b7xc6 Rd1xd4 Ra8-d8 Rd4x c4 Rd8-d1 Kg1-g2 Ke8-d7 Nb1-c3 Rd1-e1 Ra1-b1 Bf8-d6 01:04 129k 0 0 8397664 (621313) 13 -0.153 Bd7xc6 Qa4xc6 b7xc6 Rd1xd4 Bf8-b4 b2-b 3 a7-a5 Rd4xc4 Ke8-d7 Bc1-d2 Rh8-b8 Nb1-c3 e6-e5 02:37 143k 0 0 22671813 (2419347) 14 0.015 Bd7xc6 Qa4xc6 b7xc6 Rd1xd4 Ra8-d8 Rd4 xc4 Rd8-d1 Kg1-g2 Rd1-e1 Rc4xc6 Ke8-d7 Rc6-c2 Bf8-d6 Kg2-f3 Re1-h1 In the leo blitz games of course diep would go qxd1 too there at the blitz level with the new version. >By your definition of sacrifice Qxd1+ is a sacrifice. That is very obvious. you give away material for good mobility. >By my definition it is only a trade of material and >nobody sacrificed material. That is a wrong definition, because white wins material after qxd1. >I can imagine that I am going to have more mistakes >if I increase the evaluation of queen to more than 2 rooks. You could implement mobility and put it high. Qxd1 is not a winning move here. it is a speculative move. Just like in some mainlines in sicilian some GMs give up a queen for sometimes only rook + bishop. Good example is some dragon lines i play myself with white. Even Anand manages to get it a draw, but i would love to see DIEP - Anand in that position. Anand would get toasted of course. Computers are way better with more material. Another consideration is that you should print the number of times that movei is having queen+pawn versus 2 rooks, or even queen versus 2 rooks. You will see that the queen directly after that picks up material in 95% of the cases. A queen grabs material simply easier than rooks. The first thing i tell youth always is that the beginners mistake to put queen at 9 and rooks at 5 is very dumb. The old books lie here. The important other consideration is knowledge for a chesprogram. Suppose you have n patterns of knowledge for rooks and n patterns for queens. Let's just take mobmility. then obviously the 2 rooks combined get more bonus than the queen. That means that getting 2 rooks is directly like +13.0 with just mobility and that getting a good queen is +10.5 with good mobility if we assume 5+5 versus 9. You miss 2.5 pawns then for the queen! >Reducing the evaluation of queen can also cause mistake so >I am not going to change it in the near future. You forgot to mention at which values the stuff is set at now. Also you do not have mobility i already know this for sure now. It is interesting to know what the values from quark are. I bet they are closer to DIEP than they are to Movei!! >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.