HAL9000

HAL9000
"It just isn't conceivable that you can design a program strong enough to beat players like me."

April 28, 2015

Rybkamura: The last victim forever?

Komodo 9 and Stockfish new development versions promise already extra 30-40 ELO over last official versions. Who knows whether Nakamura will be the last human courageous enough to take the challenge vs a silicone monster?

Which GM remains to take a bid without odds? Anyone?

Self critic

Yes. I know it's delaying too much. More it delays, more it gets better. But it's not excuse and i can't let anyone criticise but myself :-)

Android UCI engine update: Maverick 1.00 JA reporting correct version

Soon after posting the new version of Maverick a few days ago, i've been discovering that it reports the wrong ID as "0.60 arm".

The games played between the new one and the previous v0.60 at bullet time control have proven a small detail is forgotten. The binary is compiled from v1.00 source indeed.

So i've tried to correct the reported version number directly on the compiled binary with an hex editor, the way we do since centuries.

It worked as expected. The only thing to take good care is to overwrite the error with the exact number of characters, no more, no less. The modified binary is now loading fine and reporting "Maverick 1.00 arm" tothe GUI. Cheers!

Maverick 1.00 aware of its version: HERE

April 25, 2015

Android UCI engine update: Maverick 1.0 JA

Maverick gets stronger with the release of 1.0. The author claims 2500 ELO at CCRL scale. The Android version that you may download HERE should reach 2300 approximately.
Maverick is currently playing in Rapidroid. The new version will be fully tested and included in next release.
P.S.: I've quickly tested 1.0 vs 0.60 at 60"+1" bullet time control using TCEC6 superfinal openings. 1.0 decisively won 50.5-13.5 which makes +229 ELO boost! The only issue was that the new version still identifies itself as 0.60. The author must have forgot to update it in the source.

April 19, 2015

First 3000 ELO world champ?

To compare or not ELO ratings of chess programs vs human ratings is a popular discussion topic in any chess forum. Some suggest separated populations present totally different distributions while some others defend 3000+ engines are really 3000+ because no human can dare to challenge computers since years.

It's obvious that human players faced big difficulties versus current top programs. Take Stockfish or Komodo, they both seem to be stronger than top human GMs. Remember Kasparov vs Deep Junior in 2003, Kramnik vs Deep Fritz in 2002 and lately Nakamura vs Stockfish.

Vishy Anand is also expressing sincere credits to computer programs often in interviews. How to deny the words of a former world champion?

One thing is for sure. Distributions are different in different rankings. Stating that Stockfish can be 3300 may not imply that a 2000 ELO engine selected from the same ranking overcomes a 1800 human player.

That means any 1800 ELO average player can try 1800 to 2200 ELO engines from CCRL's rankings to see if they are that difficult to crush or not.

The speculation will never get a concrete answer unless hundreds of human vs machine games are played and analyzed. I guess that will never occur and discussions will extend.

Regarding current facts to keep in mind:
* Top human Magnus Carlsen could never reach 2900, peaked 2889 only.
* Caruana played around 3300 during first 7 rounds of last year's Sinquefeld Cup.
* No human could ever beat a 3000+ machine in an OTB game in tournament conditions.
* As of today there are only 44 human players over 2700 ELO in the whole planet, only 1 over 2800.

Nodes

This time it's not the nodes per second that counts
but rather the nodes per square meter.

What kind of a Knight?

- You'd better choose your color.
- Nevermind, come lay down beside me.
- I can't. The queen is waiting for me.

Chess in paintings - 001

The Chess Game by Sofonisba Anguissola

April 18, 2015

RAPIDROID: Change of the rule

I'm quite unhappy to delay a new release of Rapidroid nowadays.

There are two reasons of the unexpected fact:
1) Extreme load at work. Classical. Almost no free time left except sundays.
2) Engines terminating games on their own.

The first issue can be somehow bypassed when automated tourneys run without early terminations. As this is linked to the solution of the second issue, i must do something with the bad boys, be it something radical.

Therefore, i've decided to exclude all engines that don't play games until checkmate, regardless of the correctness of their behaviour, of course except for legal resigns which remain allowed.

This change will trim down the engine fleet but i have no other choice to ensure the whole experiment progresses as scheduled.

The first victim is Sjaak 1.0 which caused me a lot of replays. I tried UCI and XB on all three processors i'm using but found no way to get it playing reliably. I don't see why it stops when the eval is just +2.00 something.

The rule now changes to: Only the engines which play until checkmate can be selected to play in Rapidroid. Exceptional early terminations are limited to maximum 5% of whole games played by the engine. Beyond 5%, the engine will be deleted from the database and the ranking. Above rule doesn't apply to resigns complying with the protocol being used.

New Android UCI engine: Sugar 050415 JA

Sugar is a Stockfish derivative which had been subject to cloning discussions in chess forums since its first release.

After the author released its source code, the opposing community seemed somehow calmed down and further releases came in parallel to Stockfish development.

My personal opinion here is that such parallel developments so close to the original, be it legal or not, lack the taste of alternative colors.

Some state Sugar is even stronger than Stockfish but i don't take this to serious unless Sugar finds its own way to overtake the fish considerably. I don't think it's possible to overcome an engine progressing with the help of thousand of cores testing countless patches continously over a wide network.

Anyway it's everyone's free choice to try and use Sugar on Android. You can just get it: HERE

Android UCI engine update: Sting 5 JA

JA stated tihs version is for analysis mode only. Enjoy: HERE

Rapidroid in the car: Quad core on quad wheel

It's hard to concentrate on your personal hobbies when the job takes all your juice. No time left to anything else other than swallow the dinner quickly and go sleep. In such situation, you should give  priority to your profession. This is exactly what happens to me often.

Yep, alright, understood. But couldn't it become possible to make some of the things run automated in parallel? Maybe, barely.

This is what i've done lately as seen in the photoshot. Two tablets are triggered to run two tourneys simultaneously on the floor of the car. Another one not seen in the frame runs in the front to complete the mobile fleet.
Time saving is guaranteed, no doubt. And if battery issues cause problems, there's a 10800 mAH external battery pack to help. Of course, tablets with micro-USB charge connectors are the winners here.

Then, what song to play on the deck? I guess "Make Your Move" from Rainbow is perfect.

April 15, 2015

Android UCI engine update: Bobcat 6.4b JA

Right after the first port to Android through v3.25, Bobcat is now updated to its latest version 6.4b by Jim Ablett. It's quite a strong engine close to the top and i hope the latest one will bring more competition to Rapidroid.

I'd just started introducing v3.25 with gauntlets. It seems it's time to stop that and switch to 6.4.

Download Bobcat: HERE 

April 12, 2015

Possible to add 150 ELO to Deep Saros 2.3f for Android?

When it comes to naming engine parameters in programming phase, practice and theory often provoke conflicts. Xboard and UCI standards are quite old when compared to the speed the programming techniques improved since their specifications were released.

I won't repeat the endless discussions about how we must call the protocol option that sets the number of parallel searches will use the engine. It has never come to a common agreement. Engine programmers still use different namings that not all GUI's interpret correctly.

The last case i had encountered was almost one year ago with Gaviota for Android where Miguel Balicora used Cores that Chess for Android was prepared for but Droidfish was not. The same binary could run on quad cores when loaded into CfA but when loaded into Droidfish it was delivering one third of the kNps expected. Just because of one word in the source code.

I'd come to a double solution with kind help from both authors. Miguel compiled a version which sends Threads command instead of Cores (version 1.0d) while Peter Österlund added the option to understand Cores command into Droidfish.

Today i've noticed Deep Saros 2.3f may be suffering from the same issue because it uses none of the above namings but "cpus". I'm not sure if multi threaded search will run if Chess for Android understands it. Porting SMP functionaly to Android seems tricky sometimes but it's worth giving a try.

One solution is to set the GUI to understand "cpus" command. Another easier solution, just the same as we'd found with Gaviota, is to change the command in the source code from "cpus" to "threads" so that the GUI's can use it.

I think there's a chance to add about 150 ELO to Deep Saros 2.3f and push it above 3000 if it's only a naming issue which prevents using all the cores. I hope Jim Ablett will give a hand on the question.

P.S.: In case you don't agree with +150 ELO, you are absolutely right. The reason for me to go wrong deliberately was only to attract attention to the subject. Now the truth is that if Deep Saros coıuld go 4 threads from 1, it would gain roughly: log (3 to 4 times more knps) / log2 x 60 =  95 to 120 ELO more.

Android UCI engine update: CPW 1.1.11 JA fixed

I'd banned CPW from Rapidroid because it used to keep "ponder on" all the time whatever was the GUI setting. This fixed version promises to heal the issue. I'm happy i can push CPW to the battlefield again given that compatibility wise i'd seen no issues at all other than that ponder problem.

Download fixed CPW 1.1 JA: HERE

P.S.: Just tried the benchmark test on dual core RK3066. Unfortunately the issue repeats. The engine starts to ponder as soon as loaded, even at initial position. I see one core active before first move. Importing another engine frees the cpu. Again, CPUstats is a great tool to verify core usage issues.

Android XB engine update: Micromax 4.8W JA

Following Fairymax, JA provided a compile of its brother engine Micromax, intended to become one of the smallest engines in code size. They are both written by H.G. Müller from Germany.

In Rapidroid games, i used to run an old compile of the same 4.8w version, provided by Aart Bik in Rapidroid. Interesting thing is that AB compile is only 18K while JA is 225K! I guess it's related to nothing but the different tools used to obtain the Android build.

Anyway, what we should care about is the outcome, not the size. I plan to organise a head to head match between to see what difference do they show in practice.

Download JA compile: HERE
Download AB compile: HERE

New Android XB engine: Hoichess 0.12.1 JA

Hoichess is an engine, i've first heard of after the Android compile has been posted by Jim Ablett. It's written by Holger Ruckdeschel from Germany.

According to CCRL, v0.10.3 64 bit performs at 1790 ELO on a PC at 40m/4' time control. On Android, one must expect roughly 1600.

This version 0.12.1 is newer and unknown to CCRL. Therefore, we're getting a good chance to overcome the PC guys :-) HERE

Those interested in the older version 0.10.3 rated by CCRL are not forgotten. They may download it: HERE

Android XB engine update: GK 0.90 JA fixed

I've just noticed i'd forgot to include GK in the engine collection repository at box.com. Thanks to the release of a fixed version by Jim Ablett, i take the opportunity to add GK there.

GK already played Ficher clock games only in Rapidroid and had 1996 ELO so far. As the fixed version is supposed to play fixed time per move too, a wider range of opponents is now possible to meet and that should ensure more games per round for better ELO accuracy. Remains to be tested and used.

The new version will not produce log files anymore, which is also preferable.

You may download the latest build of GK for Android: HERE


April 9, 2015

Shortest broadcast chess news :-)

* TCEC-8 may delay for months because M.Thoresen got a new job,
* Komodo 9 and Hiarcs 15 on the way,
* WCC by ICGA soon in summer,
* Nakamura broke 2800 barrier momentarily, Carlsen still on top,
* Only 44 human players above 2700!

Android XB engine update: FairyMax 4.8V JA

I don't know what serious additions 4.8V brings on 4.8L but an update is an update. So, HERE's the newest version for those interested.

Don't forget to install fmax.ini file too. Without it Fairy won't work at all.

Regarding RAPIDROID, replacing 4.8L with 4.8V by replaying all previous games is not my priority for the moment, unless i discover that the latest one plays "fixed time per move" games without problem.

4.8L is currently playing Fischer clock only in RAPIDROID, which reduces the number of samples per round.

April 8, 2015

New Android UCI engine: Bobcat 3.25 JA

Bobcat is written by Gunnar Harms from Netherlands. It's not a famous engine like Komodo but it plays quite strong, above average. It has a CCRL rating of 2727 which makes it interesting (for me at least).

Bobcat has the SMP functionality, so you can use multi threads with it and try to melt down your processor. When i tried with 1, 2 and 4 cores, i've seen way too close kNps values like 196, 212 and 226, suggesting a bad multi core management or no SMP functionality at all. Hopefully the search depths were different enough to confirm there's no problem. It may also be that the kNps reporting does not reflect the correct values but who cares anyway if the engine plays ok?

Download Bobcat: HERE

Android UCI engine update: Maverick 0.60 recompiled from correct source

The Android port of Maverick 0.60 posted previously is recompiled by Jim Ablett due to an error which caused it to contain v0.51 code.

The version based on the correct 0.60 source can be downloaded: HERE

Although the download link given above remains the same, the file has already changed. This means the initial post will link to the corrected file too.

Android UCI/XB engine update: GNU Chess 6.2.1 JA

Two days ago, i had discovered a compile of GNU v5.60 thanks to Jim's remark and i'd instantly posted it here.

Next to it, Jim kindly released another compile he made from GNU v6.2.1.

Please note that these two branches have nothing in common and you can even call them two different engines. In brief, GNU6 is not a update of GNU5 at all, but a totally different code initiated by Fabien Letouzey, the author of Fruit, forked from Fruit 2.1.

The interesting point is that GNU5 is still active on its own, maintained by another team and performing much better than GNU6.

Regarding GNU6 under Android, i used to suffer a lot with previous 6.0.2 compile of GNU which randomly terminates games on its own. Even if it proves to be correct after the post analysis, that kind of termination is something i hate as a tester because Chess for Android assumes it's a loss, which is not correct at all in most cases. It's a disaster when an engine ends a game while winning clearly but the GUI decides that it has resigned.

My hope this time is that 6.2.1 plays until checkmate and becomes a new member of my future experiments.

Download GNU 6.2.1 Xboard: HERE

P.S.: There's the usual uci compile which followed XB version, provided by JA again. UCI is to be prefered since XB versions don't allow using more hash than default one.

Download GNU 6.2.1 UCI: HERE

New Android UCI engine: Strelka 5 JA

I think this one will rock our smartphones and tablets.

Strelka is one of those countless strong engines spread from Russia, widely blamed for cloning issues here and there. Repeating a lot of statements and claims makes no sense here. Instead, let's just keep in mind that Strelka has incorporated ideas from Rybka 3, Ippolit, Houdini, Critter and Stockfish, as well as own innovations according to its author Jury Osipov. So be it.

Knowing Strelka 5.1 can hit 3139 ELO on a PC is enough to predict how seriously it can compete with current toppers. We may easily expect 2900+ from v5 compiled by Jim Ablett.

To my side, i hope Strelka will play fine automated tournaments without issues, so that i can't wait for pushing it inside Rapidroid. Another 2900+ engine certainly add more excitement to the top.
Download Strelka for Android: HERE

April 5, 2015

Fighting with cpu throttling

Processors never like chess engines because they want to suck out every bit of the silicone power provided.
Cpu manufacturers build and promote their chips for maximum frequencies they can reach.
The problem is how long they can remain fully loaded at max speed. Almost every processor available today uses strategies to avoid melting down, mostly by shifting load between cores and/or lowering freqs, separately or together.
We as testers, just can't accept that.  We need advertised speeds all the time. We are mean!
In contrast to regular public opinion chess is much more demanding than Call of Duty or Need for Speed and none of the specific cpu stress utilities can force our processors more than Stockfish 6 running for minutes on all cores. Try and see with your own eyes how hot will become your tablet so easily.
In Rapidroid experiment my three trustable weapons against cpu thtottling are:
1) A refrigerator at 12C max. He he :-)
2) Cputemp in Statusbar: A nice utility to monitor the temp without having to touch anything,
3) Cpustats: A perfect tool to monitor the freqs per each core on the statusbar, again without touching anything.
This is the only formula to be sure every tournament starts at max speed and continues until the end without power loss. In case the cpu shows up in throttle state,  i can at least detect the unexpected and replay impacted games, even the whole tourney is repeated when necessary.
Exynos 4412 is fully compatible with above tools while cheap Rockchips can't show the temps. They don't have the sensor for that. When they are forced too much, they simply freeze but it's even better because there are no games played in throttle state and i can just start over.
Another trick I recently verified is to run the processor at half speed but with double time control. I know it's slower but it's safe anyway and the total nodes per move are very similar.
Thus, if you test multi core engines vs other multi core ones, you should absolutely use efficient monitoring tools to be sure occasional abnomalies or degradations won't make your experiments suffer.

Android UCI engine update: GNU 5.60

I must have missed somehow that more recent version of GNU. Since months i used to run 5.50 in Rapidroid as submitted in the updated ranking yesterday.
Thanks to Jim Ablett, i woke up to retrieve the latest version 5.60 compiled for Android in the Toga downloads page. Looks like they compiled it in November-2014!!!
Better be late than outdated...
Download GNU 5.60: HERE

New Android UCI engine: Gunborg 1.0 JA

Gunborg 1.0 was released in July 2014 for Windows. Months after that release, it's finally ported to Android by Jim Ablett.

Gunborg is written by Torbjörn Nilsson from Sweeden. Its 64 bit Windows version is rated 1862 ELO by CCRL at 40/4 time control. I guess Android version may reach 1700, a possible challenge for a good human player.

You may download Gunborg for Android: HERE

Note that you have to install the specific Polyglot adapter to be seen by the GUI. The main engine is used by the adapter and must be installed separately. The ini file which must be also installed in the same folder serves as a link between the adapter and the main engine. Do NOT rename the ini file because its name is coded inside the adapter.

PS: A quick set of test games revealed Gunborg uses very low time in fixed time per move mode. Regarding Fischer clock, it behaved unstable in tournament mode, playing instant moves with black after an accurate play with white. Gunborg doesn't look eligible for Rapidroid yet.

RAPIDROID RELOADED: April-2015

The promise is kept on time. So, you may come taste the mobile insanity that nobody else has gone into but me. Despite neverending technical issues which caused me a lot of deleted, replayed, adjudicated games and interrupted tourneys, i like the way things progress. Ugh n' buff!

The loss of tempo may be blamed on countless new engines and versions which kept appearing in march. I prefered to introduce them into the pool on the run, rather than rigorously testing separately before introduction. As expected, some of them bounced back, causing headache.
After all, thanks to a debug and replay session done at the end of Round-3, the resulting PGN database is clean, free of termination errors, therefore good for ELO calculation.

Changes were:
* 104 engines increased to 120, only Gunborg 1.0 not yet evaluated,
* CPW Engine excluded because of "ponder always on" issue,
* Komodo 3 stopped after Doch and Beekay came in, enough Komodos...
* First non-Android member Colossus 4.0 C64 introduced with 2 versions (:-))
* Rockchip 3188 introduced as new processor capable to simulate Exynos 4412.

To do's are:
* Introduce Hiarcs on iOS5,
* Introduce Shredder Android,
* Introduce Chess Genius Android,
* Introduce 2 new RK3188 devices, for 5 Androids in total.

Targets for next release expected in May-2015 are:
* Error margins below +/-60 ELO from 89,
* Number of games standard deviation reduced to 16% from %20,
* Standard deviation of score% reduced to 9% from %13,
* Average ELO change +/-50 from 87.

As seen above, accuracy indicators are not healthy yet. Quite normal... They will need about 100 games per engine to start speaking seriously.

If one states 3300+ is too exaggerated for Android, i'd admit it could be true for the moment, in this very early steps. In parallel to increased number of samples, i expect the list will get compressed, the way it happened in previous Rapidroid experiment.

BAYES RATINGS AFTER 3112 GAMES PLAYED BY 120 PROGRAMS
Rnk Name                     O/S T  Elo   +   - ga sco oppo dra
  1 Stockfish 6              A32 4 3339  91  82 54 82% 3099 31%
  2 Komodo 8                 A32 4 3305  88  81 54 77% 3103 31%
  3 Firenzina 2.4.1 xTreme   A32 4 3176  78  76 54 58% 3117 39%
  4 Sting SF 4.8.4 JA        A32 4 3136  80  78 54 60% 3055 31%
  5 BlackMamba 2.0           A32 4 3123  74  74 54 50% 3123 52%
  6 Critter 1.6a             A32 4 3122  74  74 54 50% 3123 48%
  7 BeeKay                   A32 4 3118  75  76 54 49% 3123 46%
  8 Texel 1.05               A32 4 3092  76  77 54 45% 3126 39%
  9 Komodo32 3 AB            A32 1 3041  92  91 36 53% 3024 50%
 10 cheng4 0.38              A32 4 3006  81  80 54 56% 2970 30%
 11 RobboLito 0.085e4l       A32 1 2998  79  76 56 61% 2925 39%
 12 DeepSaros ver.2.3f       A32 1 2976  80  80 54 51% 2973 28%
 13 Senpai 1.0               A32 4 2926  76  76 56 50% 2933 43%
 14 Arasan 17.5              A32 4 2905  76  75 56 55% 2871 32%
 15 Gaviota v1.0-d           A32 4 2904  78  76 54 59% 2831 37%
 16 Fruit reloaded 2.1       A32 1 2824  80  85 56 30% 2976 29%
 17 Grapefruit 1.0           A32 4 2822  78  78 54 48% 2840 30%
 18 Cyclone 3.4              A32 1 2818  75  78 56 33% 2944 48%
 19 Doch32 1.3.4 JA          A32 1 2815  73  75 56 40% 2881 41%
 20 Toga II 3.0              A32 1 2802  78  79 54 44% 2842 30%
 21 Deep Saros 0.9           A32 4 2751  82  83 54 44% 2784 22%
 22 DiscoCheck 5.2.1         A32 1 2751  84  81 50 63% 2663 30%
 23 Hakkapeliitta dev 120115 A32 1 2730  82  82 52 48% 2741 27%
 24 Deuterium v14.3.34.130   A32 1 2699  80  79 50 55% 2669 34%
 25 Murka 3 ARM UCI          A32 1 2698  77  77 54 50% 2687 41%
 26 IvanHoe 9.46b            A32 4 2693  80  81 52 46% 2720 31%
 27 Scorpio_2.7.7.JA_xb      A32 1 2691  91  85 52 74% 2497 25%
 28 RedQueen 1.1.97          A32 4 2689  80  79 52 57% 2639 33%
 29 Rhetoric 1.4.1           A32 1 2634  81  79 54 58% 2563 28%
 30 Rodent 1.7 build 1       A32 1 2618  80  81 52 44% 2647 31%
 31 Crafty_24.1.JA_xb        A32 1 2601  84  83 54 54% 2566 15%
 32 Gull 1.2 JA              A32 1 2594  78  80 52 39% 2650 37%
 33 GNU Chess 5.50-32        A32 1 2563  83  81 54 57% 2509 22%
 34 TheMadPrune 1.7.04       A32 4 2540  95  94 42 51% 2530 12%
 35 Tucano_5.00.JA_xb        A32 1 2512  83  80 52 62% 2427 31%
 36 Bison 15.1               A32 1 2468  82  84 54 43% 2519 19%
 37 Alfil 12.10              A32 1 2468  81  80 58 53% 2437 17%
 38 Rotor 0.8                A32 1 2431  80  82 56 41% 2499 21%
 39 Daydreamer 1.75 JA       A32 1 2429  77  77 58 49% 2441 29%
 40 Dirty_030411.JA_xb       A32 1 2372  77  78 54 47% 2394 31%
 41 Vajolet 2.03             A32 1 2372  75  76 58 47% 2393 26%
 42 Danasah_5.07.JA_xb       A32 1 2353  77  78 54 44% 2396 33%
 43 BBChess 1.3b JA          A32 4 2349  84  90 54 28% 2532 19%
 44 Octochess arm r.6556     A32 1 2335  75  75 56 50% 2340 32%
 45 Phalanx_XXIV.JA_xb       A32 1 2322  82  80 54 58% 2260 20%
 46 Sloppy_0.23.JA_xb        A32 1 2319  78  79 52 46% 2350 31%
 47 Pepito v1.59             A32 1 2309  79  80 56 46% 2343 20%
 48 GarboChess 3             A32 1 2302  82  81 56 56% 2255 20%
 49 GreKo_12.5.JA_xb         A32 1 2299  81  80 54 54% 2263 22%
 50 DanasahZ_0.4.JA_xb       A32 1 2283  82  82 50 49% 2285 30%
 51 Glaurung Mainz           A32 1 2236 111 125 36 25% 2458  6%
 52 Ifrit_M18.JA_uci2xb      A32 1 2232  86  88 52 40% 2316 15%
 53 Jazz Orchestra 8.40      A32 4 2228  89  84 56 71% 2032 23%
 54 BetsabeII_1.47.JA_xb     A32 1 2215  79  77 56 61% 2143 25%
 55 Typhoon_1.0.r358.JA_xb   A32 1 2172  76  75 56 54% 2148 32%
 56 Amy_0.8.JA_xb            A32 1 2166  79  78 60 54% 2133 15%
 57 Diablo 0.5.1b JA         A32 1 2155  77  75 62 57% 2098 21%
 58 Pawny_1.0.JA_uci2xb      A32 1 2136  76  76 60 49% 2136 22%
 59 Sungorus 1.4 JA          A32 1 2119  77  76 60 58% 2061 22%
 60 Natwarlal_0.14.JA_xb     A32 1 2114  80  80 54 50% 2111 22%
 61 Simplex 0.9.8            A32 1 2083  78  78 62 53% 2033 13%
 62 MangoPaola_1.1.JA_xb     A32 1 2079  80  78 58 58% 2019 16%
 63 Olithink_5.3.2.JA_xb     A32 1 2076  77  77 60 52% 2065 17%
 64 Myrddin_0.86.JA_xb       A32 1 2074  84  81 54 62% 1981 20%
 65 TJchess 1.1U             A32 1 2070  76  77 60 47% 2069 27%
 66 Rattate_Nosferatu.JA_xb  A32 1 2062  78  80 62 44% 2106 11%
 67 Maverick 0.60 arm        A32 1 2051  75  75 62 48% 2063 23%
 68 NGplay_9.86.JA_xb        A32 1 2041  85  84 48 52% 2031 29%
 69 Scidlet_2.61b2.JA_xb     A32 1 2023  82  81 54 55% 1986 20%
 70 Ges_1.34.JA_xb           A32 1 2005 116 105 34 72% 1841 15%
 71 Gk_0.90.JA_xb            A32 1 1996 122 121 26 56% 1914 19%
 72 KmtChess_1.21.JA_xb      A32 1 1983  86  86 46 54% 1950 30%
 73 Resp_0.19.JA_xb          A32 1 1978  79  79 58 50% 1958 21%
 74 DanChess_1.04.JA_xb      A32 1 1977  80  84 60 38% 2072 10%
 75 Kurt 0.9.2.2 JA          A32 1 1909  80  81 58 47% 1921 17%
 76 Sjaak II 1.0             A32 1 1870  88  83 60 74% 1650 18%
 77 Samchess_JA_xb           A32 1 1868  88  91 46 39% 1954 17%
 78 robocide                 A32 1 1865  83  81 62 60% 1778 13%
 79 Faile_1.44.JA_xb         A32 1 1864  92  88 44 60% 1790 16%
 80 Romichess_p3k.JA_xb      A32 1 1860  84  90 54 31% 1995 19%
 81 Surprise_4.3.b13.JA_xb   A32 1 1851 126 117 30 65% 1709  1%
 82 Fairymax_4.8L.AB_xb      A32 1 1834 104 107 32 39% 1928 28%
 83 Knightcap_3.7F.JA_xb     A32 1 1834  92  95 42 42% 1903 21%
 84 Witz_Alpha21.JA_xb       A32 1 1823  91  88 48 64% 1713 19%
 85 ZCT-0.3.2500             A32 1 1820  81  82 58 46% 1859  9%
 86 BikJump v1.8             A32 1 1805  78  80 58 43% 1860 17%
 87 Woodpecker_2.11.JA_xb    A32 1 1774  89  92 44 40% 1850 20%
 88 Clubfoot r2293 JA        A32 1 1771  94 100 46 34% 1886 20%
 89 Cilian_4.14.JA_xb        A32 1 1751  77  77 58 47% 1777 24%
 90 AdroitChess0.4 JA        A32 1 1751  80  82 58 44% 1791 16%
 91 Sjeng_1.12.JA_xb         A32 1 1742  91  90 48 56% 1682 17%
 92 Leonidas_r83.JA_xb       A32 1 1742  87  89 46 45% 1777 15%
 93 Ecce rev. 508            A32 1 1736  81  82 58 45% 1778  7%
 94 smash                    A32 1 1731  85  82 62 64% 1593 11%
 95 Umax_4.8W_xb             A32 1 1715  96 103 36 32% 1856 25%
 96 Colchess_8.0.JA_xb       A32 1 1686  85  86 50 47% 1706 22%
 97 Bismark_1.1              A32 1 1651  83  80 62 65% 1523 19%
 98 Claudia v. 0.5           A32 1 1556 113 120 30 37% 1655 20%
 99 Sayuri 2015.03.02        A32 4 1533  83  83 60 51% 1534 15%
100 Zzzzzz_3.5.1.JA_xb       A32 1 1532  84  85 50 47% 1553 30%
101 zurichess                A32 1 1519 110 103 36 72% 1312 28%
102 Jester_0.84.JA_xb        A32 1 1485  93  90 48 67% 1315 25%
103 Tscp_1.8.1.AB_xb         A32 1 1482  90  91 48 51% 1452 23%
104 Rocinante 2.0 JA         A32 1 1476  76  76 66 58% 1391 27%
105 Chenard_2014.07.11.JA_xb A32 1 1445  97  96 48 58% 1350 17%
106 Kitteneitor_060513.JA_xb A32 1 1443  84  84 50 49% 1433 38%
107 Pulse 1.5-cpp            A32 1 1414  72  72 70 60% 1303 37%
108 VIRUTOR CHESS 1.1.1      A32 1 1375  81  78 68 68% 1188 22%
109 Colossus 4.0 100X        C64 1 1370 287 249  6 83% 1058 33%
110 Chess for Android        A32 1 1325  82  80 68 63% 1192 18%
111 Carnivor_1.0.JA_xb       A32 1 1260  76  78 68 46% 1285 31%
112 Evolchess_0.6.JA_xb      A32 1 1241  88  93 50 35% 1354 34%
113 Trappy_Beowulf_2.0.JA_xb A32 1 1181  90  94 66 35% 1340  9%
114 Toledo_Nano_2010.JA_xb   A32 1 1147 117 125 36 24% 1457 19%
115 K2 v.071                 A32 1 1105 125 136 36 18% 1513 19%
116 Colossus 4.0             C64 1 1044 278 236  6 50% 1058  0%
117 Byak 8.10.14.JA          A32 1  987 105 121 44 19% 1264  7%
118 Xadreco_5.7.JA_xb        A32 1  937 117 136 40 14% 1277  8%
119 Joanna_2002.JA_xb        A32 1  858 126  98 48 10% 1286  8%
120 OliveChess 0.2.7         A32 1  772 145  20 52  5% 1290  6%


Rapidroid test platform:
* GT-N7100 4*1.7Ghz, 256MB hash: All Android progs,
* Codegen Novatab 4*1.4Ghz, 256MB hash: All Android progs,
* Polypad 1010IPS 2*1.61Ghz, 128MB hash: Android progs single core,
* HTC Diam 528Mhz, 16MB hash: Windows Mobile,
* i7 M620 2*2.67 Ghz, Arena 3.5, 2GB hash: Windows 64
* iPhone5S A7 2*1.3Ghz: iOS programs
* DosBox 1.74: DOS programs,
* WinVICE 2.24: Commodore-64,
* Messtiny UCI adapters or CB-Emu2014: Mephisto programs,
* Openings: 20 ply by A.Hair or 16 ply from TCEC, no Q exchange, +0.15 to +0.40 eval by Stockfish & Komodo @ D20, played twice both colors,
* Repeating openings and twin games not allowed between two programs,
* Tablebases and pondering off,
* Time control: 10" to 30"/move or 600+0 to 1800+5 or closest.


April 3, 2015

2 more devices to boost Rapidroid even further

I really liked that quad core Rockchip 3188 for its acceptable Exynos-like performance in exchange of a very low cost. Therefore, i've made ready two more 100 bucks tablets with 2GB RAM and 16GB storage, all hidden under a very bad low resolution TFT screen, way too horrible to look at, just enough to handle engine tournaments.

The small chess fleet is finally composed of:
* 1 Exynos 4412 @ 4 x 1.7 Ghz
* 3 RK3188 @ 4 x 1.4 Ghz
* 1 RK3066 @ 2 x 1.6 Ghz

This is not a dream team indeed but an humble parallel testing network which costs less than the price of a Galaxy S6.

Open source engines ready to compile for Android

Let me help Jim Ablett provide us with some more new engines to run on our Android devices. The following ones are open source C or C++ code, so they can be ported too:

Engines never ported to Android are:
* Chesser
* Conqueror
* Deep Smash
* Dreamer
* Eia
* Exacto
* ExCE
* Galjoen
* Gambit (Gupta)
* Gibbon
* Gunborg
* Madlenka
* Silchess
* Superpawn
* Supra

Updated versions of existing compiles are:
* GNU 5.60
* GNU 6.2.1
* Gull 3 (CERTAINLY A MUST HAVE!)
* Myrddin 0.87
* Sayuri 2005.03.27

Previously ported with serious issues:
* Jazz 8.40 (unstable terminations with Fischer clock)
* CPW (Ponder always ON)

RAPIDROID very soon: 3000+ games by 117 programs

Yes, it's coming. At last, probably by the end of this week.

Latest engine to take part was Kurt, which seemed healty enough. If only it would use full allotted fixed time per move instead of too safe 2/3... Hopefully it plays Fischer clocked games very well, to compensate any slight loss.

I'm about to finish 3 rounds of 11 groups of 10 to 12 engines each. That has become way too much pain to overcome because of many abnormal terminations to analyze and decide between banning or not new engines and/or versions. Gauntlets never ended since weeks thanks to Jim Ablett's compiles.

I must state a new kind of issue has arised for me lately and it costed a lot of games already played by CPW-Engine, a newcomer which seemed perfect for tournament play at first sight. Unfortunately i was tooooo late to notice CPW keeps pondering ON all the time. As this is against Rapidroid conditions, i had to delete that engine and all related games.

If i predict Stockfish 6 on top, closely chased by Komodo8 that should not be surprising at all. I guess who's gonna come 3rd draws more attention than this.

Another good news is the mighty retro-wonder Colossus C64 which will be showing up as the first non-Android engine in Rapidroid with two versions, one at stock speed and another at 100X speed. This is a nice double channel trial to conclude how much extra ELO comes with extreme hardware.

Stay tuned!