extremely bad time management of stockfish1.9

[Uri Blass]

I looked at game 26 of the match that is in the rybka forum and I saw that stockfish used only
1 second for a non forced move in 120/40 time control game when it has an average of 3 minutes per move(it used more than 14 minutes for move 91 but it was too late).

http://rybkaforum.net/cgi-bin/rybkaforu ... =ReplyPost

I see no logical reason for it and with more time stockfish could play the better move Qb8+ that maybe save the game(I did not do a long analysis to be sure if 90.Qb8+ save the game but stockfish1.9 on my slow hardware find it at depth 23.

Looking at the game it seems that the problem was not for a single move and I wonder how much elo can stockfish earn by better time management.

[Uri Blass]

I also think that maybe the problem is of the interface because both programs use illogical small times for some moves.

Thinking about it based on the pgn it seems that the interface gave 120 minutes/40 moves for moves 11-50 but even in that case it is not logical to use only few seconds at moves 41-50 or moves 81-90 for both engines unless the interface gives some misleading information.

[bob]

I looked at game 26 of the match that is in the rybka forum and I saw that stockfish used only
1 second for a non forced move in 120/40 time control game when it has an average of 3 minutes per move(it used more than 14 minutes for move 91 but it was too late).

http://rybkaforum.net/cgi-bin/rybkaforu ... =ReplyPost

I see no logical reason for it and with more time stockfish could play the better move Qb8+ that maybe save the game(I did not do a long analysis to be sure if 90.Qb8+ save the game but stockfish1.9 on my slow hardware find it at depth 23.

Looking at the game it seems that the problem was not for a single move and I wonder how much elo can stockfish earn by better time management.

Not enough information. For example, stockfish predicts move X will be played. It sets a target time of 5 minutes. Opponent takes 6 minutes and actually plays move X. Stockfish, Crafty (and I as a human) would likely move instantly unless something catches my attention...

If we had a crystal ball to letus know problems were coming, this might be handled differently...

[Uri Blass]

I looked at game 26 of the match that is in the rybka forum and I saw that stockfish used only
1 second for a non forced move in 120/40 time control game when it has an average of 3 minutes per move(it used more than 14 minutes for move 91 but it was too late).

http://rybkaforum.net/cgi-bin/rybkaforu ... =ReplyPost

I see no logical reason for it and with more time stockfish could play the better move Qb8+ that maybe save the game(I did not do a long analysis to be sure if 90.Qb8+ save the game but stockfish1.9 on my slow hardware find it at depth 23.

Looking at the game it seems that the problem was not for a single move and I wonder how much elo can stockfish earn by better time management.

Not enough information. For example, stockfish predicts move X will be played. It sets a target time of 5 minutes. Opponent takes 6 minutes and actually plays move X. Stockfish, Crafty (and I as a human) would likely move instantly unless something catches my attention...

If we had a crystal ball to letus know problems were coming, this might be handled differently...

It seems to me that the game was ponder off but it also seems that the problem is with the interface that gave wrong information to the engines because I see that both engines had bad time management and use less time in moves 41-50 or 81-90

It seems that the interface added 2 hours for the engines only at move 51 and move 91 because the first 10 moves are book moves and the interface understood 2 hours/40 moves as 2 hours/moves 11-50 but the engine believed it is 2 hours for moves 1-40 and used most of their time in moves 1-40

[michiguel]

I looked at game 26 of the match that is in the rybka forum and I saw that stockfish used only
1 second for a non forced move in 120/40 time control game when it has an average of 3 minutes per move(it used more than 14 minutes for move 91 but it was too late).

http://rybkaforum.net/cgi-bin/rybkaforu ... =ReplyPost

I see no logical reason for it and with more time stockfish could play the better move Qb8+ that maybe save the game(I did not do a long analysis to be sure if 90.Qb8+ save the game but stockfish1.9 on my slow hardware find it at depth 23.

Looking at the game it seems that the problem was not for a single move and I wonder how much elo can stockfish earn by better time management.

Not enough information. For example, stockfish predicts move X will be played. It sets a target time of 5 minutes. Opponent takes 6 minutes and actually plays move X. Stockfish, Crafty (and I as a human) would likely move instantly unless something catches my attention...

If we had a crystal ball to letus know problems were coming, this might be handled differently...

It seems to me that the game was ponder off but it also seems that the problem is with the interface that gave wrong information to the engines because I see that both engines had bad time management and use less time in moves 41-50 or 81-90

It seems that the interface added 2 hours for the engines only at move 51 and move 91 because the first 10 moves are book moves and the interface understood 2 hours/40 moves as 2 hours/moves 11-50 but the engine believed it is 2 hours for moves 1-40 and used most of their time in moves 1-40

That is what happens when certain engines misunderstand TC or do not take into account the clock part of FEN, or the GUI screw this up. To avoid the problem, it is safer to start with pgn files rather than epd. There was an extensive discussion about this in WB forum.

Tester should pay more attention to these things too... The serious ones do.

What GUI was used here?

Miguel

[Uri Blass]

I looked at game 26 of the match that is in the rybka forum and I saw that stockfish used only
1 second for a non forced move in 120/40 time control game when it has an average of 3 minutes per move(it used more than 14 minutes for move 91 but it was too late).

http://rybkaforum.net/cgi-bin/rybkaforu ... =ReplyPost

I see no logical reason for it and with more time stockfish could play the better move Qb8+ that maybe save the game(I did not do a long analysis to be sure if 90.Qb8+ save the game but stockfish1.9 on my slow hardware find it at depth 23.

Looking at the game it seems that the problem was not for a single move and I wonder how much elo can stockfish earn by better time management.

Not enough information. For example, stockfish predicts move X will be played. It sets a target time of 5 minutes. Opponent takes 6 minutes and actually plays move X. Stockfish, Crafty (and I as a human) would likely move instantly unless something catches my attention...

If we had a crystal ball to letus know problems were coming, this might be handled differently...

It seems to me that the game was ponder off but it also seems that the problem is with the interface that gave wrong information to the engines because I see that both engines had bad time management and use less time in moves 41-50 or 81-90

It seems that the interface added 2 hours for the engines only at move 51 and move 91 because the first 10 moves are book moves and the interface understood 2 hours/40 moves as 2 hours/moves 11-50 but the engine believed it is 2 hours for moves 1-40 and used most of their time in moves 1-40

That is what happens when certain engines misunderstand TC or do not take into account the clock part of FEN, or the GUI screw this up. To avoid the problem, it is safer to start with pgn files rather than epd. There was an extensive discussion about this in WB forum.

Tester should pay more attention to these things too... The serious ones do.

What GUI was used here?

Miguel

I saw that the games started with pgn
I read in the rybka forum that it was some beta GUI of ChessOk and hopefully the programmers are going to fix the bug.

Uri