Author: Jeremiah Penery
Date: 17:52:15 12/14/03
Go up one level in this thread
On December 14, 2003 at 17:32:12, Robert Hyatt wrote: >On December 14, 2003 at 02:54:32, Jeremiah Penery wrote: > >>On December 13, 2003 at 13:07:00, Robert Hyatt wrote: >> >>>On December 12, 2003 at 19:32:45, Slater Wold wrote: >>> >>>>On December 12, 2003 at 15:13:40, Will Singleton wrote: >>>> >>>>>On December 12, 2003 at 13:49:35, Robert Hyatt wrote: >>>>> >>>>>>On December 12, 2003 at 13:40:07, Will Singleton wrote: >>>>>> >>>>>>>Here's a position from a recent game, this one gives Amateur fits. >>>>>>>How long to see black is busted? >>>>>>> >>>>>>>[d]4rr2/1b3p1k/pp5p/2b1pQ2/q7/P2R2B1/1PP3PP/2KR4 b - - >>>>>> >>>>>> >>>>>>For Crafty, the bottom begins to fall out at 6 seconds or so, where >>>>>>it begins to see it is not ahead. By 60 seconds it has failed low >>>>>>to white a pawn up and climbing: >>>>>> >>>>>> 12-> 4.84 -0.69 1. ... Kg8 2. Bxe5 Bf2 3. Rd4 Rxe5 >>>>>> 4. Qxe5 Re8 5. Rd8 Bc6 6. Rxe8+ Bxe8 >>>>>> 7. Rd6 Qb5 8. Qf4 Qg5 <HT> >>>>>> 13 6.66 -- 1. ... Kg8 >>>>>> 13 15.28 0.31 1. ... Kg8 2. Bxe5 Bf2 3. Bf6 Re2 4. >>>>>> Rd8 Bc5 5. R8d4 Rxc2+ 6. Qxc2 Qxc2+ >>>>>> 7. Kxc2 Bxd4 8. Bxd4 Be4+ 9. Kb3 Rd8 >>>>>> 13-> 17.91 0.31 1. ... Kg8 2. Bxe5 Bf2 3. Bf6 Re2 4. >>>>>> Rd8 Bc5 5. R8d4 Rxc2+ 6. Qxc2 Qxc2+ >>>>>> 7. Kxc2 Bxd4 8. Bxd4 Be4+ 9. Kb3 Rd8 >>>>>> 14 35.70 0.00 1. ... Kg8 2. Bxe5 Bf2 3. Bf6 Re2 4. >>>>>> Rd8 Bh4 5. Rxf8+ Kxf8 6. Rd8+ Re8 7. >>>>>> Rxe8+ Qxe8 8. Bxh4 Qe3+ 9. Kb1 Qg1+ >>>>>> 10. Ka2 Qxg2 11. Qf4 Qxc2 12. Qxh6+ >>>>>> 14-> 40.01 0.00 1. ... Kg8 2. Bxe5 Bf2 3. Bf6 Re2 4. >>>>>> Rd8 Bh4 5. Rxf8+ Kxf8 6. Rd8+ Re8 7. >>>>>> Rxe8+ Qxe8 8. Bxh4 Qe3+ 9. Kb1 Qg1+ >>>>>> 10. Ka2 Qxg2 11. Qf4 Qxc2 12. Qxh6+ >>>>>> 15 51.44 -- 1. ... Kg8 >>>>>> 15 1:30 0.98 1. ... Kg8 2. Bxe5 Bf2 3. Rd4 Rxe5 >>>>>> 4. Qxe5 Re8 5. Rg4+ Qxg4 6. Qxe8+ Kg7 >>>>>> 7. Qe5+ f6 8. Qc7+ Kg6 9. Rd7 Qg5+ >>>>>> 10. Kb1 Bxg2 11. Rg7+ Kf5 12. Rxg5+ >>>>>> hxg5 >>>>> >>>>>That's only half the story. Let it think another half hour. >>>>> >>>>>Will >>>> >>>>White(1): setboard 4rr2/1b3p1k/pp5p/2b1pQ2/q7/P2R2B1/1PP3PP/2KR4 b - - >>>>Black(1): st 9999999 >>>>search time set to 9999999.00. >>>>Black(1): go >>>> clearing hash tables >>>> time surplus 0.00 time limit 166666:39 (166666:39) >>>> depth time score variation (1) >>>>starting thread 1 >>>> 7 0.14 -1.34 1. ... Kg8 2. Bxe5 Bf2 3. Kb1 Rxe5 >>>> 4. Qxe5 Bxg2 >>>> 7-> 0.14 -1.34 1. ... Kg8 2. Bxe5 Bf2 3. Kb1 Rxe5 >>>> 4. Qxe5 Bxg2 >>>> 8 0.23 -- 1. ... Kg8 >>>> 8 0.34 -0.89 1. ... Kg8 2. Bxe5 Bf2 3. Rh3 Qe4 4. >>>> Qxf2 Rxe5 5. Rxh6 Qxg2 6. Qxb6 >>>> 8-> 0.36 -0.89 1. ... Kg8 2. Bxe5 Bf2 3. Rh3 Qe4 4. >>>> Qxf2 Rxe5 5. Rxh6 Qxg2 6. Qxb6 >>>> 9 0.95 -1.12 1. ... Kg8 2. Bxe5 Bf2 3. Qf6 Rxe5 >>>> 4. Qxe5 Re8 5. Rd8 Rxd8 6. Rxd8+ >>>> 9-> 0.98 -1.12 1. ... Kg8 2. Bxe5 Bf2 3. Qf6 Rxe5 >>>> 4. Qxe5 Re8 5. Rd8 Rxd8 6. Rxd8+ >>>> 10 1.83 -0.97 1. ... Kg8 2. Bxe5 Bf2 3. Qf6 Rxe5 >>>> 4. Qxe5 Re8 5. Qc7 Bxg2 6. Rd7 >>>> 10-> 1.90 -0.97 1. ... Kg8 2. Bxe5 Bf2 3. Qf6 Rxe5 >>>> 4. Qxe5 Re8 5. Qc7 Bxg2 6. Rd7 >>>> 11 7.12 -0.84 1. ... Kg8 2. Bxe5 Bf2 3. Qf6 Rxe5 >>>> 4. Qxe5 Qg4 5. g3 Qg5+ 6. Qxg5+ hxg5 >>>> 7. Rf1 Bc5 8. Rd7 >>>> 11-> 7.72 -0.84 1. ... Kg8 2. Bxe5 Bf2 3. Qf6 Rxe5 >>>> 4. Qxe5 Qg4 5. g3 Qg5+ 6. Qxg5+ hxg5 >>>> 7. Rf1 Bc5 8. Rd7 >>>> 12 13.15 -- 1. ... Kg8 >>>> 12 18.67 -0.43 1. ... Kg8 2. Bxe5 Bf2 3. Rd4 Rxe5 >>>> 4. Qxe5 Re8 5. Rd8 Bc6 6. Rxe8+ Bxe8 >>>> 7. Rd8 Kf8 8. Kd2 >>>> 12-> 20.06 -0.43 1. ... Kg8 2. Bxe5 Bf2 3. Rd4 Rxe5 >>>> 4. Qxe5 Re8 5. Rd8 Bc6 6. Rxe8+ Bxe8 >>>> 7. Rd8 Kf8 8. Kd2 >>>> 13 28.70 -- 1. ... Kg8 >>>> 13 1:02 0.31 1. ... Kg8 2. Bxe5 Bf2 3. Bf6 Re2 4. >>>> Rd8 Bc5 5. R8d4 Rxc2+ 6. Qxc2 Qxc2+ >>>> 7. Kxc2 Bxd4 8. Bxd4 Be4+ 9. Kb3 Rd8 >>>> 13-> 1:08 0.31 1. ... Kg8 2. Bxe5 Bf2 3. Bf6 Re2 4. >>>> Rd8 Bc5 5. R8d4 Rxc2+ 6. Qxc2 Qxc2+ >>>> 7. Kxc2 Bxd4 8. Bxd4 Be4+ 9. Kb3 Rd8 >>>> 14 2:21 0.00 1. ... Kg8 2. Bxe5 Bf2 3. Bf6 Re2 4. >>>> Rd8 Bh4 5. Rxf8+ Kxf8 6. Rd8+ Re8 7. >>>> Rxe8+ Qxe8 8. Bxh4 Qe3+ 9. Kb1 Qg1+ >>>> 10. Ka2 Qxg2 11. Qf4 Qxc2 12. Qxh6+ >>>> 14-> 2:34 0.00 1. ... Kg8 2. Bxe5 Bf2 3. Bf6 Re2 4. >>>> Rd8 Bh4 5. Rxf8+ Kxf8 6. Rd8+ Re8 7. >>>> Rxe8+ Qxe8 8. Bxh4 Qe3+ 9. Kb1 Qg1+ >>>> 10. Ka2 Qxg2 11. Qf4 Qxc2 12. Qxh6+ >>>> 15 3:35 -- 1. ... Kg8 >>>> 15 6:51 0.98 1. ... Kg8 2. Bxe5 Bf2 3. Rd4 Rxe5 >>>> 4. Qxe5 Re8 5. Rg4+ Qxg4 6. Qxe8+ Kg7 >>>> 7. Qe5+ f6 8. Qc7+ Kg6 9. Rd7 Qg5+ >>>> 10. Kb1 Bxg2 11. Rg7+ Kf5 12. Rxg5+ >>>> hxg5 >>>> 15-> 31:42 0.98 1. ... Kg8 2. Bxe5 Bf2 3. Rd4 Rxe5 >>>> 4. Qxe5 Re8 5. Rg4+ Qxg4 6. Qxe8+ Kg7 >>>> 7. Qe5+ f6 8. Qc7+ Kg6 9. Rd7 Qg5+ >>>> 10. Kb1 Bxg2 11. Rg7+ Kf5 12. Rxg5+ >>>> hxg5 (s=2) >>>> 16 32:56 -- 1. ... Kg8 >>>>? 16 193:15 1/3* 1. ... Kg8 >>>>Black(0): ? >>>> time=193:18 cpu=197% mat=-3 n=20834682079 fh=99% nps=1796k >>>> ext-> chk=-1921721336 cap=10974209 pp=1617516 1rep=423412478 mate= >>>>1390210 >>>> predicted=0 nodes=20834682079 evals=670369456 >>>> endgame tablebase-> probes=0 hits=0 >>>> SMP-> split=3793 stop=417 data=12/4096 cpu=382:29 elap=193:18 >>> >>> >>>I have seen worse. What happens is that the aspiration window initially cuts >>>off all mate scores. But once it fails low, it can't just find any old mate >>>and cut off analyzing that line. Now it has to follow _all_ mates and that >>>kills performance. This also happens in endgames where you can potentially >>>promote a pawn, but it always gets lost, so you can search to great depth. >>>And eventually you find that you can promote it there, but now all those lines >>>you got cutoffs on before (where the pawn promoted, but it wasn't forced) now >>>explode and while the depth 25 search took 10 seconds, the depth 26 search might >>>not take 10 days... >>> >>>It's a known problem with no known solution. >>> >>>Even not using aspiration search won't solve it. >> >>You can not lower the window to -INF, but rather on the initial fail low move it >>down -1, then next do -3, and so on. Or just do it for fail-high and that >>should affect fail-low also (since a fail-low on one side is a fail-high for the >>other). > > >Yes you can. We documented that in "using time wisely" and "using time wisely, >revisited" which were published in the JICCA. But it only solves it for >some cases, not the pawn promotion case. Because you _must_ drop it low enough >to let the promotion stand, and that is low enough to let _all_ promotions >stand, which blows the search right out of reality. Dropping it low enough to let promotions stand does not have to be the same as dropping it low enough to let it find mates everywhere. You can still get a huge improvement over dropping immediately to -INF.
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.