Having worked in engineering for 44 years there is a well used proverb that is probably equally applied to running chess matches. "Experience gained is directly proportional to the number of mistakes made".Sedat Canbaz wrote: ↑16/01/2022, 14:39 No no...all my testings are meaningless, because I run PONDER ON
I have started yesterday these testings...for this reason,
And in case of next time, before running any test...I may ask a help from you..)
And it seems,
You are the biggest expert that I have ever seen...!! you know everything right ?
That's why I think you must know all these issues...should not be so hard for you...
So one last question, it seems you missed,e.g I asked before...
Can not explain please about the time loses on my latest test via Ponder OFF
And Why when is Ponder ON, Cfish does not loose on time.. ?
Thanks
For your attention ...
There is another issue with playing games on 2 PCs with ponder off and ponder on that is exacerbated by fast blitz games. There is an RS232 data transmission and data handling time cost for each transmitted move.
The 4 core ponder on games record confirms the minimum time to send and then receive and interpret the 8 core ponder off move data is close to one second. Therefore in addition to the ponder hit gain while the 8 core ponder off machine is "thinking" the 4 core ponder off engine also enjoys a further bias in its favour of an additional 1 second per move for each ponder hit. The 8 core ponder off engine cannot benefit from this on the return move.
For a typical 90 move game and a conservative 60% ponder hit rate this would give the ponder on engine an extra 48 seconds plus the ponder hit time gain based on 10 moves openings book = 80 moves x 60% = 48s at 1 s/move time delay.
This would not be such an issue for long time controls but at Game in 180s an additional 48 seconds is a massive advantage for the ponder on engine and disadvantage for the ponder off engine that inevitably distorts the result. Certainly a 1 second time bonus for each ponder hit in the usual mad time scramble ending is significant.
As for time losses, other than the engine being unable to apportion sufficient time to manage the game length that is a very difficult task for longer games, the only time I have experienced time losses was with the introduction of Windows 10. The power plan advanced settings need to be checked to ensure the disk drive is not being shut down that will introduce a time delay to restart the disk when Syzygy or Nalimov Table bases need to be accessed.
The inactivity measurement criteria is based on no keyboard or mouse activity to determine the start of idle time that begins the countdown to shutting off the hard drive. Some anti-virus and protection software also use this too but they are best disabled for matches.
On my later machines Syzygy bases were installed on the SSD but Nalimov bases remained on the hard disk and were used by the GUI. The time loss issue in this case was not caused by the engine but by the GUI waiting for hard disk access when the game went into a 6-man Nalimov endgame position. Adjusting the hard drive time out to never shut down overcame the issue. The time should be set to 0 (zero).
Peter