Chess for Android v5.0: Time Control

[Gurcan Uckardes]

The way 5.0.2 tries to handle problem engines didn't help yet. I must now give up black Mamba for ex. It used to quit claiming wins or draws (never wrong by the way when i check) but the result recorded by CfA was loss for BM. Then i used to modify tour.pgn to correct the terminations. I waste additional time but still use the buggy engine.
Now CfA just stalls and the tourneys stop at the same point with lost engine message. The terminated game is not recorded either. Can't we just have cfa skipping to next game like before and adding a trace to the pgn such as white/black disconnected?
Another suggestion for wb engines, please remember my idea to add 0.1 seconds overhead per move to prevent unfair flags.

[supersharp77]

Chess for android is working very well but of course not perfectly (never will)
quite a few of these engines have issues and probably will have difficulty running in Chess for android........Black Mamba -actually had trouble getting a working version in w-32...some versions worked ok ....others never worked at all! GES......sometimes stops in engine tourneys......Napoleon....first created as a w64 engine......w32 came later......knightcap......tricky to run in any gui....(complex engine) firenzina....cfg setting is causing troubles in android.......Textel 1.05...not running arm 5....Ilfil......no luck.......Kiwi......a crapshoot ....plays better out of tourney mode....Rocinate same thing....quite a few of these engines play ok but crash in the tourney mode.......some play fine after you reboot the device......long complicated tourneys probably a no go due to too many unreliable results in the Chess for android gui caused mostly by engine crashes/errors..........but who cares we now have 100's of engines to choose from so......be happy!!

[Gurcan Uckardes]

In fixed time 10 sec: Latest Kiwi, latest Dirty, latest KnightCap are all ok. Rocinante 2.0 is ok too, very rare pb for me.
But better stay tuned for a huge compatibilty list of all. I'll post to my blog in a few days.
Regarding Fire.cfg it works ok for me. Texel 1.05 is ok too. If you have arm5 device arm7 compile doesn't work. Who's still using arm5 anyway?

[Ferdy]

The way 5.0.2 tries to handle problem engines didn't help yet. I must now give up black Mamba for ex. It used to quit claiming wins or draws (never wrong by the way when i check) but the result recorded by CfA was loss for BM. Then i used to modify tour.pgn to correct the terminations. I waste additional time but still use the buggy engine.
Now CfA just stalls and the tourneys stop at the same point with lost engine message. The terminated game is not recorded either. Can't we just have cfa skipping to next game like before and adding a trace to the pgn such as white/black disconnected?
Another suggestion for wb engines, please remember my idea to add 0.1 seconds overhead per move to prevent unfair flags.

I tried Blackmamba 2.0 32bit matching vs Senpai at TC 80 moves / 1 minute, no problem whatsoever using the 10 random book moves.
Engines are set at hash 16mb, 1 core and without egtb.
My Device is Quad-core 1.3 GHz Cortex-A7, 1Gb RAM.

One thing I noticed Blackmamba is using around 60mb when used in analysis, though it was set at 16mb hash. I have not tried to measure it during play.

[Gurcan Uckardes]

I didn't say Black Mamba doesn't play correct games. It does.
What i mean is that it may exit when it reaches PV scores as high as 10, probably claiming "It seems i surely win here, why continue playing?".
Same applies to series of moves with 0 scores. It just disconnects when it's surely a dead draw. This doesn't happen all the time.
Play hundred games to catch such terminations 3 to 5 times.
GNU 6.0.2 has exactly the same behaviour and now i've stopped running these engines in my tourneys.

In following game from my Rapidroid database, tour.pgn recorded 0-1 when actually B.Mamba was crushing Discocheck very clearly:

Code: Select all

[Event "Rapidroid"]
[Site "?"]
[Date "2014.09.07"]
[Round "2.0209"]
[White "BlackMamba 2.0 32bit"]
[Black "DiscoCheck 4.3"]
[Result "1-0"]
[PlyCount "102"]
[TimeControl "1/30"]
[TimeControl "1/30"]
[Device "GT-N7100"]
[Device "GT-N7100"]
[Opening "AH011130"]
[ECO "E52"]

1. d4 Nf6
2. c4 e6
3. Nc3 Bb4
4. e3 O-O
5. Bd3 d5
6. Nf3 b6
7. O-O Bb7
8. a3 Bd6
9. cxd5 exd5
10. Re1 Nbd7 {from PGN 0.005}
11. Qc2 {[16]9/c7c5 30.441} c5 {[15]-23/ 30.566}
12. Bd2 {[15]-3/g7g6 30.024} c4 {[16]-18/ 30.976}
13. Bf5 {[19]12/g7g6 30.016} g6 {[18]-8/ 30.985}
14. e4 {[20]16/g6f5 30.024} gxf5 {[17]13/ 15.975}
15. e5 {[17]0/d6e5 0.549} Bxe5 {[18]8/ 30.451}
16. Qxf5 {[18]19/e5d6 30.029} Bd6 {[16]48/ 30.971}
17. Bg5 {[18]25/a7a5 30.021} Be7 {[16]25/ 30.98}
18. Bh6 {[19]63/f6e8 30.013} Kh8 {[16]-50/ 30.986}
19. Ng5 {[19]100/b7c8 30.027} a6 {[16]-90/ 30.973}
20. Re3 {[17]218/d7c5 30.029} Nc5 {[15]-198/ 30.972}
21. Qe5 {[17]217/c5b3 30.028} Bd6 {[16]-168/ 30.972}
22. Bxf8 {[18]260/d6e5 30.013} Bxe5 {[17]-212/ 15.986}
23. Nxf7+ {[17]260/h8g8 0.071} Kg8 {[]/ 0.002}
24. Nxd8 {[19]280/a8d8 30.022} Bxh2+ {[18]-237/ 30.905}
25. Kxh2 {[18]313/a8d8 30.022} Rxd8 {[18]-246/ 15.978}
26. Be7 {[17]313/f6g4 0.212} Ng4+ {[20]-287/ 30.788}
27. Kg1 {[20]316/d8e8 30.028} Nxe3 {[21]-282/ 15.972}
28. Bxd8 {[19]451/e3c2 30.004} Nc2 {[21]-289/ 30.996}
29. Rd1 {[17]451/c5b3 0.072} Nb3 {[20]-299/ 30.928}
30. Bxb6 {[18]484/b7c6 30.024} Kf7 {[20]-308/ 30.976}
31. f3 {[16]495/b7c6 30.004} Bc6 {[17]-316/ 30.996}
32. Kf2 {[14]495/f7e6 0.021} Kg7 {[17]-335/ 30.979}
33. Rh1 {[16]531/c6b7 30.03} a5 {[17]-350/ 30.97}
34. Ne2 {[19]590/g7g8 30.019} Be8 {[18]-360/ 30.981}
35. Bc7 {[17]624/g7g8 30.02} Kf8 {[17]-379/ 30.981}
36. Rxh7 {[16]829/f8g8 30.022} Ncxd4 {[17]-371/ 18.877}
37. Nxd4 {[18]892/b3d4 30.027} Nxd4 {[17]-361/ 15.073}
38. Bxa5 {[16]842/e8b5 0.078} Nb3 {[18]-374/ 30.922}
39. Bb4+ {[14]916/f8g8 30.005} Kg8 {[19]-406/ 30.995}
40. Rc7 {[13]916/b3c1 0.008} Nd4 {[18]-421/ 30.992}
41. Rc5 {[9]892/g8f7 30.005} Nb3 {[17]-470/ 30.995}
42. Rxd5 {[7]907/g8f7 0.006} Nc1 {[17]-488/ 30.994}
43. Bc3 {[6]898/c1b3 30.004} Bc6 {[16]-489/ 30.996}
44. Rd6 {[8]955/c6e8 30.004} Ba4 {[17]-555/ 30.997}
45. g4 {[12]1099/a4e8 30.024} Bc2 {[19]-594/ 30.975}
46. g5 {[11]1184/g8f7 30.005} Kf7 {[17]-630/ 30.996}
47. Ke3 {[11]1285/c1a2 30.027} Bd3 {[18]-708/ 30.972}
48. Rf6+ {[10]1478/f7e8 30.032} Ke7 {[18]-912/ 30.968}
49. a4 {[9]1512/e7e8 30.017} Nb3 {[19]-1012/ 30.983}
50. a5 {[12]1857/e7d7 30.029} Nxa5 {[21]-1126/ 30.972}
51. Bxa5 {[17]2374/e7e8 30.033} Bc2 {[20]-1313/ 30.965} 1-0
{white resigned:engine exits} {CHECKED AND ADJUDICATED}

[abik]

Hi pal:
I have played both the Android and Windows 32 bits versions. First of them looks stronger. I defeated easily the one for windows, not the case with android. Or just chance?
Fern

Depends quite a bit on what processor was used on the Android device and what processor on the Windows device. Also, were all other settings (hash sizes etc) the same?

[abik]

What is open exchange format?

It is more or less the open counter part of the Android Chessbase compatible engines format. It simplifies engine setup, since the GUI finds the engine on the device (rather than having to copy the engine first from SD card).

[abik]

The way 5.0.2 tries to handle problem engines didn't help yet.

Next release will probably also include more improvements in this area, such as stopping the tournament on engine failure, but recording the game as lost (so that resuming the tournament eventually makes progress). However, if an engine really misbehaves that badly, why continue testing it?

[fern]

Both around half a minute the move.
The Pc is fast, 3mb hash 3 g processor
tablet is Asus, good one.
PC version did too many silly moves. I did an examination with Fritz 12 and he found a mistake at each step.

Fern

[Ferdy]

I didn't say Black Mamba doesn't play correct games. It does.

Sure. I am just showing my test results with given conditions, your TC is even different from my TC used .