The OBender Decree

[Graham Banks]

THE OBENDER DECREE

Athlon 64X2 4200+ 2.2GHz
Fritz 10 GUI
Xmas2640-12.ctg
256MB hash for each
3-4-5 piece egtbs
40/45 time control
2 cycles 30 rounds 240 games
Run by Charles Smith


Standings after Round 16

12.0 - ZCT 0.3.2451
10.5 - Parrot 07.01.16
10.0 - Matilde 2008
9.5 - AliChess 4.16
8.5 - OBender 3.2.4
8.0 - Hamsters 0.0.6
8.0 - Cheese 1.2
8.0 - Neurosis 2.4
8.0 - Atak 6.0
6.5 - Mustang 4.97
6.0 - BikJump 1.8
6.0 - Matheus 2.3
5.5 - Clarabit 1.00
5.0 - Feuerstein 0.4.5.1
5.0 - Gibbon 2.01b
3.5 - Buzz 0.07

[Graham Banks]

THE OBENDER DECREE

Athlon 64X2 4200+ 2.2GHz
Fritz 10 GUI
Xmas2640-12.ctg
256MB hash for each
3-4-5 piece egtbs
40/45 time control
2 cycles 30 rounds 240 games
Run by Charles Smith


Standings after Round 23

18.5 - ZCT 0.3.2451
15.5 - Matilde 2008
15.0 - AliChess 4.16
14.5 - Parrot 07.01.16
14.5 - Atak 6.0
14.5 - Neurosis 2.4
12.0 - Hamsters 0.0.6
12.0 - Cheese 1.2
10.5 - OBender 3.2.4
10.5 - Clarabit 1.00
9.0 - Feuerstein 0.4.5.1
8.5 - Gibbon 2.01b
8.0 - BikJump 1.8
8.0 - Matheus 2.3
7.5 - Mustang 4.97
5.5 - Buzz 0.07

[Graham Banks]

THE OBENDER DECREE

Athlon 64X2 4200+ 2.2GHz
Fritz 10 GUI
Xmas2640-12.ctg
256MB hash for each
3-4-5 piece egtbs
40/45 time control
2 cycles 30 rounds 240 games
Run by Charles Smith


Final Standings

22.5 - ZCT 0.3.2451
21.5 - Matilde 2008
19.5 - AliChess 4.16
19.0 - Atak 6.0
18.0 - Hamsters 0.0.6
18.0 - Neurosis 2.4
16.5 - Cheese 1.2
16.0 - Parrot 07.01.16
14.0 - Clarabit 1.00
13.5 - OBender 3.2.4
12.0 - Feuerstein 0.4.5.1
11.0 - Matheus 2.3
10.5 - BikJump 1.8
10.5 - Gibbon 2.01b
9.5 - Mustang 4.97
8.0 - Buzz 0.07


The complete tournament pgn (zipped) can be downloaded here:
http://www.kirill-kryukov.com/chess/dis ... p?id=12025

[Graham Banks]

Matilde 2008 doesn't appear in our rating lists because it has a small opening book built into the exe which can't be disabled.

[swami]

Congrats Zach, Really good performance from ZCT.

Don't really know what exactly the bug is with the 0.2451, other than faulty MP implementation. Does it play little(>30 elo) worser than 0.2447?

Could I still go through these CCRL games for 0.2451, Zach?

[Zach Wegner]

Congrats Zach, Really good performance from ZCT.

Don't really know what exactly the bug is with the 0.2451, other than faulty MP implementation. Does it play little(>30 elo) worser than 0.2447?

Could I still go through these CCRL games for 0.2451, Zach?

Thanks Swami!

I don't remember exactly all the details of these releases, but I think neither of those releases had a fully working SMP search. .2447 hung I believe, but .2451 had a bug where one processor would not search around half of the time.

I can't remember exactly what it was, but I do remember finding a bug introduced in .2451 that wasn't there in .2447. .2451 should be about 50 elo weaker. I'm looking around a bit, but I can't remember exactly what it was.

BTW, today I was working on generating graphs of ZCT versions. Here's a prototype: http://zct.sourceforge.net/graph.png

A few versions don't have games played, and the rating probably isn't very reliable. All of the engines use default settings, and no books. This is rather temporary, until I get around to setting everything up properly, and getting some more engines to test with. So take that graph with a truckful of salt. There have been a bunch of changes lately, and a bunch of new bugs introduced. I found a big one today, and so with .2479 as a base I'll make the engine more solid and work on the eval.

I've also been working on SMP search, I think I will have a good speedup sometime soon...

[Graham Banks]

I don't remember exactly all the details of these releases, but I think neither of those releases had a fully working SMP search. .2447 hung I believe, but .2451 had a bug where one processor would not search around half of the time.

I can't remember exactly what it was, but I do remember finding a bug introduced in .2451 that wasn't there in .2447. .2451 should be about 50 elo weaker. I'm looking around a bit, but I can't remember exactly what it was.

Hi Zach,

sounds like ZCT 0.3.2451 running on single-cpu is fine then?

Graham.

[Zach Wegner]

Hello Graham,

Actually .2451 should be a little better. .2451 should be better than .2447 with respect to MP (if .2447 even works).

I know the UEL had .2451 as worse, but I went to check the site and it says that there is a problem with .2451--doesn't work in Arena. Don't know what that is...

Anyways, all this doesn't really matter too much. ZCT has changed a lot recently, and I don't tend to concentrate on old versions much. I don't plan on releasing again until it's much stronger. It's mainly important here because Swami was looking for some more ZCT games to look at, which should be rather similar to newer versions, if weaker.

P.S. I'm sounding like a bit of a broken record... "I can't remember exactly what it was." I should probably read my posts before I submit them.

[Graham Banks]

Hello Graham,

Actually .2451 should be a little better. .2451 should be better than .2447 with respect to MP (if .2447 even works).

I know the UEL had .2451 as worse, but I went to check the site and it says that there is a problem with .2451--doesn't work in Arena. Don't know what that is...

Anyways, all this doesn't really matter too much. ZCT has changed a lot recently, and I don't tend to concentrate on old versions much. I don't plan on releasing again until it's much stronger. It's mainly important here because Swami was looking for some more ZCT games to look at, which should be rather similar to newer versions, if weaker.

P.S. I'm sounding like a bit of a broken record... "I can't remember exactly what it was." I should probably read my posts before I submit them.

Okay - thanks Zach. Looking forward to your next release, whenever it may be.

Cheers, Graham.

[swami]

Thanks, Zach. Graph is really useful, probably fun for you to keep track of your engines progress.

Perhaps I will go through the CCRL games first, 50 elo weaker is not so bad, it did indeed perform well in this tournament, so it should be around 0-30 elo weaker if anything.

I will organise the ZCT gauntlet on longer time controls on the other machine. May I get the beta version especially the one that is higher in the graph? Version number 0.2460 something. That would really help me find weaknesses in there and will probably submit the report to you in weeks time.