HAL9000

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

October 6, 2015

Android UCI engine update: Stockfish Dev 2015-09-10 (towards Stockfish 7)

Each hot championship is a source for new public demands. Since TCEC-8 has started last month, chess freaks didn't stop asking for a new release of Stockfish. Version 6 was released months ago and development version have already progressed a lot, although it's not yet considered enough to release an official update.

On the other side on the medallion, Komodo 9.2 has already fulfilled the expectations so far during TCEC-8. The reptile currently leads the Stage-2 ahead of the fish, somehow making me smile because i still remember the speculations stating that TCEC had to be delayed by the organisers just to make sure a recent version of Komodo stronger enough to cope with the latest Stockfish version becomes available.

Be it true or not, the tournament is in place, running at full speed and i'm not too concerned about possible commercial timings and manipulations. Both teams deserve respect, no matter the status, commercial or open source.

Computer chess can't progress without competition after all. Besides, i wish there could be more candidates to chase the leaders. Unfortunately there's only two engines and all others following far behind. Isn't it sad to see Houdini 4 still playing for 3rd place after so long time?

Regarding Android side of the two musqueeters on top, Komodo 9.2 is already available while Stockfish-6 was the latest among the fishes, until i heard of an alternative development version compiled by Peter Ă–sterlund, the author of Droidfish and his own engine Texel.

I see no reason for that development version play weaker than v6, given that Peter's previous compiles performed very well and that it's based on arm7 or 8 already.

I will schedule a head-to-head match between that dev't version and v6 to see what's the gap as soon as i can. Rapidroid ranking will wait for an official release anyway, regardless of the result.

Meanwhile you may download the package which includes all possible versions compiled and test for yourself: HERE


3 comments:

Anonymous said...

hi. tnx for ur interest of 070915 android,so i didnt make a match but
i have compared the droidfish 6 vs 070915 on initial position the kn/speed, and i have found this: on an 1cpu arm a9 budget phone
droidfish 6 120 kn/sec 070915 108 kn/sec average from several tests
with 2 mins infinite anal.
so i dont know why shows less kn/s but yes only a match will reveal
the true difference.btw the latest sf dv master regression showed 44 elo points over sf 6.i will wait your results ( my bet +30 elo).
(from an fanatic android chess-stockfish user)

Gurcan Uckardes said...

Don't take small kNPs differences too seriously. Even the line analyzed at first can change the numbers. It can start by thinking on e4 then switch to d4 or in antoher test it can start with e4 and stick on it until the end. The nodes may be different because not the same path at all.
The best is to arrange a match and continue until the calculated gap decreases below the error margin. I think in this case, some hundreds of games will be necessary to conclude.

Gurcan Uckardes said...

A small correction: I've had to rename to post title and the binary zip to 150910. It seems the correct date of the source is 10-Sep-15 and not 07-Sep-15, as per the datestamp of the binaries as well as the engine ID that shows up when imported.