Zappa Mexico II available now

[ernst]

Hello,

A new version of Zappa Mexico called Zappa Mexico II is now available on shredderchess.com. Anthony Cozzie has managed to further improve the playing strength of his world class chess engine that was computer chess world champion in 2005.

This is a free update for all customers of Zappa Mexico. Simply download the program again via the download link in your order confirmation email to get the latest version. The download counts for all customers are reset.

Regards,
Ernst.

[Uri Blass]

Hello,

A new version of Zappa Mexico called Zappa Mexico II is now available on shredderchess.com. Anthony Cozzie has managed to further improve the playing strength of his world class chess engine that was computer chess world champion in 2005.

This is a free update for all customers of Zappa Mexico. Simply download the program again via the download link in your order confirmation email to get the latest version. The download counts for all customers are reset.

Regards,
Ernst.

This version was tested and there is no proved improvement.
It even seems that there is a small reduction in playing strength

http://www.computerchess.org.uk/ccrl/40 ... _length=30

1 Zappa Mexico 64-bit 4CPU 3064 +20 −19 64.8% −97.6 43.7% 863
64.9%
Zappa Mexico II 64-bit 4CPU 3056 +33 −32 58.0% −50.0 47.7% 287

Zappa Mexico 64-bit 2CPU 3035 +32 −31 62.2% −81.1 42.1% 316
Zappa Mexico II 64-bit 2CPU 2996 +101 −94 69.1% −118.2 38.2% 34
78.3%


Zappa Mexico 64-bit 2938 +20 −20 57.3% −47.3 44.4% 784
61.5%
Zappa Mexico II 64-bit 2931 +34 −34 55.1% −32.8 47.3% 264
54.7%

[IWB]

Hello

This version was tested and there is no proved improvement.

Please do not confuse a 'bugfix' of Zappa Mexico which was long ago privately called by the CCRL testers "Zappa Mexico II" with the official Zappa Mexico II which was released a few minutes ago!

Bye
Ingo

PS: To avoid confusion maybe the CCRL can rename their 'Mexico II' to something else (e.g. 'Mexico bugfix')

[ernest]

This version was tested and there is no proved improvement.

Sometimes, Uri, you are awfully quick to pull out your sword!

Please read the question before you answer it...

[Martin Thoresen]

Yes, we in CCRL tested the bugfixed 1st version under the name Zappa Mexico II.

It will not be mistaken with the "real" Zappa Mexico II which has just been released, we will clearly fix this.

[Dr.Wael Deeb]

This version was tested and there is no proved improvement.

Sometimes, Uri, you are awfully quick to pull out your sword!

Please read the question before you answer it...

Uri was simply confused with the CCRL renamed bug fixed version,he didn't do anything wrong....

[Ryan Benitez]

This version was tested and there is no proved improvement.

Sometimes, Uri, you are awfully quick to pull out your sword!

Please read the question before you answer it...

Uri was simply confused with the CCRL renamed bug fixed version,he didn't do anything wrong....

I agree. He made a natural assumption, the situation was just confusing.

[Graham Banks]

Zappa Mexico II Release Notes 1/25/2008

1. Changelist:

added futility/extended futility pruning
a lot of speed optimizations that were mostly eliminated by calling
eval() at every node
singular extensions are now automatically disabled for >8 cores
max hash table size increased to 1TB !
some very small extra eval terms.

Overall not a lot; this is strictly a maintenance upgrade. Should
be similar to the original Zappa Mexico in most positions except
for (hopefully) getting a bigger depth due to extra pruning.
Perhaps +20 elo can be expected.

2. Instant Move

I have gotten some questions about this. Zappa will move instantly
whenever a move is singular, i.e. much better than the others.
It's a feature, not a bug These are not always immediately
obvious recaptures, but if you analyze in multipv mode the
alternatives to the singular move should always be considerably
lower in score (> 0.5 pawns).

3. Strelka

Since this release comes immediately after the release of the
Strelka source code, I assume that there will be a hue and cry from
the unwashed masses about how I am pirating the intellectual
property of Vasik Rajlich. Let me just immediately say that there
are simply no new ideas in Strelka anyway; my shock at reading the
Strelka source code was not generated by the brilliancies contained
therein but by the lack thereof. Looking at Strelka finally
convinced me to do some things that everyone else has done for a
long time (like futility pruning), and to clean up my code a bit
since I had always been developing in "crunch" mode. I did not add
any eval terms from Strelka (my view being that Zappa's eval is way
better than Strelka's anyway) and I also did not do the crazy
futility pruning where it drops into the qsearch at depth 3. In
summary, Strelka was more a source of motivation than chess ideas.
Of course, you will just have to take my word for it until some
enterprising Russian hacker disassembles Zappa

4. Opening Book

I have also gotten some questions about an opening book for Zappa.
My current recommendation is Sedat Canbaz's confidently named
Perfect.ctg for Chessbase GUIs, free from SedatChess.com.

enjoy,

anthony

[Tony Thomas]

Anthony has never been wrong with his elo estimations. Matter of fact, we got better ratings than he expected. If this current version is indeed a 20+ elo improvement, it would cut down the rating difference between Rybka and Zappa on a quad to almost nothing.

[ChessAddict]

Can someone who has zappa mexico II run some engine vs. engine test and see how it performs? I would like to see how much better it is. Thanks.