Hi!,

A new chess engines rating list site just completed the infrastructure. It is a fast blitz rating site at 1 minute + 1 second increment time control with tournaments running in a single core CPUs and Windows XP 32bit operating system (virtualized).

The goal of the rating site is to have more statistical information together with the other popular rating sites such CCRL, CEGT, IPON, LightSpeed, etc. This may provide data for people who need to know which engines are the best in the categories of ultra fast, fast, average and long tournaments. In my observation, Houdini 3 Pro by Robert Houdart is the best in all these categories and the engines with low KN/sec get better ranking in longer tournaments which incidentally also increase the draws.

The new rating list site also includes old engines, superseded versions and maybe controversial engines to get some data on the status of the chess engines art and science, whether the last few years really made huge advances and by how much progress.

You may visit the site at http://chessowl.blogspot.com

Thank you.
drj

Owl Computer Chess Engines Blitz

Hello,

In an exploratory match, Fire 2.2 xTreme beat Stockfish 2.3.1 with a 69% score in a 200 games match with both engines re-compiled using Visual Studio 2010 compiler and the same build process. Fire is currently being revived by Firenzina, the engine to watch.

The "most powerful open source chess engine" crown may no longer be with Stockfish.

Results at: http://chessowl.blogspot.com/2013/03/showdown-strongest-open-source-computer.html

Throwing more wood into the Fire 2.2 xTreme

I'm attempting a 400-game match between Fire 2.2 xTreme (NS Compile, x64) vs Stockfish 2.3.1 (Default)

5+3 Time Control, 1 CPU, 256MB Hash, No TB's GM_Polyglot.bin book.

Castle00 wrote:Throwing more wood into the Fire 2.2 xTreme

I'm attempting a 400-game match between Fire 2.2 xTreme (NS Compile, x64) vs Stockfish 2.3.1 (Default)

5+3 Time Control, 1 CPU, 256MB Hash, No TB's GM_Polyglot.bin book.

This should be a good experiment, where compile differences will kick in! Fire 2.2 xTreme was "under-compiled" to begin with, it seems. Despite that, the outcome is difficult to predict.

Chessowl was using the w32 version while I am using x64. Both read "Compiled by NS (Norman Schmidt)" although the files may be different.

I know of a QUO-compile but I lack the necessary DLL's. There is also Fire 2.2 Tr4, and Fire 2.2 Good Hardware, which is weaker than Tr4 according to tests. You can also consider testing the new Firenzina program, which I think is a near-clone of Fire 2.2, with permission by Norman.

And as a side-note, I will consider including Fire/Firenzina in the tournament called WORLD'S STRONGEST FREE ENGINES

UPDATE #1: 49 Games complete, Stockfish LEADING 28.5-20.5
UPDATE #2: 77 Games complete, Stockfish leads 43.0-34.0

It appears that this isn't the best Fire 2.2 compile. Next, I will test with Firenzina vs the best-known Stockfish build.

I had also recompiled Yuri's Firenzina and ran a short match of 100 games against my own Fire compilation. The result is a 2 point advantage of Fire over Firenzina, nothing significant.

Like, Yuri said, "Little has changed", and in fact the major changes are the addition of the labels Firenzina and "Yuri Censor (Dmitri Gusev)" in all the files. Other notable change is the ability of Firenzina to be compiled in x64 mode under Visual Studio 2010.

I copied the win64bits.h from Yuri's Firenzina so that my Fire is now successfully compiled for 64 bit mode. Thank you Yuri.

Later today, I may be able to post the results with Fire vs. Stockfish in 64 bit combat.

I had to emphasize that the purpose of my Fire vs Stockfish match is to determine who really is the strongest open source chess engine program using similar compilation and testing environment. No personal grudges, because Stockfish is also my favorite..

drj

Under longer time controls, I believe Stockfish remains the strongest open-source engine.

UPDATE #3 Stockfish 2.3.1 leading 83.5-74.5

Hi! Castle00,

May I know what time control you used so I can set match for comparison? Also, are you using the original Stockfish or the Stockfish owl against Fire owl?

Thanks.
drj

I've stated that it was a 5+3 time control. Stockfish owl & Fire owl compiles are w32.

I am using an "NS" compile of Fire 2.2 x64 and Default Stockfish 2.3.1 x64. The compile might be different from the Fire owl compile.

The best known versions of Fire 2.2 I believe is either the QUO-compile, or
Fire 2.2 Tr4

Hi! Castle,

I believe that we are not standing on the same ground. My intention is to know which chess engines are the strongest, meaning who has the best implementation of data structures and algorithm that constitutes a program applied to chess. The desired result of the best implementation is the one who wins most of the time.

But unfortunately, most of the strongest engines we see in the top of the rating list are doped. It is like the infamous cycling races in France where the champions use performance enhancing drugs.

The executables made by Norman Schmidt in his Fire 2.2 chess engine is probably produced with the old Visual Studio 2008 while the Stockfish 2.3.1 executables made by Jim Ablett is probably produced with the latest Intel compiler. There is a huge difference in performance with these two compilers as I learned from the Internet. If the 2 chess engines in question are roughly equal in strength, using the compiler that can produce the fastest code in one of the chess engines will decide who will consistently win most of the time. It will become a battle between two Computer Compilers, not a contest between two chess engines.

That is why, I compiled these two chess engines using the same compiler, in this case Visual Studio 2010 to eliminate the factor that will favor the other tremendously. It does not matter whether the compiler used is Microsoft's, Intel, GCC or any other compiler, as long as the same compiler will be used in the compilation of the two opposing chess engines.

I think it would be useless to compare the strength of any two comparatively strong chess engines if they are using different compilers because the engine that is compiled with the fastest code executing compiler will prevail.

Thank you for your effort.

drj

I am only testing for comparison, and not conclusion. It is true your testing conditions were probably more thorough and more accurate.

I would like to see x64 compiles of firenzina and the latest stockfish build tested against each other to see how far the engines have progressed.

Hi! Castle00

I am going in that direction of testing the latest builds. But in the meantime, I have to timeout. My computer system just crashed, it will probably take a couple of days to replace the 1TB hard disk and restore everything to its former location, if it is still there. Also, my rating list site will be dormant for a few days.

Have a nice day.

Devastaing. I hope the data loss is minimal

I think I know where the best Firenzina compile is so we can get an accurate comparison

Or if you could, link to the customized compiles.

Hi! Castle00,

Of course, the authority on Firenzina is Yuri Censor. He has different compilations on Firenzina, but the trouble is, only the generic 32 version compilation works for me. The rest terminates with an error message "not valid win32 application". That is the consequence of being so advanced and powerful that it will not work on less blessed machines. Mine are all AMD based computers, mostly running on single core and 32 bit. That is why my rating site uses 32 bit compatible chess engines.

I had one AMD quad-core which I used for the bulk of my tests. I was able to compile Firenzina's 64-bit just a few days days ago but cannot use it for comparison against Yuri's. I was trying to test the performance of the new feature of the software doubled thread which I saw as 6 against 3 of Fire's 64 bit in my quad-core with 3 cores allocated for the virtualized Windows, when my system crashed hard. So, the test will be stalled in a few days.

Anything that uses SSE4, POPCNT, Intel specific compilation and other advanced processor features is not good for me. For example, the latest Bouquet Jukebox is dead in my system and so is Firenzina's very latest compilation with one exception.

Anyway, I am already preparing for 64 bit matches with SMP that will be incorporated in my rating list. This is truly where the best engines will be gauged as strong. It will be in small batches, because of limited computer resources. I am planning to use my own compilation for engines with source codes, if it is stronger than the official distribution to even up the competition.

Best regards,
drj

Thank you very much, drj! (John, right?! )

I've ended the gauntlets after about 280 games when stockfish pulled away. It's a confirmation that drj's suggestion is correct: I have an outdated and weaker compile.

I need to look at Firenzina 2.2, Fire 2.2 Tr4 & Fire 2.2 QUO-compile (considered best) for WSCE Season 4. I will accept only 1 of these compiles for Season 4.

Hi! Richard,

How did you know my name?

drj

drj4759 wrote:Hi! Richard,

How did you know my name?

drj

From Google+ Followers in your Blog.

Yuri Censor wrote:

Castle00 wrote:Throwing more wood into the Fire 2.2 xTreme

I'm attempting a 400-game match between Fire 2.2 xTreme (NS Compile, x64) vs Stockfish 2.3.1 (Default)

5+3 Time Control, 1 CPU, 256MB Hash, No TB's GM_Polyglot.bin book.

This should be a good experiment, where compile differences will kick in! Fire 2.2 xTreme was "under-compiled" to begin with, it seems. Despite that, the outcome is difficult to predict.

Hello Yuri,

I just downloaded the latest Firenzina files from your site but the 64bit version still does not work but with a message "is not a valid win32 application".

Your chess engine is the next in queue in the gauntlet matches for the 64bit competition in the Owl Computer Chess Engines tournament. I made an emergency compile of Firenzina yesterday and tested it quickly against Stockfish 2.3.1 x64 compiled by Jim Ablett. This is a real Intel vs. Intel C++ compiler war. The result which was finished just a few minutes ago, is a 55.5-45.5 win for Firenzina in 100 games.

If you could not provide a usable 64 bit version in a few days, I may use my own O.W.L. version of Firenzina for the duration of the tournament.

Here is the Firenzina 2.2.2 x64 owl version that fried Stockfish.

Download the Firenzina vs Stockfish PGN games here.

Hmmm, I rather naively though that was only a 32-bit version.

I'll try a new test, with Firenzina (Owl version) vs Stockfish (3/25/2013 Build)

Castle00 wrote:Hmmm, I rather naively though that was only a 32-bit version.

I'll try a new test, with Firenzina (Owl version) vs Stockfish (3/25/2013 Build)

Oops! I made a terrible mistake. When I reviewed the PGN file upon seeing your post, the Stockfish 2.3.1 version was 32bit against Firenzina's 64bit. I will make a rematch using both 64bit, the results may favor Stockfish this time.

Sorry...

drj

Wow, thats devastating. At least this will allow further testing and lead us closer to the truth!

Hello,

The rematch between Stockfish 2.3.1 x64-ja and Firenzina 2.2.2 xTreme x64-owl was completed. And the winner is... Firenzina!

Scores:
Firenzina 115.5/200
Stockfish 84.5/200

Read the complete story here.

drj4759 wrote:Hello,

The rematch between Stockfish 2.3.1 x64-ja and Firenzina 2.2.2 xTreme x64-owl was completed. And the winner is... Firenzina!

Scores:
Firenzina 115.5/200
Stockfish 84.5/200

Read the complete story here.

Thank you very much for testing Firenzina! It's good to know that it's a good product comparable to Stockfish.

Hi!,

The first 64 Bit Tournament, 100 rounds, 1 Minute + 1 Second - for Owl Computer Chess Engines Rating List is released. As usual with other rating list sites, the winner is Houdini 3 Pro x64, followed by Critter 1.6a x64 and Stockfish 2.3.1 x64.

Read the full result here.

drj