Author: Vincent Diepeveen
Date: 10:50:43 10/18/02
Go up one level in this thread
On October 17, 2002 at 10:07:41, ujecrh wrote: >When two human players play each other, if one takes a lot of time for a move >usually the opponent also takes this in account for his next move as it is often >a sign of a critical step in the game (tactically or positionnaly). > >This is somewhat used by chess programs because they ponder but isn't this worth >implementing in a chess engine? > >We can track opponent's thinking time and, without trying to match it, add some >time or search extensions when an unusual delay has occured. > >Ujecrh I can assure you that this isn't always the case. Many times a long think from a GM indicates that after he gives you a forced move :) Example is Diepeveen-Piket. after e4 c5 nf3 nc6 d4 cd nd4 nf6 nc3 e5 ndb5 h6 be3 d6 Piket thought for 16 minutes about d6. I took 3 minutes to play Nd5 there. 3 wasted minutes in fact. 8. Nd5 nxd5 ed Ne7 c3 nf5 Here some would have thought for 30 minutes but i'm in the information technology. i really saw no alternative other than Bxa7 so i played it within 3 minutes. 11.Bxa7 Bd7 played very quick. The 3 minutes on my notation form indicate that he had walked away in fact from the table, so Piket took less than the annotated 1 minute. 12.a4 despite this is a logical move i thought for 11 minutes about it. Probably this move would have been easy to find for the computer :) 12..Be7 again played immediately, however clock indicated he had thought for 2 minutes here. But he was talking so someone on the corridor. How you going to make clear to your computer that this time was lost by talking on the corridor? 13.a5,O-O 14.Bb6 Now piket thought for 13 minutes and played Bd7. 14..Bd7 Now i thought for 17 minutes and played Be2, i had to calculate Nh4. So i hope you realize i did NOT think long because Piket thought long about Bd7. Instead i just thought long about Be2 as i feared Nh4. 15.Be2 However Piket is world top 50. In fact with 2646 he's #46 of the world. So for him the Be2 move was very easy to see to be best and for him it was very easy to see that Nh4 was not possible. How does your program understand that? Most up to small GM level dudes go think real long here about Nh4. In analysis Piket showed he knew all about Nf5-h4-g2 being not optimal. For me this was not so trivial to see. Of course in post mortem it was, but anything creating chaos on my board i carefully look at. For Piket this was trivial. Super GMs in complete chaos simply feel which move they can play and which they cannot. The 'cannot' is perhaps most important feeling. So you get confronted here again with that different players at different levels need different time management to see more complicated things. How are you going to program that into your algorithm? What i see in a microsecond some 2000 rated will never see. What Piket saw in a few seconds i needed 17 minutes for. I wanted to calculate how the attack of black and white would go. I don't want to just be busy on the queen side. i want to do something on the king side as well. So Be2 is best move here. 15.Be2,Bd8 Piket had of course given Nh4 also attention but not that long. As a non-IT person he is less efficient. For him it is difficult to say against a FM: "here you get a won position in your hands which is going to be tough to lose against me". So he had problems making Bd8. The Nh4 he had seen quick. He was considering other ideas here. Basically figuring out whether Bd8 was the best move didn't take him that long, but the combination of me getting some drinks and him wanting to play the objective best move versus the subjective moves which might win against some small silly FM, THAT is what took the time. So Bd8 took 23 minutes. Now should i go think for 20 minutes too, wasting my precious time? Because Bxd8 is kind of forced here IMHO. Of course i deliberately didn't analyze it with the computer yet this position. Just 1 position i analyzed with computer. The one where i got a draw. Does 23 minutes from piket mean i must think long too about a forced move? The whole reason that the process took so long was not because he had problems seeing that Bd8 was best move, but because he had problems psychologically with it and other factors like drinks and not being directly at the board played a small role too. 16.Bxd8R the only reason why i wasted so much time was because it is Piket. But it was so obvious that it was a forced move, that i simply wasted my time. In total 3 minutes i needed for Bxd8. Note that i offered a draw when taking on d8, for those who cannot read chessnotation. 16..Qxd8 17.b4,Qg5 18.g3 i thought for 5 minutes about g3 Must piket think long now becuase it is a complicated position or must he think long now because i think long? Anyway e4 is kind of forced here. You see he is not an efficient IT man. I knew e4 was forced here. It took him however 18 minutes. I would not have wasted my time to it. By now left on clock white; black ==> 1:01 ; 0:40 So though this is not time trouble it means that i am right on schedule! 19.O-O? a very bad move from my side. Not amazingly also hard for programs who all love to castle here. 19..Rac8 took piket 6 minutes btw, but most of the time was wasted to offering me a drink. 20.Ra3,e3 21. f4,Qg6 22.nd4 nxg3 23.hxg3 must i think for 10 minutes about hxg3 becuase piket thought for 10 minutes about nxg3? NO!! I played it immediately of course. I have 33 minutes left on clock by now. Do NOT waste your time here. Do NOT give GM Piket a reason to play on at all. If i would have wasted my time like your algorithm would have in at least 60% of the cases here, then i would have lost this game as with 3 pawns against a piece, even though it's dubious; if i'm in time trouble and he too then i know who is going to win, and it isn't me.
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.