Look Out Batman! (Final Standings)

[Graham Banks]

LOOK OUT BATMAN!

Athlon 64 3200+
1GB RAM total, 256 MB for each engine
3-4-5 piece EGTBs
HS-8moves.ctg book
40 moves in 45 minutes repeating (adapted to CCRL 40/40)
4 cycles (36 rounds)
Run by Charles Smith


Final Standings

24.0 - Rotor 0.1a
23.5 - Pupsi 0.18
22.5 - Chispa 4.0.3
19.5 - Joker 1.1.06
18.0 - AliChess 4.08
17.0 - Buzz 0.08
16.0 - NagaSkaki 4.00
14.5 - Neurosis 2.3
13.5 - Counter 0.2
11.5 - Feuerstein 0.4.51


The complete tournament pgn (zipped) can be downloaded here:
http://kd.lab.nig.ac.jp/chess/discussio ... hp?id=6176

[swami]

My Prediction went right here too, Joker should have been ahead of Chispa then only I had guess the exact Top 3!

[hgm]

Sorry about that.

Note that this is a Joker version that has already been superseded, so I have a valid excuse!

Strange thing is that all these versions (1.1.06, 1.1.07, 1.1.08) contain a bug that leads to addition of a random value of +/- 0.25 pawns to the evaluation. This comes about by accessing a local array out of bounds (to get the position of the backmost pawn in the 'i-file', to be exact). The unusual thing is that the data that is accessed this way is not reproducible from run to run, apparently it is used for reading the system clock or something like that. It really has a different effect on the eval in every run (which is how I found it).

The most remarkable thing of all: I of course corrected this bug as soon as I found it. But that doesn't make Joker play any better! In fact, if anything, it seems to play slightly worse without this bug!

[swami]

Sorry about that.

That's Okay.

Note that this is a Joker version that has already been superseded, so I have a valid excuse!

I know it.

Strange thing is that all these versions (1.1.06, 1.1.07, 1.1.08) contain a bug that leads to addition of a random value of +/- 0.25 pawns to the evaluation. This comes about by accessing a local array out of bounds (to get the position of the backmost pawn in the 'i-file', to be exact). The unusual thing is that the data that is accessed this way is not reproducible from run to run, apparently it is used for reading the system clock or something like that. It really has a different effect on the eval in every run (which is how I found it).

I guess this kind of stuff belongs in programming and technical discussion forum,I'd rather read some technical business jargon

The most remarkable thing of all: I of course corrected this bug as soon as I found it. But that doesn't make Joker play any better! In fact, if anything, it seems to play slightly worse without this bug!

Thanks for the twists!

I hope you fix all the bugs and implement missing features in Joker,and one days it wouldn't be called a Joker anymore.

[Jan Brouwer]

Wow, third time Rotor won a CCRL tournament, even if only by a very small margin.
This is realy too much honour for a simple program

I am looking forward to seeing Rotor 0.2 being tested by the CCRL team, the first rating from Le Fou Numerique was a bit disappointing
(about 12 ELO points if I remember correctly, although it seems to be climbing a bit now).

Thanks again for running these tournaments!

[Graham Banks]

Wow, third time Rotor won a CCRL tournament, even if only by a very small margin.
This is realy too much honour for a simple program

I am looking forward to seeing Rotor 0.2 being tested by the CCRL team, the first rating from Le Fou Numerique was a bit disappointing
(about 12 ELO points if I remember correctly, although it seems to be climbing a bit now).

Thanks again for running these tournaments!

Thanks Jan,

Rotor 0.2 will be tested soon.

Regards, Graham.