Author: Uri Blass
Date: 05:07:35 08/16/05
Go up one level in this thread
On August 16, 2005 at 07:24:35, Robert Hyatt wrote: >On August 16, 2005 at 01:24:01, Uri Blass wrote: > >>On August 16, 2005 at 00:46:35, Robert Hyatt wrote: >> >>>On August 15, 2005 at 23:53:14, Uri Blass wrote: >>> >>>>On August 15, 2005 at 21:41:22, Robert Hyatt wrote: >>>> >>>>>On August 15, 2005 at 21:21:46, Mark Young wrote: >>>>> >>>>>>On August 15, 2005 at 20:54:51, Robert Hyatt wrote: >>>>>> >>>>>>>On August 15, 2005 at 19:35:25, Mark Young wrote: >>>>>>> >>>>>>>>On August 15, 2005 at 16:55:58, Thomas Mayer wrote: >>>>>>>> >>>>>>>>>On August 15, 2005 at 16:41:40, Dann Corbit wrote: >>>>>>>>> >>>>>>>>>>On August 15, 2005 at 16:21:31, Theo van der Storm wrote: >>>>>>>>>> >>>>>>>>>>>C:\DOSEXE\PGN.EXE -d -x wccc13.pgn >>>>>>>>>>> # Name 1 2 3 4 P BU SB G >>>>>>>>>>> 1 Crafty 3b= 4b1 10w1 9b1 3.5 8.0 6.50 4 >>>>>>>>>>> 2 Zappa 9w1 6b1 5w1 11b= 3.5 6.0 5.75 4 >>>>>>>>>>> 3 Shredder 1w= 10b1 9w1 6b= 3.0 7.5 4.75 4 >>>>>>>>>>> 4 Deep Junior 11b1 1w0 7b1 10w1 3.0 7.0 3.50 4 >>>>>>>>>>> 5 Fruit 7b1 8w1 2b0 12w= 2.5 7.5 3.75 4 >>>>>>>>>>> 6 Deep Sjeng 8b= 2w0 12b1 3w= 2.0 8.5 2.75 4 >>>>>>>>>>> 7 Jonny 5w0 11b1 4w0 8w1 2.0 7.5 2.00 4 >>>>>>>>>>> 8 The Baron 6w= 5b0 11w1 7b0 1.5 7.0 1.50 4 >>>>>>>>>>> 9 The Crazy Bishop 2b0 12w1 3b0 1w0 1.0 10.5 0.50 4 >>>>>>>>>>> 10 Diep 12b1 3w0 1b0 4b0 1.0 10.0 0.50 4 >>>>>>>>>>> 11 Fute_MT 4w0 7w0 8b0 2w= 0.5 10.0 1.75 4 >>>>>>>>>>> 12 IsiChess MMX 10w0 9b0 6w0 5b= 0.5 6.5 1.25 4 >>>>>>>>>>> >>>>>>>>>>> >>>>>>>>>>> # Name 1 2 3 4 5 6 7 8 9 10 11 12 P BU SB >>>>>>>>>>> 1 Crafty X . ½ 1 . . . . 1 1 . . 3.5 8.0 6.50 >>>>>>>>>>> 2 Zappa . X . . 1 1 . . 1 . ½ . 3.5 6.0 5.75 >>>>>>>>>>> 3 Shredder ½ . X . . ½ . . 1 1 . . 3.0 7.5 4.75 >>>>>>>>>>> 4 Deep Junior 0 . . X . . 1 . . 1 1 . 3.0 7.0 3.50 >>>>>>>>>>> 5 Fruit . 0 . . X . 1 1 . . . ½ 2.5 7.5 3.75 >>>>>>>>>>> 6 Deep Sjeng . 0 ½ . . X . ½ . . . 1 2.0 8.5 2.75 >>>>>>>>>>> 7 Jonny . . . 0 0 . X 1 . . 1 . 2.0 7.5 2.00 >>>>>>>>>>> 8 The Baron . . . . 0 ½ 0 X . . 1 . 1.5 7.0 1.50 >>>>>>>>>>> 9 The Crazy Bishop 0 0 0 . . . . . X . . 1 1.0 10.5 0.50 >>>>>>>>>>> 10 Diep 0 . 0 0 . . . . . X . 1 1.0 10.0 0.50 >>>>>>>>>>> 11 Fute_MT . ½ . 0 . . 0 0 . . X . 0.5 10.0 1.75 >>>>>>>>>>> 12 IsiChess MMX . . . . ½ 0 . . 0 0 . X 0.5 6.5 1.25 >>>>>>>>>> >>>>>>>>>>Is there a hardware listing somewhere yet? >>>>>>>>>> >>>>>>>>>>I am guessing that Fruit is the only one in the top six that is running on a >>>>>>>>>>single CPU. >>>>>>>>>> >>>>>>>>>>Against Crafty (for instance) there is about 6-7x speed difference (as a rough >>>>>>>>>>estimate). I am curious about the other hardware. >>>>>>>>> >>>>>>>>>Hi Dann, >>>>>>>>> >>>>>>>>>all I know currently is, that Diep & Crafty play on 8-way boxes, Shredder, >>>>>>>>>Junior & Zappa on 4-way, Deep Sjeng on Dual an the rest on single... For Jonny >>>>>>>>>it should be Athlon 64 with 2.6 GHz internally, Fruit is 2.4 GHz... >>>>>>>>>I think Gerd has his own Athlon 64 with him which is afaik 2.2 GHz -> The rest >>>>>>>>>is probably P4 3.0 GHz >>>>>>>>> >>>>>>>>>Greets, Thomas >>>>>>>> >>>>>>>>Crafty has really shown me some very good chess in this tournament. Crafty has >>>>>>>>shown some holes in its play before. I don't see them so far in this tournament. >>>>>>>>This has to be more then just 8 cpu's and 16 million nodes a sec helping >>>>>>>>Crafty's play. For this to happen Crafty must has also stepped up a notch on the >>>>>>>>programming side. Speed is great, but speed can't always fix holes in a programs >>>>>>>>play. As shown by Diep for example who also plays on a 8 way. Bob looks like a >>>>>>>>big threat for any program in this touranment, and it is no fluke that Crafty is >>>>>>>>in first place. Bob has to be considered a favorite to win this tournament. I >>>>>>>>would not have said that 2 days ago. >>>>>>> >>>>>>> >>>>>>>I still wouldn't say it myself. :) >>>>>> >>>>>>You may not want to say it, but you see what I see...I bet. :) It is not easy to >>>>>>bring everything together when playing in WCCC. You may have done it this year. >>>>>>All the best Bob. Good Luck. >>>>> >>>>> >>>>>There are a couple of issues at least. Yes, the thing is very fast, which tends >>>>>to mean that everyone else is searching a perfect sub-tree of what crafty is >>>>>searching. That is a serious problem for them. If you ever play against >>>>>someone with much faster hardware, you'll see this effect. You play the best >>>>>move you can find, yet your eval continually drops. This simply means you are >>>>>getting out-searched. TCB played well today, it just got out-searched pretty >>>>>badly. The other issue is the "luck factor". A good program can have bad luck >>>>>and lose against anybody in Iceland. A bad program can have a bit of luck in a >>>>>single game, and be capable of beating anyone playing there. I just have to >>>>>hope that I'm not the recipient of that bad luck. >>>>> >>>>>Also, there are some very strong opponents left. Anthony's program (zappa) is >>>>>certainly highly dangerous. Don't know much about Fruit, but from all the >>>>>comments here, it is highly capable of winning games. It is going to have its >>>>>hands full against Crafty's search speed, but I can clearly remember both Cray >>>>>Blitz and Deep Thought losing a game here and there against _far_ slower >>>>>programs that also had less chess knowledge. >>>>> >>>>>In the shredder game, Crafty played well. It pushed hard and pushed into a >>>>>position where it had an eval of almost -2.0, which means it was simply >>>>>out-searching shredder for the most part (ignoring shredder's selective search >>>>>stuff of course). >>>> >>>>I am not sure if better position means outsearching the opponent and it can be >>>>simply result of better evaluation. >>>> >>>>better hardware does not mean automatically outsearching the opponent and it may >>>>be interesting to see if Crafty outsearches fruit in middle game positions after >>>>the tournament. >>> >>>I can test any positions you want. I don't recall a search less than 15 plies >>>deep so far during the WCCC, and most are 16 and up... >> >>I think that most of fruit searchs are 16 and up(of course depths are not >>equivalent and I think that fruit does both more pruning and both extensions >>than Crafty) >>> >>> >>>> >>>> >>>> It then was a little optimistic in its evaluation with the >>>>>opponent's two connected passers, something I had seen a couple of times in test >>>>>games against shredder on ICC. Was afraid to try to fix that so close to the >>>>>tournament, because changes to the eval can have effects far beyond just the >>>>>things being looked at. >>>>> >>>>>The deep junior game saw DJ self-destruct with the BN for RP sacrifice. No idea >>>>>what caused that, other than probably king-safety turned up too high to reward >>>>>aggressive play. >>>> >>>>Maybe good preperation of peter berger to go to positions that Junior does not >>>>know how to play? >>> >>>Impossible. How to get access to their book to know where they will go? How to >>>get access to their program to know how it will react and what it handles >>>poorly? >> >>Impossible to get access to their program but guessing that they may have the >>same weakness as Junior9 is not a bad guess. >> >>About their book I agree that it is impossible but it is still possible to find >>some lines that they do not play well and if you are lucky these lines are in >>their book(there is a significant probability that line in their public book >>will be also in the tournament book). >> >>first 6 moves of Junior are in the public book of Junior and later Junior out of >>the public book. >> >>> >>>We didn't have time for that kind of experimentation, even if we had had access >>>to the current DJ program and hardware... >> >>I agree that it is not easy to do that kind of preperation but in theory it is >>possible. >> >>> >>> >>>> >>>>Note that I am not sure that the sacrifice was so bad because fruit also likes >>>>it but does not continue like Junior but go to trade pieces. >>> >>>It is bad. Period. Just based on general principles... >> >>This is also what I thought but I am not sure about it >> >>I expect in position that is good for black that the evaluation will improve >>when the program goes deeper >> >>I do not see it. >>In position that is forced some plies after the sacrifice >>Crafty suggest 17...Be6 and the evaluation for black does not improve when it >>searches deeper. >> > >Here is what I see in the log file: > >after Qxf7, score is -.70 You probably means after 14...Rxf7 , after white plays Nf3 the next move, score is -1.0, I think that 15.Nf3 was a mistake 15.Bxf7+ was better for white and after the forced line for black 15...Qxf7 16.Qxf7+ Kxf7 17.Nc4 Crafty will see smaller advantage for black. 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.