I've done a lot of bullet tourneys first, then went for blitz with little satisfaction. Not my thing at all. Rapid chess seemed to be the optimum choice for a tester because life is too short against unbearable tournament time controls. Besides, engine updates are too frequent to handle in our era.
Currently, i have come to obtain average scores stabilized between 45% and 55% in Rapidroid. The variation between number of games played by each engine is also very low. All these trends confirm that the phase of construction is completed. From 100 games per engine approximately to 220 games at present, it appears to become only fine tuning with smaller elo fluctuations. The error margins stretch in these phase but the ranking does not change too much.
Before launching the 20th round, my target was fixed as pausing the basic Android stage and continue with other platforms to mix and calibrate everything done so far. I'm really curious to see at what strength would compete the Android engines vs anchor PC engines, IOS, Winmobile programs as well as my retro tabletops.
But something wonderful happened in January: Jim Ablett has returned. Guilty of many engines compiled for Android in the past, he had stopped two years ago. I must admit his comeback has been too severe for Rapidroid. As of today, i can't believe how fast below ranking has become partly obsolete after the compiles he released in three weeks :-)
His highly appreciated assault changed all my plans now. I must continue with Android engines for some more time before invading other platforms, bearing in mind:
* Out of 56 engines rated, only 43 remain up to date.
* 13 engines have been updated. Gauntlets are necessary to replace their elder bros (2600 games expected).
* 14 new engines need to be introduced from scratch (2800 games expected).
Below release does not include any of Jim's latest works. I have just finished the ongoing round and calculated. For sure, the next update will be much more colorful with at least 69 engines.
RAPIDROID 18-JAN-2015 / 6052 GAMES PLAYED BY 56 PROGRAMS
## Name c o/s elo + - gam sco oppo dra
01 Stockfish 5 4 And32 3162 43 41 214 78% 2964 35%
02 Komodo 8 4 And32 3093 39 38 214 67% 2977 43%
03 Critter 1.6a 4 And32 3010 37 37 214 54% 2982 53%
04 Firenzina 2.4.1 xTreme 4 And32 2998 37 37 214 50% 2996 47%
05 BlackMamba 2.0 4 And32 2925 39 39 220 53% 2892 41%
06 RobboLito 0.085e4l 1 And32 2874 40 40 220 48% 2875 34%
07 Senpai 1.0 4 And32 2863 41 41 214 52% 2840 33%
08 Komodo32 3 AB 1 And32 2841 40 40 216 51% 2834 40%
09 Texel 1.05a8 1 And32 2809 41 40 220 58% 2749 27%
10 Gaviota v1.0-d 4 And32 2758 39 39 216 49% 2768 38%
11 Toga II 3.0 1 And32 2699 39 39 218 49% 2706 36%
12 Deuterium v14.3.34.130 1 And32 2641 38 38 240 50% 2639 32%
13 Arasan 15.2 JA 4 And32 2637 39 39 224 47% 2660 34%
14 DiscoCheck 4.3 1 And32 2611 39 39 228 48% 2624 29%
15 IvanHoe 9.46b 4 And32 2609 40 40 228 51% 2599 28%
16 GNU Chess 5.50-32 1 And32 2602 39 39 220 53% 2585 35%
17 Rhetoric 1.4.1 1 And32 2570 39 39 220 48% 2592 31%
18 RedQueen 1.1.3 TCEC JA 4 And32 2509 41 41 216 46% 2545 25%
19 Crafty_23.4.JA_xb 1 And32 2502 40 40 216 51% 2493 25%
20 Rodent 1.00 1 And32 2494 40 40 220 53% 2470 25%
21 Alfil 12.10 1 And32 2451 40 40 212 51% 2443 29%
22 cheng3 1.07 JA 1 And32 2429 41 40 218 53% 2409 23%
23 Rotor 0.7a 1 And32 2429 40 40 212 52% 2418 30%
24 Daydreamer 1.75 JA 1 And32 2427 39 39 220 45% 2469 30%
25 Sloppy_0.23.JA_xb 1 And32 2385 40 40 218 49% 2392 30%
26 Scorpio_2.7.JA_xb 1 And32 2378 40 40 218 52% 2360 26%
27 DanasahZ_0.4.JA_xb 1 And32 2367 40 41 216 44% 2406 25%
28 GarboChess 3 1 And32 2363 40 40 216 44% 2409 27%
29 GNU Chess 6.0.2 1 And32 2352 39 39 216 52% 2338 27%
30 Tucano_1.04.AB_xb 1 And32 2304 41 41 214 53% 2285 22%
31 Pepito v1.59 1 And32 2294 39 39 222 49% 2309 29%
32 GreKo_9.0.JA_uci 1 And32 2275 39 39 222 49% 2285 29%
33 Diablo 0.5.1b JA 1 And32 2266 41 41 216 50% 2262 21%
34 Typhoon_1.0.r358.JA_xb 1 And32 2253 40 40 218 51% 2244 25%
35 BetsabeII_1.30.JA_xb 1 And32 2252 41 41 222 54% 2219 15%
36 Phalanx_XXIII.JA_xb 1 And32 2206 42 42 216 54% 2167 17%
37 Olithink_5.3.2.JA_xb 1 And32 2182 42 42 216 52% 2155 22%
38 Sungorus 1.4 JA 1 And32 2170 42 42 218 49% 2169 21%
39 Natwarlal_0.14.JA_xb 1 And32 2138 42 42 214 51% 2122 19%
40 TJchess 1.1U 1 And32 2118 41 42 212 46% 2147 23%
41 Myrddin_0.86.JA_xb 1 And32 2116 41 41 216 48% 2131 20%
42 Jazz 6.40 JA 1 And32 2114 40 40 216 48% 2122 25%
43 Scidlet_2.61b2.JA_xb 1 And32 2060 43 43 212 53% 2024 20%
44 KmtChess_1.21.JA_xb 1 And32 2060 42 42 212 54% 2026 19%
45 AdroitChess0.4 JA 1 And32 1939 44 44 210 52% 1905 16%
46 BikJump v1.8 1 And32 1885 44 44 210 50% 1873 20%
47 Sjeng_1.12.JA_xb 1 And32 1881 46 46 210 48% 1887 13%
48 Leonidas_r83.JA_xb 1 And32 1803 47 47 210 56% 1726 15%
49 ZCT-0.3.2500 1 And32 1794 48 49 210 50% 1773 10%
50 Sjaak_4.68.JA_xb 1 And32 1751 47 47 210 51% 1716 13%
51 Tscp_1.8.1.AB_xb 1 And32 1584 49 49 210 49% 1574 11%
52 Rocinante 2.0 JA 1 And32 1577 51 50 210 52% 1550 8%
53 Zzzzzz_3.5.1.JA_xb 1 And32 1572 47 47 210 52% 1549 22%
54 VIRUTOR CHESS 1.1.1 1 And32 1412 48 49 210 43% 1475 10%
55 Chess for Android 1 And32 1293 50 52 210 30% 1486 10%
56 Simplex 0.9.8 1 And32 1047 66 76 210 08% 1535 5%
Rapidroid test platform consist of:
* Samsung Galaxy Note II @ 1.6 or 1.7 Ghz x 4 cores + 256MB hash for SP & MP Android programs,
* Polypad 1010IPS tablet @ 1.6 Ghz x 2 cores + 128MB hash for SP Android programs,
* HTC Diamond @ 528Mhz to be used for Windows Mobile programs, with 16MB hash,
* i7 M620 @ 2.67 Ghz + Arena 3.5 + 2GB hash for Windows X64 programs
* iPod Touch 64G @ 600 Mhz and iPhone 5S @ 1.3 Ghz x 2 cores to be used for IOS programs,
* DosBox 1.74 used to run DOS programs,
* WinVICE to run Commodore-64 programs,
* Messtiny UCI adapters or CB-Emu2014 used to emulate Mephisto programs,
* Own books disabled and replaced by 20 ply openings taken from Adam Hair's 10 move book, whenever possible.
* Openings selection for max variety, queens on board, no check or capture at last ply, preferably rated between +0.15 to +0.39 by Stockfish and Komodo.
* Opening positions played twice with different colors, whenever possible,
* Repeating openings and twin games avoided between two programs,
* Tablebases and pondering off,
* Time control: 15+1 Fischer clock or 15 to 30 sec/move or closest possible, identical for both programs.
No comments:
Post a Comment