Chess for Android: double blunder between Komodo & Stock

Discussion of anything and everything relating to chess playing software and machines.

Moderator: Ras

Max
Posts: 247
Joined: Tue Apr 13, 2010 10:41 am

Chess for Android: double blunder between Komodo & Stock

Post by Max »

Just watched a game between Komodo and Stockfish at my little tournament on Samsung Galaxy Tab with Aart Biks great new Chess for Android 2.9 gui. Komodo outplayed Stockfish and reached this position

[d] 5R2/8/1Bk5/P7/4K3/8/r4p2/8 w - - 0 57

Komodo blundered now with 57.Rxf2? and Stockfishs evaluation moved up to 0. But to my surprise Stockfish 2.11 did not go for the exchange and played Ra3? with -6.58 after 10 seconds.

Chess for Android setup:
- level 4 (with 10 s/move)
- no tablebases
- 32 mb hash for every engine

Looked a bit strange to me, that Stockfish always moved after about 10 seconds. While Komodo used his time more intelligent and takes between 1 and 33 seconds.

Is there anything broken in the time management of Stockfish for Android?

-Max

Code: Select all

[Event "Chess for Android 2.9"]
[Site "Galaxy Tab 7"]
[Date "2011.05.20"]
[Round "1"]
[White "Komodo 1.3 JA"]
[Black "Stockfish 2.11 JA"]
[Result "1-0"]
[PlyCount "127"]
[TimeControl "1/10"]
[Device "GT-P1000"]

1. c4 {[opening] 0.0} e5 {[opening] 0.001}
2. Nc3 {[opening] 0.004} Nf6 {[opening] 0.001}
3. Nf3 {[opening] 0.0} Nc6 {[opening] 0.001}
4. d4 {[opening] 0.002} exd4 {[opening] 0.002}
5. Nxd4 {[11]5/f8b4 9.6} Bb4 {[13/21]16/d4c6 10.99}
6. Nxc6 {[11]1/b7c6 11.111} bxc6 {[15/22]8/e2e3 10.469}
7. e3 {[12]-2/e8g8 10.325} O-O {[15/20]20/f1d3 10.524}
8. Bd2 {[13]-13/d7d5 40.043} d5 {[13/19]28/f1e2 10.297}
9. Be2 {[11]-13/b4d6 12.317} Rb8 {[14/20]20/e1g1 10.897}
10. O-O {[11]-10/b4d6 18.904} Bd6 {[14/21]40/d1c2 10.273}
11. Qc2 {[11]-16/g7g6 9.858} Qd7 {[13/19]24/c3a4 10.609}
12. h3 {[10]12/g7g6 8.467} Rd8 {[12/18]20/b2b3 11.232}
13. Rfd1 {[10]7/d7f5 13.122} Qf5 {[12/19]44/e2d3 10.983}
14. Bd3 {[12]12/f5e5 6.746} Qh5 {[14/20]52/g1f1 10.693}
15. cxd5 {[11]-5/c8h3 33.11} Bxh3 {[14/28]72/g2h3 10.335}
16. gxh3 {[12]-23/h5h3 28.205} Qxh3 {[14/22]113/d3f5 10.692}
17. Bf5 {[12]-25/f6g4 9.009} Qh2+ {[13/29]40/g1f1 10.549}
18. Kf1 {[16]-5/h2h1 11.937} cxd5 {[13/25]60/c3e2 11.219}
19. Ne2 {[11]9/g7g6 10.644} Qh1+ {[13/27]72/e2g1 10.924}
20. Ng1 {[13]-84/f6e4 9.176} Ne4 {[14/29]68/d2c3 11.109}
21. Bxe4 {[12]-98/d5e4 8.374} dxe4 {[14/28]141/d2c3 10.538}
22. Bc3 {[12]-65/d6h2 10.872} Bh2 {[14/26]214/f1e2 10.147}
23. Ke2 {[10]-74/h1g2 1.549} Qg2 {[15/30]173/g1h3 10.698}
24. Rf1 {[12]-106/c7c5 22.136} Rd3 {[12/23]238/c3d4 11.38}
25. Bd4 {[10]-106/b8d8 8.6} Rd8 {[12/24]258/a1c1 10.554}
26. Qc5 {[10]-103/g2g4 7.001} c6 {[11/20]173/c5a7 10.64}
27. Qe7 {[9]-32/d8f8 7.492} Rb8 {[13/20]105/e7a7 10.562}
28. b3 {[9]-53/g2g4 9.109} a5 {[11/19]145/a1c1 11.155}
29. Qa7 {[10]-5/h2d6 8.36} Rf8 {[11/24]139/a1c1 10.313}
30. Qxa5 {[10]26/g2g6 6.834} Re8 {[11/19]68/a5f5 10.44}
31. Qf5 {[12]62/c6c5 7.543} c5 {[12/21]-64/d4c5 10.676}
32. Bxc5 {[12]28/d3c3 15.773} Rc3 {[13/23]-109/a1c1 10.546}
33. Rfc1 {[12]38/c3c1 7.655} Rxc1 {[13/22]-141/a1c1 10.478}
34. Rxc1 {[11]40/h2g1 1.199} Bxg1 {[13/26]-173/c5d4 10.687}
35. Bd4 {[13]50/e8f8 7.551} h5 {[13/27]-185/c1c7 10.288}
36. Rc7 {[12]71/g2g4 7.062} Qg4+ {[14/28]-210/f5g4 10.515}
37. Qxg4 {[12]81/h5g4 1.819} hxg4 {[15/32]-234/b3b4 10.38}
38. a4 {[14]103/g1h2 6.859} Bh2 {[16/27]-189/c7c6 10.102}
39. Rc4 {[13]127/f7f6 10.127} Ra8 {[16/28]-189/d4c3 10.269}
40. Bc3 {[14]120/f7f5 6.805} f5 {[15/29]-238/a4a5 10.096}
41. a5 {[14]109/g4g3 15.524} Kf7 {[17/29]-214/c4c6 10.865}
42. b4 {[15]123/f7e6 9.518} g5 {[14/25]-206/c4c6 10.429}
43. b5 {[12]120/f7e6 27.594} Ke6 {[14/26]-242/b5b6 10.921}
44. b6 {[13]131/e6d5 11.485} g3 {[15/31]-319/f2g3 10.808}
45. fxg3 {[13]182/h2g3 6.559} Bxg3 {[17/33]-323/b6b7 10.402}
46. b7 {[13]217/a8a6 25.828} Ra7 {[16/30]-327/c4b4 10.697}
47. Rc6+ {[14]235/e6d7 10.132} Kd7 {[16/30]-404/c6b6 10.442}
48. Rb6 {[15]240/f5f4 15.212} f4 {[16/31]-513/e3f4 10.722}
49. exf4 {[14]249/g3f4 10.269} gxf4 {[17/33]-541/b7b8q 10.126}
50. b8=Q {[13]285/f4f3 9.383} f3+ {[17/33]-573/e2e3 10.098}
51. Ke3 {[16]287/g3b8 9.056} Bxb8 {[18/29]-589/b6b8 10.108}
52. Rxb8 {[13]287/a7c7 0.723} Rc7 {[18/27]-585/c3d4 10.471}
53. Bd4 {[15]289/c7c2 7.44} Rc2 {[17/29]-606/e3e4 10.194}
54. Kxe4 {[15]292/c2a2 8.781} Ra2 {[17/32]-626/d4b6 10.357}
55. Bb6 {[15]285/d7c6 6.638} f2 {[20/28]-630/b8f8 10.368}
56. Rf8 {[16]301/d7c6 10.176} Kc6 {[22/33]-630/f8f5 10.064}
57. Rxf2 {[17]304/a2a4 9.071} Ra3 {[20/28]-658/f2f5 10.558}
58. Rf5 {[17]316/a3a4 14.511} Kb7 {[18/33]-686/f5f6 10.384}
59. Kd5 {[16]357/b7a6 10.602} Ka6 {[19/37]-719/d5c5 10.092}
60. Rf8 {[16]370/a6b5 10.023} Kb5 {[16/32]-755/f8b8 10.112}
61. Rb8 {[16]374/a3h3 8.094} Rf3 {[19/36]-804/b6c7 10.325}
62. Bd4+ {[14]377/b5a6 8.446} Ka6 {[22/33]-8313/d5c4 10.207}
63. Ra8+ {[13]445/a6b5 10.472} Kb7 {[20/40]-1589/a8a7 10.078}
64. Ra7+ {[16]470/b7b8 9.165} 1-0
User avatar
abik
Posts: 823
Joined: Fri Dec 01, 2006 10:46 pm
Location: Mountain View, CA, USA
Full name: Aart Bik

Re: Chess for Android: double blunder between Komodo & S

Post by abik »

Max wrote:Just watched a game between Komodo and Stockfish at my little tournament on Samsung Galaxy Tab with Aart Biks great new Chess for Android 2.9 gui.
Interesting game. Thanks. I use the go movetime <T> command in the GUI (extending time controls is still on my todo list), but I noticed that some engines are indeed "a little creative" with the given time.
Others, like Diablo, always move instantaneously, which explains why it performed badly in the tournament I published earlier.
User avatar
Mike S.
Posts: 1480
Joined: Thu Mar 09, 2006 5:33 am

Re: Chess for Android: double blunder between Komodo & S

Post by Mike S. »

I can reproduce this in Windows XP 32-bit, in two different interfaces with Komodo 1.3, but not with Stockfish 2.1.1. Apparently, Komodo doesn't understand the blind bishop draw so it thinks RxP RxR BxR is won for White.

But Stockfish immediatly has RxR 0.00 here. All without endgame tables, P4/2.4 GHz.

Code: Select all

Hash table size is now 32 meg
Engine: Komodo 1.3 (32 MB)
von Don Dailey, Larry Kaufman
 14.01	 0:01 	+3.01 	57.Txf2 Ta4+ 58.Kd3 Kb5 59.Tf5+ Ka6 60.Ld8 Ta3+ 61.Kc4 Ta4+ 62.Kc5 Ta1 63.Tf6+ Kb7 64.Te6 Tc1+ 65.Kb5 Tb1+ 66.Kc4 Ta1 67.Kb4 Ta2 68.Tb6+ (545.864) 286
(...)
 17.01	 0:05 	+3.05 	57.Txf2 Ta3 58.Tf5 Kb7 59.Tf6 Ta4+ 60.Kd5 Ka6 61.Kc5 Ta2 62.Ld8+ Ka7 63.Kc4 Tc2+ 64.Kb3 Te2 65.Td6 Te5 66.Lb6+ Kb7 67.Kc4 (1.740.586) 318
 17.28	 0:05 	+3.05 	57.Txf2 Ta3 58.Tf5 Kb7 59.Tf6 Ta4+ 60.Kd5 Ka6 61.Kc5 Ta2 62.Ld8+ Ka7 63.Kc4 Tc2+ 64.Kb3 Te2 65.Td6 Te5 66.Lb6+ Kb7 67.Kc4 (1.775.258) 316
Bester Zug: Tf8xf2 Zeit: 0:05.688 min  K/s: 316.480  Knoten: 1.775.258 

Engine: Stockfish 2.1.1 JA (32 MB)
von Tord Romstad, Marco Costalba and Joona
 20/07	 0:00 	 0.00 	57...Txf2 58.Lxf2 Kb5 59.Kd3 Kxa5 (65.859) 191
Bester Zug: Ta2xf2 Zeit: 0:00.375 min  K/s: 191.450  Knoten: 65.859 
Regards, Mike
CRoberson
Posts: 2094
Joined: Mon Mar 13, 2006 2:31 am
Location: North Carolina, USA

Re: Chess for Android: double blunder between Komodo & S

Post by CRoberson »

Hi Aart,

I played Ares on a PC vs Stockfish 2.1 on an Android Tablet with the Nvidia Tegra 2 chip using your interface for the Android. I am porting a new version of Ares to the Android and will put it on the Android market. I have stalled my GUI development to work on other things, so I am planning to recommend your GUI. Thus the test.

In several games Ares beat Stockfish easily in under 30 moves. I had the time set to 30 secs per move. Several of the moves looked questionable especially one where Stockfish blundered into a mate in 4. The same thing happened with Kommodo. Thinking it might be a bug in the GUI, I tried Stockfish in another Android based GUI. This time Stockfish played much stronger.

I was going to send you email about this potential bug, but noticed this thread first. It could very well be a bug in those engines support for that type of TC (Move in Time). I remember I had to put in special code for that TC when I was testing against Steve B and Thorsten C a few years ago. When I get Ares on my Android (hopefully tomorrow or tonight), I will try it under those conditions and let you know if that seems to be the issue.

Charles
User avatar
abik
Posts: 823
Joined: Fri Dec 01, 2006 10:46 pm
Location: Mountain View, CA, USA
Full name: Aart Bik

Re: Chess for Android: double blunder between Komodo & S

Post by abik »

CRoberson wrote:I was going to send you email about this potential bug, but noticed this thread first. It could very well be a bug in those engines support for that type of TC (Move in Time). I remember I had to put in special code for that TC when I was testing against Steve B and Thorsten C a few years ago. When I get Ares on my Android (hopefully tomorrow or tonight), I will try it under those conditions and let you know if that seems to be the issue.
Thanks for your feedback Charles, I appreciate that. I am unsure how this could be a GUI bug, but I am open for suggestions of course. Do you have the PGN? I sometimes noticed that particular engines immediately reply in some situations under the fixed time control. I assumed this was because the reply was obvious, but perhaps a time management bug is responsible.
Albert Silver
Posts: 3026
Joined: Wed Mar 08, 2006 9:57 pm
Location: Rio de Janeiro, Brazil

Re: Chess for Android: double blunder between Komodo & S

Post by Albert Silver »

Is there any way to set an undetermined analysis mode so I can just choose an engine and see its analysis?
"Tactics are the bricks and sticks that make up a game, but positional play is the architectural blueprint."
mcostalba
Posts: 2684
Joined: Sat Jun 14, 2008 9:17 pm

Re: Chess for Android: double blunder between Komodo & S

Post by mcostalba »

abik wrote:
Max wrote:Just watched a game between Komodo and Stockfish at my little tournament on Samsung Galaxy Tab with Aart Biks great new Chess for Android 2.9 gui.
Interesting game. Thanks. I use the go movetime <T> command in the GUI
I'd think this is wrong for a match.

See UCI protocol:

Code: Select all

	* movetime 
		search exactly x mseconds
You may want to send wtime and btime instead:

Code: Select all

	* wtime 
		white has x msec left on the clock
	* btime 
		black has x msec left on the clock
Marco
User avatar
abik
Posts: 823
Joined: Fri Dec 01, 2006 10:46 pm
Location: Mountain View, CA, USA
Full name: Aart Bik

Re: Chess for Android: double blunder between Komodo & S

Post by abik »

mcostalba wrote:
abik wrote:I use the go movetime <T> command in the GUI (extending time controls is still on my todo list)
I'd think this is wrong for a match.
Allow me to respectfully disagree. A fixed time per move may be a bit unusual for a match or tournament, but I don't think it is more or less wrong than any other time control. In fact, it is sometimes useful to take the road less traveled to explore new code paths through an engine (for instance, a few authors discovered that they forgot to implement this part of the UCI protocol altogether as a direct result of these matches).

Chess for Android does not currently support a clock, and even the fixed time per move is not strictly enforced when an engine decides to take more time. As stated in the part of the quote that you left out, extending time controls is on my todo list (as is, for example, implementing infinite analysis). It is a mere hobby project, and I have not worked much on it the past few months (other than just compiling a few engines for Android). I hope to return to the project soon though, assuming it is still relevant :-)
mcostalba
Posts: 2684
Joined: Sat Jun 14, 2008 9:17 pm

Re: Chess for Android: double blunder between Komodo & S

Post by mcostalba »

abik wrote: Allow me to respectfully disagree.
Of course. I was misleaded by the comment of the thread author:
Looked a bit strange to me, that Stockfish always moved after about 10 seconds. While Komodo used his time more intelligent and takes between 1 and 33 seconds.
I didn't understood this is your intended behaviour. Sorry for the noise ;-)

In this case I think is Komodo that perhaps does not honor the "movetime" command...
User avatar
abik
Posts: 823
Joined: Fri Dec 01, 2006 10:46 pm
Location: Mountain View, CA, USA
Full name: Aart Bik

Re: Chess for Android: double blunder between Komodo & S

Post by abik »

mcostalba wrote:In this case I think is Komodo that perhaps does not honor the "movetime" command...
Absolutely, and unfortunately not enforcing the given time in the GUI may lead to unfair match results. Time for me to fire up Eclipse and start working...... :-)