The Spanish Inquisition

[Andres Valverde]

the name of the tournament refers to DanaSah being the highest rated

got it, only kidding

PS - DanaSah 3.13 seems a good improvement over previous versions.

Indeed!, I do know that Pedro works a lot on it

[pedrox]

At the moment I am in strike of crossed arms. I believe that I must change by a strike to the Japanese.

Some of the versions of DanaSah have begun with a low score and as they have played have raised the ELO, I believe that to the last version will happen different, she has begun very hard but she will fall at least 50 points of ELO.

I have changed little code in the last versions, perhaps a positive thing is that I have compiled with pgo giving a little more speed (10%)

My favorites are BugChess, Delphi, Twisted Logic. BugChess is a monster in bullet and blitz, I want to see it here.

[Graham Banks]

THE SPANISH INQUISITION

Dual Athlon XP1800+
Arena 1.1 GUI
128mb hash each
3-4-5 piece tablebases
Ponder off
Salvo's 8moves.abk book
40 moves in 75 minutes repeating (adapted for the CCRL)
2 cycles (30 rounds)


Standings after Round 5

4.5 - Twisted Logic 0.099x
3.5 - Tytan 9.32
3.5 - Delphil 1.8
3.0 - Hamsters 0.3
3.0 - Popochin 3.0
3.0 - Alfil 7.6
2.5 - Natwarlal 0.14
2.5 - Little Thought 1.00
2.5 - BB Chess 1.10
2.0 - Hermann 2.0
2.0 - Zeus 1.29
2.0 - DanaSah 3.13
2.0 - NanoSzachy 2.9
1.5 - Typhoon 1.00-314
1.5 - BugChess2 1.5
1.0 - GreKo 5.3

[swami]

My Prediction went right till 5th round...more to go,poor performance from Bugchess at long time controls,I guess it plays extremely well in blitz.

[FrancoisK]

Hi Swami & Graham,

I have mixed feelings about this.
What's sure here is that BugChess2 made a bad start
Unfortunately nothing more significant can be concluded from 5 games.
I usually perform 4000-game tests which can start by 0-5 against BugChess2 and end with a 60% score.

Like you, I had the feelings that Bug is far better at short time controls than long ones, but things are not that clear and simple :
- in Graham's tests, it seems a fixed 8-move book is used instead of the engines' own books (I don't know why by the way - for me the book is part of the engine) I think Bug is significantly better with its own book (which is automatically tuned by its evaluation), but i have not done serious testing to confirm this.
- Bug seems to perform well enough in the WBEC (not enough games as well of course), but there it uses its own book and does pondering.
- in the CCRL rating list, Bug 1.4.1 has the same rating in 40/40 and in 40/4 (too few games as well)
- like most engine programmers i cannot afford enough long time control games, so i only test in blitz tournaments (40/2, 4000 games, 20 various opponents)
- Bug plays with 32 MO hash tables, but this should not account for more than something like a 20-elo loss in long time controls as compared to 128 MO.

So i really do not know what to think.
Far from enough long time controls games with its own book to draw conclusions.
Anyway, I will have a close look at BugChess2 losses, losses are always instructive even if not good for the programmer's pride

Any feeback on this subject or other programmers' experience welcome

Thanks for the tournaments !
François

[swami]

What's sure here is that BugChess2 made a bad start
Unfortunately nothing more significant can be concluded from 5 games.
I usually perform 4000-game tests which can start by 0-5 against BugChess2 and end with a 60% score.

Like you, I had the feelings that Bug is far better at short time controls than long ones, but things are not that clear and simple :

Hi Francois,

I do agree with you regarding the number of games but even a small number of games at long time controls would be almost enough to conclude[Refer Official Swiss type events,they usually have around 7-13 rounds]

But Why I said it was a poor performance was that It's pretty obvious that Bugchess is better than most of these engines considering the average ratings,If this were a blitz tournament,no doubt Bugchess would be leading the table with >4 points,instead of 1.5 in the long time controls.Well Let's wait for the final results,Bugchess would have a chance to finish in top 3.

Anyway,Thanks for this interesting engine that plays way above its average in blitz

Regards.

[Graham Banks]

Anyway, I will have a close look at BugChess2 losses, losses are always instructive even if not good for the programmer's pride

Thanks for the tournaments !
François

Hi Francois,

I'm pleased that you are following our testing.
It would be nice to get an acknowledgement from some other engine authors that they do likewise (apart from the ones who've already let me know).

Playing through BugChess' losses is probably the best thing to do at this stage.

Thanks for the nice engine!

Regards, Graham.

[Alessandro Scotti]

It would be nice to get an acknowledgement from some other engine authors that they do likewise (apart from the ones who've already let me know).

Hi Graham,
add another programmer to the list then, although I'm not so happy to see Hamsters 0.3 play because of its many bugs!

[Graham Banks]

It would be nice to get an acknowledgement from some other engine authors that they do likewise (apart from the ones who've already let me know).

Hi Graham,
add another programmer to the list then, although I'm not so happy to see Hamsters 0.3 play because of its many bugs!

Hi Alessandro,

I already knew that you follow our testing, but thanks anyway!

Looking forward to a bug free Hamsters!

Regards, Graham.

[nthom]

It would be nice to get an acknowledgement from some other engine authors that they do likewise (apart from the ones who've already let me know).

I like watching LittleThought lose every tournament no really.