Author: Bas Hamstra
Date: 04:19:16 05/30/01
Go up one level in this thread
Vincent, stop bashing Jan. I don't like it. Also some advice: if you want to post a complete book here, at least try to divide it in chapters. Even better would have been to delete 99% of it. I am not saying that your post contains no info. Just that it is impossible to read it all without dying of old age. Toedeledokie! :-) Bas. On May 29, 2001 at 13:02:43, Vincent Diepeveen wrote: >From diep mailing list here a message. Some of you might be interested. >------------- >Hello dear Friends, > >Diep sailed regrettably a bit as expected last weeks. >Very high scores with white. In I-CSVN diep scored 5 out of 5 >with white, in the blitz last weekend DIEP scored 2.5 out of 3 >with white. Regrettably only 2.5 out of 5 with black. > >To start with the last weekends rapid (level 45 minutes all game 10 seconds >incremental): > round 1 : ShrikeX vs. diep 0-1 > >Diep played najdorf. Poisened pawn to be exact. Thanks to Marc L >Bernstein and Erwin l'Ami i had the courage now to let diep play >this opening! Shrike allowed diep to capture on b2. Then Shrike >nullmoved a bit after which i was of course a pawn up. Diep mishandled >exchanging to the endgame a bit and had some problems to win a rook >endgame with 2 pawns up. After reaching the rook endgame my score >dropped and dropped. Diep gave away a pawn quite quickly and i think >the rook endgame was drawable with white, but lucky for DIEP it got >some help from Shrike. Nice to debug why diep went into the rook >endgame and/or why it didn't win it! > > round 2 : diep - Amateur 1-0 > >With some transpositions of move we got into ben-oni. Diep has a >near to 100% score there against any engine if the level is not too >fast. This was the most chanceless game of the whole tournament >despite that it took 51 moves to win! >What happens is always the same. Diep goes for the majority. Score >drops a bit and at that point the opponentmakes mistakes after mistake >after which diep gets the e4 pawn to e5. A passer gets on the board >and the passer nearly promotes. When black is stopping the passer then >DIEP collects the rest of the board and has a material advantage which >easily wins the game! > > round 3 : Ferret - diep 1-0 > >Diep played Najdorf and got itself a splendid position despite that >the number of variations in my tournament book is not so big. >However bad luck struck for diep this time. Something went wrong >with king safety and diep walked its king into the center. Now despite >that Ferret kept on playing nullmoves and my score went high. So >i assume somewhere there a penalty must have had a plus instead of >a minus. Perhaps i'll figure it out tonight. Anyway. Playing with a rook >less lost quick for DIEP though. Complete crush after playing with that >rook less. Great debug info though! > > round 4 : Diep - Postmodernist 1-0 > >Last so many tournaments i get games where even from lost positions >i manage to win for just one positional pattern which is in diep and >not the opponent. This game was such a game. The game started bad for me >with a dutch opening setup (diep white). >First of all i made an operator mistake. Andrew Williams allowed me >to take back. Still need to thank him for that, because with xboard >which he was running taking back is nearly impossible without adjourning >the game a few times in between. >This game initially appeared to get a big disappointment for me. Diep >was quick out of book, long before it had castled or anything like it. >There must be more as a few things wrong with the king safety of diep, >as again it didn't castle but played h4 in a dutch setup. Black had >its queen on e7 so castling short would allow Qxh4 losing a pawn for DIEP. >Secondly long castling was impossible because DIEP had managed to get >a stupid doubled pawn on c3,c4 versus PostModernist backward d6 pawn. >I bet that was the reason why diep allowed itself to get strategically >lost. Just for some pressure on d6. Just one weakness doesn't lose >the game, but strategical disadvantage does! > >So diep was bigtime lost there, one positive thing was however that >DIEP had the bishop pair which if the position would get opened would >get real strong. >For some unknown reasons to me PM opened the position in such a way >that diep could sacrafice an exchange to get its bishops active and >take black king into a pin. Amazingly this WON the game for me after >probably from both sides some shuffle moves. > > round 5 : gambitmaster - diep 1/2-1/2 > >Apologies i forgot which program gambitmaster is, but i can't remember! >All which is in my email box is this game. It's a game to forget. The >game had big similarities with zillions of games from the years >1996 - 1997 which i played. Diep completely outplayed its opponent >in a quiet nimzo indian setup from white. White got itself a passer, >as well as diep. The white passer was weak, diep's passers weren't. >This game was dominated by shallow searches of DIEP. Probably too many >passed pawn extensions triggered stupid lines which weren't relevant. >Searching 8 ply is really too little. Several times diep saw an incredible >draw shot for its opponent. > > 1. d4 (0:01) Nf6 (0:08) > 2. c4 (0:01) e6 (0:07) > 3. Nc3 (0:01) Bb4 (0:06) > 4. Nf3 (0:01) b6 (1:01) > 5. Bg5 (0:00) Bb7 (1:05) > 6. e3 (0:00) h6 (2:51) > 7. Bh4 (0:00) c5 (3:48) > 8. Bd3 (0:00) cxd4 (4:09) > 9. exd4 (0:01) O-O (4:11) > 10. O-O (1:13) Be7 (0:08) > 11. Qe2 (1:52) Nc6 (0:51) > 12. Rfd1 (3:04) Re8 (3:17) > 13. Nb5 (0:56) d6 (2:38) > 14. c5 (1:59) bxc5 (0:54) > 15. dxc5 (0:51) d5 (0:04) > 16. Rdc1 (1:09) Nb4 (1:08) > 17. Nbd4 (0:01) Nxd3 (1:30) > 18. Qxd3 (0:01) a5 (1:30) > 19. Ne5 (0:01) Qc7 (0:50) > 20. c6 (0:01) Ba6 (0:58) > 21. Qe3 (1:31) Ng4 (0:10) > >Despite a small search depth directly after playing Ng4 i directly >had a fail low from +1.xx to draw score if white would play >Nxh6 here. > >Amazingly white missed this chance, probably because it thought white >was better here, as gambitmaster had a +0.34 score. Not a very >impressive evaluation therefore. Black is simplistically won here, >so if you get a draw chance, grab it! > >Anyway, as you can see i had lost quite some time in the game >kibitzing and also had had some phone calls in the meantime. > >Hand operating really is stupid in these events. I was kind of forced >to do it, but it really isn't what i'm going to do next time! > >All my games i several times missed either the opponent moving or diep >moving. Despite that i have great sounds, they are not yet working in my >GUI :) > >Also in the evening i couldn't turn on sound because i'm just 4 feet away >from the bedroom of my neighbours. Only a stone wall is between me >and my neighbours bedroom! > >Anyway, i got 8 ply search depths here. Diep completely blew its won >position because of that. We got into an endgame where diep wasn't very >good, but enough to draw i assumed. I was kind of right. Gambitmaster >completely blew it, but was lucky to discover that 3 versus 2 in a rook >endgame is hard to lose with the 2 pawns. So from a bad endgame diep had >won a full pawn somehow but wasn't able to win the rook endgame as expected. > >I've had games like that in 1996 and 1997 already. Hard to blame DIEP >doing something wrong. It just got lousy search depths. I can at most >work a bit harder to limit all the stupid extensions a bit at blitz >levels. > >I find any shot, but getting 8 ply simply isn't enough! I already figured >that out in 1997 in the world championship very bitterly, and only >can write that again here as being the major problem in this game. > >The same problem also dominated the next game: > > round 6 : diep vs. pharaon 1/2-1/2 > >Diep played white in gruenfeld. Note i so far only had won this in testgames, >though gruenfeld simply doesn't get played by commercial programs as it >usually loses for them. I'm pretty amazed that Franck Zibi allows Pharaon >to play this opening. Probably BECAUSE i haven't auto232 played a game or >10000 yet, he had a bit of luck here. Diep played a quiet gruenfeld line >where black remained in book long after i was out of book. Here diep made >2 strategical mistakes after which black was completely won. > >then i think from both sides a small search depth dominated the game. >Black simply did NOT want to open the position with c7-c5 which would >have easily won for it, also exchanging queens would have won for black. > >Neither of the both things happened. Whole game i was expecting black to >break the position, get into the position and win easily. Usually in >computerchess if it doesn't happen at move 30 then it happens at move 50. > >When time got less for both sides, then Zchess allowed diep to setup >a great center, after which the wing play for zchess no longer won. > >For some reason DIEP exchanged queens which still is a mystery for me, >and when thinking in the time of the opponent i got big fail lows. The >reason for this was obvious. Where passive programs like shredder have >the habit to allow the opponent to make mistakes, the activity play of >diep FORCES the opponent to make strong moves. > >This led me into a lost endgame, which diep drew by superb tactics and >endgame evaluation. > > round 7 : Patzer - diep 1-0 > >Now patzer i can describe as a program written by a very nice person, >but the program is about the most antipositional which exists on earth. >Patzer is one of those programs that LIKES to close positions and likes >to put everything on the wrong color and play on the opposite wing as >where its king is. > >If patzer castles short then it attacks at the queen side, if it castles >long it attacks at the king side. You can blindfolded write that down, >whether it's good or not, it simply is doing it. > >Also a passer in patzer at 7th rank is worth +3.0 pawns, with very few >exceptions. > >Now that's about all knowledge that's in patzer. > >Apart from this very bad evaluation it is tactical very strong and searching >in a very correct way. Roland definitely has made a program which searches >very correctly. > >Amazingly the children moves from programs like Patzer also get diep into >confusion always. Usually diep blunders with patzer in a positional unsound >way. > >That happened here also. In a completely closed position of course the chance >to win for patzer is bigger as in an open position. All (half) open positions >i ever had against patzer, there diep won very easily from patzer. > >I don't need to mention that i probably was completely outsearched every >move, but that doesn't matter. The whole thing here was that diep got itself >a better endgame, BUT CLOSED. Diep managed to get a good bishop versus patzer >a very bad one. > >How i then lose such an endgame is a complete mystery to me. Diep allowed >patzer to get too much play on the queen side for sure, but there were some >factors in the position which i also thought would easily win for me! > >diep was up +1.7 at times, patzer was up +1.3 to +2.0 at the same time. > >So that's more or less 3 pawns difference in scoring. > >Now i completely agreed with diep every time, but there was one small factor >of luck which patzer had, which is always happening in computerchess. And >that is that all knights were exchanged. > >Only a good vs bad bishop was on the board and 2 rooks each. Also both >had a passer and obviously diep overevaluated its own passer and patzer >also its own. > >Then some stupid pawn level which i was expecting all the time, after >a move or 15 finally was seen by patzer and i lost the game! > >A big blow in my face, though i need to admit that knowledge in closed >positions simply works different as in open positions, so getting always >a closed position with a very antipositional program is a very smart >thing to do. It again won against diep here! > >I'm pretty sure at a bit slower level DIEP would have WON this >game chanceless as it would not have given patzer its passer! > >Because diep could have EASILY prevented patzer from getting its passer! > >The passer of patzer was completely hung in my evaluation probably, but >by tactical means could be defendend. Big luck for patzer there, because >i've had plenty of games against patzer where diep COULD easily win the >+3.0 passer and win chanceless. > >Patzer is the kind of program that is always dangerous because it gives >such a HUGE scores for one factor, just like tiger does in some cases. > >From both sides very unimpressive play however. Dominated by weak moves >of patzer, answerred by weak moves of diep. > >Thereby patzer had the luck to get quickly into an endgame against DIEP. >In a middlegame it would have been CHANCELESS again, with induction to >knowledge and because of having a bunch of pawns on the wrong color... > >Nevertheless i have great debug info again from this game to improve >diep's endgame. > > round 8 : crafty - diep 0-1 > >For some easy reason i had again black here. The higher your ranking, >the bigger the chance you have more white as black. In this case crafty >was higher placed as diep, so i had black. > >Now this is very bad because getting 5 games black and only 3 games white >with an engine that wins every non-dead lost white game is not a good thing! > >This is why all tournaments have an odd number of rounds, to prevent such >big differences in white & black. TWO games more with black. > >Note that the pairing every round took 40 minutes or so. The dudes >paired by hand. Big losers on ICC. Note that at the homepage it said: >"pairings made by program". > >Anyway easier is to get a better ranking next time because that means >5 white games instead of 3 white games. I feel we're not getting less >or more rounds next time with this cct tournament... > >This pairing was extra bad, because if there was a program which deserved >a bit more points this tournament then it was crafty! > >But to get to the game. Same line as against ferret up to 12th move >of black. I was pretty amazed >actually. Any automatic generated book would have played in poisened pawn >the Qd2 move allowing me to get a REAL poisened pawn by taking on b2. > >However crafty played Nb3. I DEFINITELY like to know why, because a >posting of Bob: > Posted by Robert Hyatt (Profile) on May 28, 2001 at 00:46:07: > > " > [snip] > > > > >It shouldn't be the case that someone signs a contract with > >ChessBase and then > >sighs with relief because they know they're going to gain > >an extra couple of > >points at the next tournament, because they can use > >opening book technology > >shared by all of the Chessbase products. > > > >bruce > > I won't disagree. But it is already happening. If I ever play > in what I consider an important event, it will be one of my > goals to take some sort of evasive action, book-wise. > Or take my main goal: to make hand-tuned books > obsolete. I believe that is possible over time. > I'm going to try to do it." > >Oh well obviously crafty isn't playing with a non-hand tuned thing here, >because any automatic generated book will see that after > e4 c5 nf3 d6 d4 cd nd4 nf6 nc3 a6 bg5 e6 f4 qb6 > a) 8.Nb3 has a bad score for white > b) 8.Qd2 is played way more times and has a better score for > white. > >But main factor is that Qd2 is at least 10 times more played as Nb3, >which will render any thought of playing Nb3 useless! > >Anyway my surprise became even BIGGER, when the exact same line as >against ferret came onto the board. Only at move 13 a different move >was selected, probably even better as the book move played by ferret. > >If that's non-hand tuned then i'll eat my hat! > >Anyway, DIEP frightened its programmer here by quickly castling, but after >a few nullmoves of crafty, things were very clear in favour for diep then. > >I need to note that in past diep was very bad with opposite castling, but >that this has improved dramatically last year when i fixed a bunch of >old king safety patterns after a disastreous game Diep-Patzer in DCCC2000. > >So the endresult was 5 out of 8 score. > 2.5 out of 3 with white > 2.5 out of 5 with black > >In short i couldn't complain too much about book except for against Pharaon. >Of course there were some reasons too to complain about the opening against >PostModernist, but those complaints are basically showing the strategical >lack of DIEP (and engines in general). > > GENERAL CONCLUSIONS > >It's very obvious nowadays that engines have become stronger and stronger >because of better testing and simply more time that has been put into the >engines as one could do in the past. I do not find it surprising that some >so called 'amateurs' score very well against commercial engines nowadays. > >To examine the advantages which commercial engines had in the past >versus nowadays: > > 1) SEARCH DEPTH ADVANTAGE > >Commercial engines were faster a few years ago because they were >optimal programmed in assembly, so they searched deeper or were >tactical stronger usual as amateurs. A good example is the dominating >genius engine which got even at a slow Pentium 200 already 200000 >nodes a second. Until a ply or 10 tactics completely dominate in the >middlegame. > >Nowadays everyone gets that depth, except some badly programmed programs, >and the programs that use expensive knowledge that scans. It's not that >the smaller nodes a second (65k nodes a second for diep at a p3-800 dual) >is giving such a bad search depth, it's also the huge diversity >of scores which takes care that ordering moves is much tougher, as well >as the fact that i cannot do a cheap quiescencesearch, combined with >the fact that current diep version is doing all kind of silly extensions, >in order to solve testsets quicker. > >I bet no other program except for Quest and Schach were faster as Genius. > >Nowadays however at tournament level *everyone* gets decent depths. Even >8 ply DIEP in the year 2001 is not comparable with 8 ply DIEP 1997. > >With 8 ply searches with the current diep i completely destroy anything >from 1997/1998. Very seldom Jan Louwman plays programs like nimzo98 just >for fun with DIEP and it's a complete walk over then. > > > 2) SUPERB TESTING > >Now this is still the big difference between most commercial engines >and amateurs. Though some claim they play zillion of games on chess servers, >my question is always: "And what did you do with it then?" > >Now that most commercial engines join the chess servers this is of course >changing, but definitely a few years ago playing online was fun, but >from DIEP perspective it was not so much worth. > >This might be different for others though. But for DIEP it's very important >that i get loads of games WITH the log files of diep and WITH the scores >of the opponent. > >This is also why i learn so much in tournaments, as i can ask for the >score of the opponent. > >With a huge evaluation function as DIEP a very easy way to always win >from opponents is to basically have the same score and if there are a >few patterns that boost DIEP's score, then i win with induction of course! > >So if i lose a game because DIEP had the wrong score for whatever reason >except a small search depth, then i nowadays work on that. At ICC nor FICS >i get games emailed with scores of DIEP + opponent. At auto232 i DO GET >games from Jan Louwman with scores of both sides! > >Of course playing at home under my own interface against a winboard >opponent is no big difference there, because i can collect from both >sides the scores then too. > >But most people simply DO NOT REALIZE how many games commercial engines >tested before they play a bit decent chess. We talk about thousands of >games for each beta version at slow levels like 40 in 2. > >Diep has a huge evaluation function and loads of code to >guide the search. Especially in the first i change weekly things, so >testing is very crucial thing for me. I need to mention that despite >hard work of Jan Louwman, i still didn't test diep with the current >openings book at the auto232 player!! > >This might really amaze you, but last weeks i have tried all kind of >different opening books. > >From Marc L Bernstein books, to books of Cock de Gorter, >as well as automatic generated books. > >So when taking commercial standards, DIEP was laughable tested, >as i didn't play any game with the book i played the tournaments >I-CSVN or CCT3 with. > >With commercial testing i mean: testing an engine using the same >settings as where you play a tournament with! > >In this respect especially Shredder and DeepFritz and Tiger >still have a major advantage towards the rest of the world. > > > 3) Superb Opening book > >A few years ago the openings book issue was also there, though less >important as it is nowadays. Nowadays this is probably the only advantage >which some commercial programs have compared to most of the world. > >Shredder at random book completely gets annihilated by DIEP for example, >just like it lost at CCT3. > >However for tournaments these guys have a special very well tested and >well prepared book at home which definitely favours them, because what >the book makers already realize for years: you need a position where >an engine plays strategically well! > >Even a bad engine can win with a strategic advantage nowadays if it >realizes it! > >The most simplistic case for DIEP are ben oni lines where DIEP has white >and clearly realizes its pawn majority. It will always push it forward >to create a passer on d5. So also programs that are very happy with >passers and search very deep might do very well in that opening with >white :) > >A very bad case is if i play KID with DIEP with black. That's >completely suicidal against a good book, because diep will get >attacked on the queen side and react on it by defending on the queenside! > >This is a completely wrong strategy in KID! > >Nowadays all engines, and not only because of search depths, have become >much stronger in endgame, middlegame and everywhere. Even Gambit >Tiger 2.0 has nowadays some sense of bad/good bishops seemingly. > >Also most engines do not play g2-g4 when being castled short just like >that! > >Such laughable moves really are fixed in the EVALUATION of the engines >nowadays. > >Now we all should know that chess is a weakest chain game. If SOMETHING >of you is very bad, then you lose becuase of that. > >Definitely opening is from most programs the weakest chain nowadays >where it wasn't in the past. > >The real reason for this is because the programmers themselves hardly >have any chess rating themselves, or in my case hardly invested time >in making their own book. > >The tournament book which DIEP has for example is around 6000 moves. >Behind that i use several books, either from De Gorter, automatic >generated, or from Bernstein. > >Yet entering those 6000 bookmoves cost me at most a week of work, >where making my engine has cost me 7 years now. > >Most persons will wonder why, well there WAS a good reason to do so. > >Every week new theory gets discovered, usually refuting the lines >from a week before, so updating your book is weekly work! > >However, if engine gets improved you keep that improvement forever! > >Further doing everything myself: > - interface > - engine > - book > - homepage > - daily answerring loads of email > >That's a bit much work of course! > >Now compare my book work to that of a normal 2200 chessplayer, who >weekly checks out the latest theory, buys every few months a new openings >book, and is doing anything to keep up with theory! > >After losing a SINGLE game such a player goes checkout what went >wrong in OPENING. At least half of the analysis after the game are >dedicated to the opening usually! > >KNSB (dutch chess federation) rating list 1 april 2001: > >2400 7814059 Kaan J.E.F. M 1245* 14- > >Now compare that with a 1245 rated 'born loser' Jan Kaan who >played in I-CSVN with Yace with his own home made book. Of >course that book doesn't make a chance, not to mention that >it lost strategically in every line for Yace! > >Because that is the bitter reality. Now you don't need too much rating >yourself to make a good book, as making a book is completely different >from playing chess in a tournament. > >However, some strategical insight is at least needed to make a book. >I know some 1200 rated KNSB players who play correspondence chess and >manage to get good openings positions there, but of course Jan Kaan >is completely chanceless against guys who DEFINITELY know strategics >a bit like Pesce who made a book for Yace and Pharaon. > > 105872 Pesce, Carlos ARG 2045 0 19.04.61 > >I'm really amazed that programmer of Yace allowed Jan1245 to delete >at home the very good Pesce book and replace it by a Jan Kaan >"automatic generated beginners book". > >Because strategically Jan Kaan is of course infinitely and with >induction to forever worse than Pesce, who is at a chess level that >he might outbook any IM and some GMs also. > >'Kaan automatic generated beginner books' of course >also have no chance of course against for example: > >Cock de Gorter (KNSB rating): > 2400 6164532 Gorter C.H. de M 1950 3- > >Now this 1950 tournament rating is extra good >considering that Cock is somewhat older and he's more busy >with correspondence chess anyway. > >Jeroen Noomen (KNSB rating): > 2400 6495632 Noomen J. M 2165 4* > >Jeroen is always won after opening in his games (Jeroen plays >masterclass league in netherlands, so every round he faces >either an IM/GM, or someone with a rating he'll never reach; Jeroen >is one of the lowest rated masterclass players which play this >year masterclass league). > >Jeroen gives however regurarly pieces away himself. > >From Alexander Kure i can give rating info from first hand, >but he isn't on any FIDE list. We have >played quite some blitz over the times and in all games he was won >out of the opening, but somewhere blundered away a piece every game. > >So i won them all but one. > >Alexander Kure doesn't play actively tournament chess anymore but >i would estimate him at at least 2100 if he would start playing now. > >Alex is analytically definitely also in IM/GM league, especially >considering that his profession is programming, this makes him very >dangerous against automatic generated opening books as he knows >exactly how they get generated! > >Vincent Diepeveen (KNSB rating): > 2400 6614817 Diepeveen V. M 2285 27* >Vincent Diepeveen (FIDE rating): > 1003054 Diepeveen, Vincent NED 2281 14 16.10.73 > >I am usually lost after opening, and i hardly give away a piece, >and i usually nail my opponent by exchanging into a safe to win >technical endgame. If i have more as 15 minutes before the first >time control (40 moves usual) and my opponent 5 minutes or less >for a bunch of moves, then my record is a 100% score (that >includes GMs & IMs). Note that all my team members call me 'diep' >usually, just like my whole family is called in slang already >for 250+ years here in this town. > >But quite a contradicting style between me and the book makers. > >Not amazingly with a style that's technical and positional better >as any program, and tactical quite a bit less, but good compared >to most of my opponents, my rating is going up and up, though with >small steps of a few points each list, because i play very little >rated games internationally. > >However the real weak chain in my own play is the same weak chain >as that of most engines: book. > >Yet in contradiction to most engines i nearly >ALWAYS PLAY THE SAME OPENING. > >So the bookjob for me is quite simple compared to the book job which >most would have for their engine if they would maintain it. > >Jeroen for years played certain slav lines with Rebel and of course >Tiger plays at tournaments mainly these well checked lines, whereas most >engines play every round a different line. > >Also Alexander Kure (maker of fritz & nimzo & all chessbase books >that join in tournaments) has the same record here. > >In a world championship you can already be prepared that the same >prepared openings line gets always played by Kure. No exceptions and >i cannot blame him there. Outbooking it is nearly IMPOSSIBLE, unless >one did some real good work at home. > >So far only Noomen qualified for that. In past also Necchi did. > >I clearly saw how Noomen's book improved, as well as the big >quantity of lines that Kure uses to kill automatic generated books, >and also adding to that very modern lines for the quality at world >champs. > >Necchi still has to proof that his level also has increased, but >i don't doubt he'll manage to proof his point. > >Just generating a book from GM games in an automatic way is only going >to be a sure zero against Kure and nowadays also Noomen. THEY ARE PREPARED >FOR ENGINES HAVING AN AUTOMATICALLY GENERATED BOOK! > >Necchi already showed an outbooking mentality at SSDF years ago... > >That's a matter of a few thousands of games at a number of auto232 players. > >Now a few very inexperienced computerchess fans perhaps >already wonder: "do the book makers have such bad ratings, how can >1950 to 2200 rated persons make books for 2600 beating programs?" > >Well here is the bare truth: programs are strategical rated 1500 points >at most. Meaning they either do something or they do not do something >which is needed to do. > >So strategically all these guys are hell better as any engine. > >Secondly if Jeroen Noomen would play me in a match then i probably >win it. My rating is higher, so i have a higher chance to win, but >analytically he is in the same league. Add to that that he understands >quite something from opening! So he might lose because i nail him >somewhere in the endgame, where i definitely know more, or i might >trick him in the middlegame with some tactics, but nowhere in the opening >he'll be in danger that he doesn't understand something which i do. > >Same is true for most IM/GMs. I have no problems to follow games of 2600+ >rated persons. Same applies to all these guys. Of course, 2600+ will >always win from me in a match. Especially opening where they prepare >daily. But take for example Vaganian. He plays a lost openingsline with >black! > >Yet by means of superb tactics he has a 100% score against me!! > >Analytically he is in no way better in opening or middlegame as i am! > >Because of his age and professional play for half a century >he'll know hell more of endgame as i ever will. > >But that doesn't mean i cannot follow his games. I can in fact follow >his games very well. Idem for the above book makers! > >So strategically every player mentionned here, with exception of Jan Kaan, >will blindfolded beat a program strategically. Of course they all lose >tactically, but opening is not about tactics only. It's basically >setting up the game in a STRATEGIC sound way. > >Here amateurs only slowly catch up with commercial programs. But we >must not forget that some commercial programs of the past who failed >to keep up with book, positional insight, that those have completely >fallen away. > >Good examples are of course Genius, Kallisto, Wchess, >Virtual Chess and the list can go on. > >Though all these programs still could do well in blitz, they positionally >and strategically are no longer in the same league as the modern >chess programs. > >So there is a kind of natural decline there. For sure the programs >with good books survived and others joined a company who offered >good books, but my expectation is that this big book advantage is >also going to get less when some 'amateurs' prepare well. > >Some in fact already do. > > > GUESSING THE FUTURE > >Where the book of Noomen and Kure have kicked many butts the last 2 years, >usually getting near to 100% scores against bad prepared amateurs, >others have kept up with it. > >Splendid jobs have been performed by for example Dan Wulff who makes >book for Gandalf and Josef Zwinger who makes book of Insomniac (though >its book lines are basically favouring white), and >still the lines played by Cock de Gorter (The King, Tao, and also DIEP >sometimes i test with his book) still can't be called suicidal as they >all are strategically sound. Pesce definitely made impression last CCT3 >tournament! > >Sometimes old lines also work well if they're strategic sound you know! > >The more new openings theory gets invented, the more lines there are >of course which can be called strategic sound. So in the future, >just like in the past, after this chain has been made a bit stronger, >in engine-engine matches the strongest engine will again win. > >What we already see is that some 'amateurs' are taking out commercial >engines by surprise. In the future the world champion will be the >engine that can beat amateurs the best! > >Best Regards, >Vincent
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.