Author: Vine Smith
Date: 21:47:17 05/20/01
Go up one level in this thread
On May 20, 2001 at 16:00:32, Uri Blass wrote: >On May 20, 2001 at 15:35:17, Uri Blass wrote: > >>On May 20, 2001 at 14:26:15, Vine Smith wrote: >> >>>On May 20, 2001 at 13:24:29, Christophe Theron wrote: >>> >>>>On May 20, 2001 at 04:25:41, Frank Phillips wrote: >>>> >>>>>On May 19, 2001 at 23:48:48, Christophe Theron wrote: >>>>> >>>>>>On May 19, 2001 at 23:37:31, Ratko V Tomic wrote: >>>>>> >>>>>>>>I'm extremely surprised that my creature managed to survive more >>>>>>>> than 30 moves, given a 300 times speed handicap. >>>>>>> >>>>>>>The flip side is that the current programs running at some >>>>>>>future machines at 300 GHz won't be able to crush the current >>>>>>>programs on 1 GHz any more convincingly (in terms of how >>>>>>>many moves the slower machine can hang on) than what happened >>>>>>>in this matchup. >>>>>>> >>>>>>>This is the same effect that many players have experienced >>>>>>>when upgrading their hardware to 2-3 times faster one and >>>>>>>then being disapponted, after all the expense and hopes, >>>>>>>when they can't even notice any difference in the perceived >>>>>>>program strength (aginst humans). >>>>>> >>>>>> >>>>>> >>>>>>You are absolutely right. >>>>>> >>>>>>I think we are already beginning to experience the effects of dimishing returns >>>>>>in chess on current hardware at long time controls. >>>>>> >>>>>> >>>>>> >>>>>> Christophe >>>>> >>>>>Would someone take the time to explain this simply and clearly, to me. I can >>>>>understand that if you are already beating humans (or some other group of >>>>>players) most of the time, then increasing the speed still means you are beating >>>>>them most of the time and maybe a bit more, but until a machine can see _all_ >>>>>there is to see how would it not improve by seeing more and how can you say >>>>>(apriori) that it will improve only at a diminishing return? In other words, I >>>>>can believe that results against a set of players is aysomtopic, tending towards >>>>>100 percent, but do see why this is necessarily true of the game played by two >>>>>otherwiseequally matched entities. >>>> >>>> >>>> >>>>In my opinion it has to do with the fact that in a given chess position the >>>>number of moves is limited. Generally you have between 20 and 50 legal moves. >>>> >>>>From these moves, only an even more limited subset does not lead to an obvious >>>>loss. >>>> >>>>And from this subset there is an even more limited subset of moves (2 or 3 >>>>generally) that can be played, and chosing between them is a matter of >>>>preference because the amount of computation needed to prove which one is better >>>>is too big for any computer. >>>> >>>>So once you reach the stage where you can see which 2 or 3 moves are playable, >>>>it would take an additional huge computation to see further. >>>> >>>>I think some chess programs on current computers at long time controls have >>>>already reached this stage, and this is why is becomes increasingly difficult to >>>>say which one is better. >>>> >>>>This is a very simplistic explanation which lacks mathematical support, I know, >>>>but that's how I explain dimishing returns. >>>> >>>> >>>> >>>> Christophe >>> >>>Is it possible that there is also a problem with bad evaluations infecting whole >>>branches in the tree of analysis? In Fritz vs. Gambit Tiger at Leiden, Fritz >>>played 21.b4, shutting in its queen. Was this not a dreadful move? Yet, I had >>>Fritz analyze after this point through 18 ply, and the evaluation was just +0.06 >>>(after which it mysteriously halted analysis). And Tiger 14 has reached 20 ply >>>looking at this same position, with an evaluation of just +0.46 after 21...Nc3 >>>22.Rd3 Qf6 23.Kf1 Ne4 24.Bd4 Qf7 25.Bb2 g5 26.Rde3 Bf4 27.Bxe4 fxe4 28.Rxe4 Rxe4 >>>29.Rxe4 Qxd5 30.Re7. Actually, the final position is lost for White after >>>30...Qd3+ 31.Re2 Qb1+ 32.Ne1 Bf5, but White doesn't need to play 30.Re7. The >>>point is that neither program, given even 10-12 hours to think (on a PIII 850) >>>appreciates the disastrous effects of White's missing queen. >> >>I prefer black's position after b4 but >>I am not sure about the disastrous effect of white missing queen >> >>Following the main line of chess programs is a bad idea because it is known that >>the end of the main line is often wrong. >> >>The real test is to give programs to play. >>Tiger beated Fritz in the game but it is not a convincing proof because it is >>only one game. >> >>It is better to play games at slower time control between programs and only if I >>see that black wins all the games I will say that black is probably winning. >>Before seeing it I can only say that I prefer black but I am not sure if black >>has enough to win the game. >> >> >>I think that evaluating it more than christophe does may cause problems even if >>black can win the specific position because sacrificing for the idea of putting >>the queen at this situation is risky because the queen can finally escape by >>ideas like a4. >> >>Finally I can add that it seems that 8 processors may help Fritz to avoid the >>problem >> >>Here is some analysis of Deep Fritz on PIII800(64 mbytes) >> >>Fritz - Gambit Tiger 2.0 >>r3r1k1/2pb1q1p/p2b2p1/Qp1P1p2/4n3/P3BN1P/1PB2PP1/3RR1K1 w - - 0 1 >> >>Analysis by Deep Fritz: >> >>21.Bxe4! >> ² (0.47) Depth: 1/3 00:00:00 >>21.Bxe4 fxe4 22.Ng5 >> ² (0.44) Depth: 2/6 00:00:00 >>21.b4! >> ² (0.47) Depth: 2/8 00:00:00 >>21.b4! Qf6 >> ² (0.53) Depth: 2/8 00:00:00 >>21.b4 Nc3 22.Ng5 >> ² (0.44) Depth: 3/10 00:00:00 >>21.b3! >> ² (0.47) Depth: 3/12 00:00:00 1kN >>21.b3 f4 22.Bxe4 Rxe4 23.Ng5 >> ² (0.38) Depth: 4/15 00:00:00 2kN >>21.b4! >> ² (0.41) Depth: 4/15 00:00:00 3kN >>21.Bxe4! >> ² (0.44) Depth: 4/15 00:00:00 3kN >>21.Bxe4! fxe4 22.Ng5 Qf6 >> ² (0.50) Depth: 4/15 00:00:00 3kN >>21.Bxe4 fxe4 22.Ng5 Qf5 23.Rd4 h6 >> ² (0.44) Depth: 5/14 00:00:00 8kN >>21.b3! >> ² (0.47) Depth: 5/21 00:00:00 14kN >>21.b3 Qf6 22.Bxe4 fxe4 23.Nd4 Qe5 24.Nc6 Qh2+ 25.Kf1 >> ² (0.47) Depth: 6/24 00:00:00 31kN >>21.b3 Qf6 22.b4 Rad8 23.Bxe4 Rxe4 24.Bd4 Qf7 >> ² (0.47) Depth: 7/23 00:00:00 115kN >>21.b3 Nf6 22.Rd4 Kg7 23.Ng5 Qg8 24.Red1 h6 >> ² (0.34) Depth: 8/25 00:00:00 333kN >>21.Re2! >> ² (0.38) Depth: 8/25 00:00:01 528kN >>21.Re2 Nf6 22.Ng5 Qe7 23.g3 f4 24.gxf4 Bxf4 25.Rde1 Be5 >> ² (0.31) Depth: 9/29 00:00:02 1577kN >>21.Bb1! >> ² (0.34) Depth: 9/29 00:00:04 2348kN >>21.Bb1 Bc8 22.Bxe4 fxe4 23.Ng5 Qf6 24.Rd4 Qe5 25.Rxe4 Qh2+ 26.Kf1 Rxe4 >> ² (0.28) Depth: 10/29 00:00:08 4721kN >>21.b3! >> ² (0.31) Depth: 10/29 00:00:10 6301kN >>21.b4! >> ² (0.34) Depth: 10/29 00:00:14 8313kN >>21.b4 Bc8 22.Bd3 Bb7 23.Bxe4 fxe4 24.Ng5 Qf5 25.Ne6 Rec8 26.Nd4 Qxd5 >> = (0.25) Depth: 11/31 00:00:22 13182kN >>21.b4 Bc8 22.Bd3 Bd7 23.Bxe4 fxe4 24.Ng5 Qf6 25.Rd4 Qe5 26.f4 Qf5 >> = (0.16) Depth: 12/31 00:00:41 24964kN >>21.Bb1! >> = (0.19) Depth: 12/31 00:00:58 35516kN >>21.Rd4! >> = (0.22) Depth: 12/34 00:01:40 62541kN >>21.Rd4 Bc5 22.Rd3 Bd6 23.Bd4 Bc8 24.Rde3 Bb7 25.Bxe4 fxe4 26.Rxe4 Rxe4 >> = (0.22) Depth: 13/35 00:03:18 124595kN >>21.Rd4 Bc5 22.Rd3 Bd6 23.Bd4 Bc8 24.Nd2 Nf6 25.Rxe8+ Nxe8 26.Nf3 Bb7 >> = (0.19) Depth: 14/35 00:05:00 190098kN >>21.b4! >> = (0.22) Depth: 14/38 00:06:33 248720kN >>21.b4 Rad8 22.Rd3 Bc8 23.Bb3 g5 24.Nd4 f4 25.Bc1 Qg6 26.Bc2 >> = (0.13) Depth: 15/41 00:15:23 586010kN >>21.Rd4! >> = (0.16) Depth: 15/41 00:18:07 695940kN >> >>(Blass, Tel-Aviv 20.05.2001) >> >> >> >>This analysis suggests that with 8 processors Deep Fritz may avoid b4 at >>tournament time control. >> >>Uri > >and after more time Deep Fritz prefers b3 > > >21.b3! > = (0.19) Depth: 15/41 00:40:11 1535989kN > >Uri Hi -- I do realize that the ending of program analysis lines is often wrong -- that's why I pointed out that White doesn't need to play 30.Re7 at the end of Tiger's 20-ply variation. If White chooses something else, the game is still not clearly won, although the unavailability of the queen is beginning to tell. I might try program vs. program games at tournament time controls starting from the point after 21.b4, but suppose White won some of these. It would not prove that 21.b4 was not so bad, just that it was not so bad against other programs. I suspect that many, if not most, programs do not appreciate the trouble that White gets into from playing 21.b4. Sunne Larsson (I hope I got the spelling right) tested quite a few, and found that Hiarcs and Deep Fritz (after 4 min on something like a PIII 800) select this move as well. I tested a couple of Winboard programs after that and found Phalanx XXII chooses 21.b4, while Yace 0.99.01 prefers 21.b3, but evaluating after 21.b4 (just a quick 11 ply deep) does not think anything is wrong, scoring the position at -0.03. So if I play program vs. program games, the one playing Black may allow White to play a4 without any interference, or even move the c7 pawn (Tiger's analysis at 17 ply has Black doing this). What I'd like to do is get a GM opinion, so I think I may submit an inquiry to Alterman at KasparovChess. Besides asking about the true nature of 21.b4, I'd like to know if he can see any reason why so many programs want to move the b-pawn at all. I don't understand the motivation for even 21.b3, which just seems to weaken a3 and c3 (both squares under attack by Black) in order to strengthen a4 and c4 (neither square under attack by Black). I was glad to see that Deep Fritz might see enough on the Kramnik match 8-way machine to avoid moves like 21.b4, although its shifting back and forth on move choices did not seem to demonstrate that it had decisively rejected b4. It may be that the real problem (if any; I'm trying to keep an open mind) with 21.b4 could be a matter of seeing 25-30 ply deep, in which case Deep Fritz might be deciding against 21.b4 at the end of your analysis based on meaningless differences of a few hundredths of a pawn. Regards, Vine
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.