Author: Vincent Diepeveen
Date: 14:27:14 11/05/01
Go up one level in this thread
On November 05, 2001 at 14:39:43, Robert Hyatt wrote: >On November 05, 2001 at 07:56:42, Vincent Diepeveen wrote: > >>On November 04, 2001 at 14:52:47, Gian-Carlo Pascutto wrote: >> >>not restarting your program is allowed, i fail to find that >>anywhere in the tournament rules. >> >>if you use winboard, face reality then. You'll lose on flag >>before it's even clear that your engine blows the position. >> >>Thomas Mayer has found some kind of nerd solution to not lose on time. >>Well either write your own interface, hack winboard like Bruce Moreland >>did for the world champs, find a nerd solution like Thomas did, >>or don't use winboard! >> >>I never used it in the tournaments i played when diep was a console >>application only. Note that initially it wasn't even allowed to make >>time corrections. Now it's allowed to synchronize time, which is a good >>thing. Next tournament you want to restart your engine? >> >>O wait here it goes cool. Bob's quad doesn't have my own EGTBs. Nalimov >>code doesn't compile on gcc 3.0.2, so the reason i didn't run 4x 700 last >>weekend but my own 800Mhz machine (note a dual AMD, as the mainboard >>stopped working suddenly) was because i have 3.0.2 installed here at home. > > >Why not compile on my quad. I use 2.95.2 still and it compiles all the >EGTB stuff just fine... i didn't have time for that. i already had uploaded with superhumane effort a new book to the quad and the executable. The ISDN line was dead slow. like 500 bytes a second or so (connected to wanadoo.nl may i ask you, this already gives for 2 years problem after problem, whereas xs4all.nl i never have had problems with, perhaps Theo next year uses other provider?). Despite that isdn is on paper like 16KB/s or something? Anyway i had slept for 3 hours and was not very clear otherwise i would have thought of that solution. > >> >>So at Bobs machine i couldn't use EGTBs at all, apart from the fact that >>i could only allocate 32MB ram there. > >shmmax was set to 400000000. In fact, I found at least one 320000000mb shared >segment left lying around after you finished. Perhaps you couldn't get another >until you released that one. (ipcrm) Right, we got disconnected from the internet and DIEP already had lost loads of time because of this, so we continued game on the dual 800. A major mistake perhaps, because i lost the game i later analyzed because of that. However my operator was very happy with the decision as operating diep in textmode is not exactly a peanut. I lost against tiger because of the many 8 ply searches and some 6 ply and 7 ply searches later in the game. I know that with a big hashtable on the quad that would have gone ok. However i had to leave the tournament hall to play for my chessclub in the national league. At that moment it was 6 AM at your place. And the crucial game was at move 30 by then or something. I am sure that with a 400MB hashtable and a quad the game which Tiger blew, would have been a win for DIEP. However what happened now is that DIEP didn't survive tactical stage and the few 6 and 7 ply searches were pure blunders. Also some 8 ply searches were bloody nonsense from DIEP. With 10 ply it would've been a win there already. The quad makes that easily. The good thing about the quad is the worst case. Worst case is simply not there. There is always some processor that has seen the non-pv lines, so whatever move opponent makes, you get like 9 ply directly out of hashtable. Also i found out later that i had overloaded my dual machine a bit. it has 256mb ram. 150mb hashtable and 2x20mb for egtb indices and 2x10mb for egtb cache. that's together 210MB workload if bad luck. Now that still would be not a major problem, but this old 800Mhz machine i use as my development machine. i have installed all kind of databases on it and all that crap gets loaded at boot time already and isn't friendly getting cached away. So in short the broken dual K7 mainboard here at home took care for a tiger win in this tournament. That's my viewpoint. Not to mention that i didn't have time for all kind of other things because of hardware problems (more things were broken in the office as well which demanded time last week). Note the book used this day was the pesce book. it really was used because my own book is a horror still and i was afraid for all kind of stupid sidelines from opponents. On paper diep had an equal position against tiger after opening, but i think it plays way easier for white. We've seen that in the game. This is the problem with all Carlos lines. On paper it's equal , but it demands strategics which most engines do not have. Also 1.e4,e6 sucks in principle. Nevertheless i was happy i still could use this august 16 book from Carlos for a few games. Sunday i again used my own book which i regretted bigtime again because the whole line against isichess is from computerviewpoint a bloody draw. However in both these games i didn't feel the hardware lack,b ut i'm sure it would have helped bigtime at crucial moments, 8 ply is simply not enough. Now we all know that if a game is a bit longer because of whatever circumstance (for example i arrived late because in tournament hall at sunday and my clock was already running for 10 minutes by then, of course i'm the only participant where this gets done to, not to others AFAIK, 10 minutes can be pretty much in a 90 0 game). More influence in the games at sunday however was a stupid mistake from me in preparation. I had a few good booklines, but hadn't tested them with tournament book mode at tournament level times that week, for the obvious reason that my machine was broken. Otherwise i would have noticed that some lines i prepared just had a normal chance to get played just like other lines. so the other lines came on the board against isichess and kallisto. both dead drawn lines IMHO for computer-computer games. Now i won't complain about the kallisto game, because i still had the opportunity for some hour and a few minutes to try to win the game after the Qxe2?? blunder from kallisto. I give it 2 questionmarks because it is game deciding, despite that it doesn't give away material, which normally gets 2 question marks. Nevertheless in the isichess game definitely chosing a non-tournament bookline was game deciding. Now we had quickly a drawn position and despite a spectacular sacrafice from diep on h6, that still didn't decide the outcome of the game. It was and remained a draw, which given the way the game had developed was a deserved result for both sides. So in short i think a combination of better testing and not having broken hardware is going to be interesting for future events. I feel sorry for Crafty though that it suffered more from all this than any other program. The reason why crafty lost that game against Tiger is pretty easy to define: that's the 8 pawns on the board penalty in the evaluation from crafty. A week ago or so i made some remarks about this feature inside craftye evaluation. Jeroen was one of the guys who heard this and he has mercilous converted the knowledge into a win by setting up a closed position onto the board against crafty. Crafty of course desperately exchanged a pawn which directly resulted into a lost position in the long term. I was live spectator how crafty got slaughtered after that in a chanceless way. In short crafty was victim of its own open source policy. Now some people will say: "it sucks", but that's not a true judgement. There is simply more known about crafty than any other opponent and this very clearly resulted in less points this tournament than crafty deserved! The opposite happened with this spiderchess program. Nothing was known about it, other than that its programmer lacks more sleep than the total of sleep lack of the other participants. His 6.5 points result is real good of course. Especially considering age of his program. If we however examine HOW he got his points then we see that the program sucks everywhere in open positions, but that he got his points by playing very closed type of style of positions. Like patzer used to play some years ago. Not sure how many years he has worked onto spiderchess, but 6.5 points is a *real* good result, especially considering he has made many draws against strong programs, my own included. Next tournaments most people will realize however that he got his points because he had build in sneaky lines to get to closed positions. In short they'll open the position by good book usage and he'll suffer defeat after defeat, if engine doesn't progress that is. Spiderchess simply shows how important it is to play with a clear thing above playing with something which is unclear and untested. Because IMHO there is not much difference in approach from spiderchess versus crafty (with exception of the 8 pawn penalty anti human feature). Both rely upon very basic chess knowledge and are tactical weak. both are nearly world champ doing nothing and not very good defenders. Crafty is better tuned in open positions and of course has no problems when it faces programs like Spiderchess. But for other programs like DIEP, playing spiderchess is very similar in style to playing crafty or Patzer, because especially the first 2 have such limited knowledge, that the play of these engines is very passive but playing on very important basic things. That means that strategically the luck of these engines is bigger. Now about crafty is more known, so it gets knocked down. Let's hope that Spiderchess doesn't face the same problems coming years, but it's realistically to assume that Spiderchess will get knocked down bigtime next tournaments, when people realize that its playing on such basic knowledge that all you need to do is to open the position against it and it will commit suicide on itself because of a passive playing style. >> >>Restarting diep with an old diep version that was compiled with gcc 3.0.0 >>which does compile nalimov egtb.cpp file, would of course solve that problem. >> >>Even only restarting it at my own 800Mhz machine with egtbs in far >>endgame would solve that problem. >> >>That isn't allowed either is it? >> >>Let's see: >> paderborn99 i used diep in console and not in winboard >> world champs 99 i used diep in console and not in xboard >> dutch champs 99 i used diep in console and not with xboard >> paderborn2000 i used diep in console and not with xboard >> world champs2000 i used diep in console and not with winboard >> in world champs 2000 blitz i used diep in console. That was hell, >> but i had to do it that way, i had no other options. >> paderborn2001 i used diep in console and not with xboard >> i-csvn2001 i used diep in console and not with xboard >> world champs 2001 i used my own GUI and because i never tested it in blitz >> i missed world title blitz by some few seconds as the interface needed >> like 5 seconds anyway to just start search. So i forfeited chanceless >> against goliath one last round. >> >>Now i play Tao, and he uses winboard. So he takes risk. When i arrived >>in tournament hall (i had to play national competition at the same day >>and i had lost) i took over operating DIEP again. Bas couldn't complain >>about losing much time in the start of the game, but when i took over >>in the 2 bishops for diep against Tao Rook and 2 pawns endgame (and some >>other material) endgame, then i warned him after a few moves that time >>was a crucial factor. >> >>For years i played in console as you can see above. Many people always >>forfeited, i was even laughed at in CCC when i forfeited against junior >>in WMCC2000. Now i win on time because of something which i always prevented >>by playing in textmode. Playing in textmode/console is hard. Now i must >>be such a stupid guy to offer my opponent a draw because he is that stupid, >>apart from the fact that just like the tournament before he could lose >>the game again for some silly suicide reason of the engine? >> >>Remember tournament before this i won against tao from a position where >>i was a queen down. Hardly compensation, but then it was a draw. then >>after some time i still won. >> >>Now i have endgame, guy begging for a draw. he forfeits before even proving >>whether his engine is improved. I take the point of course, just like others >>took my point and just like i had suffered all these years for running in >>textmode as i had a winboard engines for some years. >> >>If you use winboard, you take a RISK. If that still isn't clear, i advice >>to use winboard without preparations next tournament. >> >>Bas had before game allocated only 10 minutes operator time for the whole >>game. Game was near 90 moves or so. In short he shouldn't complain. If >>i would have wanted him to forfeit on time he would have forfeited at move >>50 or so. >> >>>On November 03, 2001 at 18:01:24, Bas Hamstra wrote: >>> >>>>Against Diep Winboard's clock was slightly out of sync with the real clock. >>>>Don't ever let that happen! Amazing how fast the difference grows and you CAN'T >>>>adjust the clock in Winboard. A couple of minutes difference is *deadly*. And, >>>>though it is allowed to adjust the engine clock, apparently it is not allowed to >>>>restart Winboard+engine with a corrected time. Vincent was the first to point >>>>that out to the tournament directors. "Not allowed! If I can't restart my engine >>>>for table bases, HE can't bla bla bla..." (Tao +0.90 Diep +0.001 but drawish >>>>IMO). Whatever I would have done in this situation, *not* this pityful "no no, >>>>not allowed!". The 30 nullmoves from Diep that followed to push Tao through the >>>>flag I can forgive, but not the "nono". >>> >>>I personally strongly disagreed with the arbiters decision not to allow >>>you to correct the clock. If that requires restarting WinBoard, so what? >>> >>>Vincent's comment is nonsense. The arbiter was there, he saw what you >>>were doing. He would have seen you changing options in the engine if >>>you did so. Synchronising the clock is expressly _allowed_ by the rules. >>> >>>Another thing is that in the game Sjeng-XiniX I accidentally started >>>without setting my 'operatortime'. I called the arbiter and he allowed >>>it. So why wouldn't he allow it there? >>> >>>If it weren't for the fact that you were running on a laptop, I >>>would have *gladly* asked Miss Winboard to 'accidentally' flip >>>the wrong switch again ;) >>> >>>-- >>>GCP
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.