Winboard Forum

Posted: **18 Mar 2006, 02:41**

I recompiled the WV2 code using profile-guided optimizations, and the result is 15% faster in the opening and early midgame, and drops off to about 5% faster by the late endgame.

You can get WV2.1 on the Slow Chess page
http://www.3dkingdoms.com/chess/slow.htm
(click on the wv2.1 info)

The playing code isn't exactly the same as WV2, there were a few small changes but I don't remember what was changed. They probably didn't affect strength, I hope it isn't weaker.

I debated whether or not to put this one up, since a 15% speed increase probably only amounts to a single digit ELO increase, and I wish I had time to actually work on the program, but I decided it was better than nothing =)

Posted: **18 Mar 2006, 04:41**

Jon Kreuzer wrote:I recompiled the WV2 code using profile-guided optimizations, and the result is 15% faster in the opening and early midgame, and drops off to about 5% faster by the late endgame.

You can get WV2.1 on the Slow Chess page
http://www.3dkingdoms.com/chess/slow.htm
(click on the wv2.1 info)

The playing code isn't exactly the same as WV2, there were a few small changes but I don't remember what was changed. They probably didn't affect strength, I hope it isn't weaker.

I debated whether or not to put this one up, since a 15% speed increase probably only amounts to a single digit ELO increase, and I wish I had time to actually work on the program, but I decided it was better than nothing =)

Sometimes, when you see that the previous ply took 3 minutes, and you see that you have 5:45 on the clock for this move, and the search runs out before the ply finishes, you wonder what a teeny bit more juice would have done.

So quick, like a bunny, I ran and got your new version.

Thanks.

Posted: **18 Mar 2006, 06:26**

Thanks Jonathan.

I sincerely hope that you can find the time and motivation to further improve your fine engine! 8-)

Regards, Graham.

Posted: **19 Mar 2006, 15:54**

Hi Jonathan,

SlowChess Blitz WV2.1 would obtain a first rating by 2591

Tournoi d'entr?e HS-Book.abk
Le Fou numerique, 2006.03.18
Score 1 2
----------------------------------------------------
1: SlowChess Blitz WV2.1 59.0 / 84 17w= 14b+
2: Zappa 1.1 2.0 / 2 1b+ 1w+
3: Glaurung 1.0.2 1.5 / 2 1b= 1w+
4: SlowChess Blitz WV 1.5 / 2 1b= 1w+
5: Spike 1.1 1.5 / 2 1b+ 1w=
6: Tao 5.6 1.5 / 2 1b+ 1w=
7: Yace Paderborn 1.5 / 2 1b+ 1w=
8: Fruit 2.1 1.5 / 2 1b+ 1w=
9: Nejmet 3.07 1.0 / 2 1b= 1w=
10: Pharaon 3.3 1.0 / 2 1b+ 1w-
11: Dragon 4.6 1.0 / 2 1b- 1w+
12: Ktulu 4.2 1.0 / 2 1b+ 1w-
13: SOS 5.1 1.0 / 2 1b= 1w=
14: List 5.12 1.0 / 2 1b+ 1w-
15: Little Goliath Evolution 1.0 / 2 1b= 1w=
16: Naum 1.91 1.0 / 2 1b- 1w+
17: Arion 1.7 1.0 / 2 1b= 1w=
18: Pepito v1.59 0.5 / 2 1b- 1w=
19: Abrok 5.0 0.5 / 2 1b- 1w=
20: Ruffian 1.0.5 0.5 / 2 1b= 1w-
21: Aristarch 4.50 0.5 / 2 1b- 1w=
22: Atlas 2.20 0.5 / 2 1b- 1w=
23: Ayito 0.2.994 0.5 / 2 1b= 1w-
24: Booot 4.75 0.5 / 2 1b= 1w-
25: Chispa 4.0.3 0.5 / 2 1b= 1w-
26: Delphil 1.5b 0.5 / 2 1b- 1w=
27: Diablo 0.4 0.5 / 2 1b= 1w-
28: Homer 1.02 0.0 / 2 1b- 1w-
29: Queen 3.09 0.0 / 2 1b- 1w-
30: Amyan 1.597 0.0 / 2 1b- 1w-
31: Sjeng 12.13 0.0 / 2 1b- 1w-
32: Frenzee 300 0.0 / 2 1b- 1w-
33: Snitch 1.4.5 0.0 / 2 1b- 1w-
34: AnMon 5.60 0.0 / 2 1b- 1w-
35: Muse 0.899b 0.0 / 2 1b- 1w-
36: Gaia 3.3 0.0 / 2 1b- 1w-
37: Terra 3.4 0.0 / 2 1b- 1w-
38: Tornado 0.84.2 0.0 / 2 1b- 1w-
39: Ufim 7.01 0.0 / 2 1b- 1w-
40: WJChess v1.64 0.0 / 2 1b- 1w-
41: Aice 0.97 0.0 / 2 1b- 1w-
42: Patzer 3.61 0.0 / 2 1b- 1w-
43: Hermann 1.6 0.0 / 2 1b- 1w-
----------------------------------------------------
84 games: +28 =24 -32

In the UEL, SlowChess Blitz WV has 2557 but starts with 2649. This was a year ago, since some engines will be updated.

PGN here

Regards,
Patrick

Posted: **20 Mar 2006, 14:25**

Jon Kreuzer wrote:I recompiled the WV2 code using profile-guided optimizations, and the result is 15% faster in the opening and early midgame, and drops off to about 5% faster by the late endgame.

You can get WV2.1 on the Slow Chess page
http://www.3dkingdoms.com/chess/slow.htm
(click on the wv2.1 info)

The playing code isn't exactly the same as WV2, there were a few small changes but I don't remember what was changed. They probably didn't affect strength, I hope it isn't weaker.

Hi Jon,

Solving the PET v1 test I noticed, at least it seems to me, that Slow cannot underpromote to bishop. In position 26 there are 16 possible moves, including the underpromotion to bishop, but it seems Slow never considers the move, although it does count 16 move possibilities.

Here is the position:

[diag]8/5k2/4p3/4Pp1p/5P1P/3Rn2K/6p1/8 b - - 0 1[/diag]

G?bor

P.S. This is exactly how Rybka behaves. Is there something unnaturally difficult about underpromotion to bishop?

Posted: **20 Mar 2006, 20:28**

Dann: I'm surprised at how many people have downloaded WV2.1 already, hopefully it finds some better moves in time. But the number of downloads just makes me wish I had done more on it =)

Graham Banks:
This is still on my list of things to do someday. Lately programs have gotten so strong, maybe it's from using ideas found in Fruit, or even if there was no Fruit, it's probably natural progression. I could tinker with my code once in a while, but I feel it will probably take some concentrated programming effort( and testing on my part ) over a period of time to improve the Slow engine significantly and not mess things up.

Patrick:
Looks like a good result to me, with how strong the competetion is.

Gabor:
I'm sure this is mostly just a matter of importance. Originally Slow didn't have any underpromotions, and I once saw it lose 2 games in a row to underpromotions to a knight. Not knowing underpromotion to a rook has caused some screwy evaluations after searching, but I don't think I've ever seen it change a game result. Bishop underpromotions I've only seen effective in studies.

Since I'm as interested in proper analysis as I am in game playing, someday I'll add the Rook and Bishop underpromotions in the search.

Posted: **20 Mar 2006, 22:25**

Jon Kreuzer wrote:Graham Banks:
This is still on my list of things to do someday. Lately programs have gotten so strong, maybe it's from using ideas found in Fruit, or even if there was no Fruit, it's probably natural progression. I could tinker with my code once in a while, but I feel it will probably take some concentrated programming effort( and testing on my part ) over a period of time to improve the Slow engine significantly and not mess things up.

Jon,

Good to see you doing a bit of work on SlowChess (which always seemed like an oxymoron to me)! :wink:

BTW, I think you can partially blame Tord for all the progress computer chess has made recently. His sharing of ideas has definitely made a number of programs stronger. Of course, Fabien and Ed must also take part of the blame!

Seriously, one of the biggest contributions (IMHO) that both Fabien, Tord and Ed made was to get people away from the way Crafty does things. Bob's approach is excellent, but it's not the only way to program an engine. Once people started trying other ideas, I believe progress was inevitable.

take care,
--tom

Posted: **22 Mar 2006, 14:07**

Jon,

Thanks for the new version - good luck in finding the time to further improve this great engine!

Shaun

Winboard Forum

Another not quite new Slow Chess version

Another not quite new Slow Chess version

Re: Another not quite new Slow Chess version

Re: Another not quite new Slow Chess version

Re: Another not quite new Slow Chess version

Re: Another not quite new Slow Chess version

Re: Another not quite new Slow Chess version

Re: Another not quite new Slow Chess version

Re: Another not quite new Slow Chess version