nanoSzachy 1.7

[Piotr Cichy]

New version of nanoSzachy is available.

[Jim Ablett]

Hi Piotr,
Maybe a time control bug in ther new version.

Timed-out versus Mscp at fast time control
near move 40.
(40-05-0 / Arena 1.1 /WinXp-Pro)

Regards,
Jim.

[Piotr Cichy]

Hi Jim

Thank you for the information.

The change in version 1.7 is increasing time for move in certain situations. This is of course some risky, especially near move 40, 80 etc...
Maybe this is the reason, however in my tests with version 1.7 (about 50 games) timeout never occured.

I never give more than 3/n of the remaining time, where n depends on remaining moves and is always at least 4. But time is checked every 16384 nodes, so real time for move may be a bit longer. Maybe this is a reason of timeout at fast control?

Do you have Arena log of this game?

[Jim Ablett]

Here it is.

Sorry I took so long, my Arena debug file is
31000 pages long (66mb) and Nanoszachy was right at the end.
Have you ever tried to edit a text file that big. My computer
ground to a standstill!

By the way Nanaoszachy was about to lose the game
to Mscp before the timeout. I guess it just couldn't
take to humilliation and refused to play any longer!

Download here >

http://homepages.tesco.net/henry.ablett/nano-debug.rar

Regards,
Jim.

[Piotr Cichy]

Hi Jim.

Thank you for the debug. I found something very strange in it - the time shown by nanoSzachy and Arena is different and the difference is over 1 second per each move. For last 3 moves time per move shown by Arena is: 6.4s, 10.25s, 9.25 s and shown by nanoSzachy: 4.18s, 8.81s, 6.60s.
Here are some lines from debug:

Code: Select all

2557485*2*Start calc, move no: 71
2557844>2:f3f4

2563860<2:stat01: 418 1723224 8 0 26 h7h6 411Knps
2563891<2:move e5c5
2563891*2*Found move:Re5-c5

2569891*2*Start calc, move no: 73
2570250>2:c2e4

2580141<2:stat01: 881 5282895 9 15 24 f7f6 599Knps
2580141<2:move f1c4
2580141*2*Found move:Bf1-c4

2588329*2*Start calc, move no: 75
2588688>2:b3b7

2597547<2:stat01: 660 4640715 9 13 22 g7f8 702Knps
2597579<2:move a7a5
2597579*2*Found move:a7-a5


Can somebody explain how is it possible? I use Windows function GetTickCount() to read the time.

---

By the way, you wrote: "NanoSzachy was about to lose the game to Mscp before the timeout".
I don't agree. The position seems to be equal or even with small advantage for black!

[Jim Ablett]

The bug's happening quite often here's a debug from
another game >

16701125<2:stat01: 4 39333 6 9 26 e5d4 836Knps
16701125<2:stat01: 4 39334 6 8 26 e3c3 836Knps
16701125<2:stat01: 4 39335 6 7 26 f3f2 836Knps
16701125<2:stat01: 4 39336 6 6 26 e5d6 836Knps
16701125********Game end: 1-0 {1-0 Black forfeits on time}
16701141<2:stat01: 4 39337 6 5 26 h7h6 836Knps
16701141<2:stat01: 4 39338 6 4 26 a7a6 836Knps
16701141<2:stat01: 4 39339 6 3 26 h7h5 837Knps
16701141********Last game won by Mscp 1.6 with White in 39 moves
16701141<2:stat01: 4 39340 6 2 26 h8g7 837Knps
16701141<2:stat01: 4 39341 6 1 26 h8g8 837Knps
16701141********Mscp 1.6 - NanoSzachy 1.7: 1-0-0
16701156<2:stat01: 4 39342 6 0 26 e5f4 837Knps
16701156<2:stat01: 4 39343 7 25 26 e5g7 837Knps

Jim.