Level 11 Perft statistics

Discussion of chess software programming and technical issues.

Moderators: hgm, Rebel, chrisw

andreasoverland
Posts: 3
Joined: Sun Jan 27, 2019 10:28 am
Full name: Andreas Øverland

Level 11 Perft statistics

Post by andreasoverland »

Hi everyone.

This is my first post, so be gentle..

TLDR; My "chess engine" has calculated full perft statistics for level 11. Is it of any interest to post on it the www.chessprogramming.org website?

The engine is a purely statistical engine, so no validation is done. All quantity, no quality.
It was born in Java, lived one day as NodeJS-version, converted to C, then assembly, then back to C after I realised my assembly coding skills was no match for the C-compilers when it comes to optimisation.

Speed wise, right now the engine churns through all possible boards (including level 8 for any boards in a check state on level 7), from level 0 to level 7 in 100 seconds on an 3,4Ghz i5 iMac, running on one thread. By splitting up results into work units the engine can run distributed on any number of cores & machines and the counts can be combined fairly easily.

Anyway, I have understood that to update the chessprogramming website, I need to go through Gerd Isenberg. I find it hard to track him down :) So if these stats are of interest to CPW, I would be very thankful if anyone could give him a nudge in my direction.

Below are an example of stats output from the program.
It shows the same as the fields on the CPW Perft table, but has two differences.
One in the way Discovered&Double checks are counted. The way Alessandro Iavicoli's Ada Chess-engine does it, is by keeping Discovered check counts and Double check counts separate, mine don't just now, but I will adjust my engine to do the same as Ada Chess.
In addition, I keep track of checks that happened after a promotion, an en passant capture or a capture. Come to think of it, I should count how many check that arise from a castling as well.

Anyway. Even though I this is my first written post, I am an avid used of the CPW which is an invaluable tool for noob chess programmers like myself, and I would like to thank all past, current and future contributors for all the work that have been done.

Andreas Øverland
Oslo, Norway

0, N: 1, Caps: 0, EP: 0, Cast: 0, Prom: 0, Checks: 0, Disc: 0, Disc(promo): 0, Disc(capture): 0, Disc(ep): 0, Dbl: 0, Dbl(promo) : 0, Dbl(capture) : 0, Dbl(ep) : 0, Mates: 0, S.Mates: 0
1, N: 20, Caps: 0, EP: 0, Cast: 0, Prom: 0, Checks: 0, Disc: 0, Disc(promo): 0, Disc(capture): 0, Disc(ep): 0, Dbl: 0, Dbl(promo) : 0, Dbl(capture) : 0, Dbl(ep) : 0, Mates: 0, S.Mates: 0
2, N: 400, Caps: 0, EP: 0, Cast: 0, Prom: 0, Checks: 0, Disc: 0, Disc(promo): 0, Disc(capture): 0, Disc(ep): 0, Dbl: 0, Dbl(promo) : 0, Dbl(capture) : 0, Dbl(ep) : 0, Mates: 0, S.Mates: 0
3, N: 8902, Caps: 34, EP: 0, Cast: 0, Prom: 0, Checks: 12, Disc: 0, Disc(promo): 0, Disc(capture): 0, Disc(ep): 0, Dbl: 0, Dbl(promo) : 0, Dbl(capture) : 0, Dbl(ep) : 0, Mates: 0, S.Mates: 0
4, N: 197281, Caps: 1576, EP: 0, Cast: 0, Prom: 0, Checks: 469, Disc: 0, Disc(promo): 0, Disc(capture): 0, Disc(ep): 0, Dbl: 0, Dbl(promo) : 0, Dbl(capture) : 0, Dbl(ep) : 0, Mates: 8, S.Mates: 0
5, N: 4865609, Caps: 82719, EP: 258, Cast: 0, Prom: 0, Checks: 27351, Disc: 6, Disc(promo): 0, Disc(capture): 0, Disc(ep): 0, Dbl: 0, Dbl(promo) : 0, Dbl(capture) : 0, Dbl(ep) : 0, Mates: 347, S.Mates: 0
6, N: 119060324, Caps: 2812008, EP: 5248, Cast: 0, Prom: 0, Checks: 809099, Disc: 375, Disc(promo): 0, Disc(capture): 71, Disc(ep): 0, Dbl: 46, Dbl(promo) : 0, Dbl(capture) : 46, Dbl(ep) : 0, Mates: 10828, S.Mates: 0
7, N: 3195901860, Caps: 108329926, EP: 319617, Cast: 883453, Prom: 0, Checks: 33103848, Disc: 19654, Disc(promo): 0, Disc(capture): 3539, Disc(ep): 91, Dbl: 1628, Dbl(promo) : 0, Dbl(capture) : 1592, Dbl(ep) : 4, Mates: 435767, S.Mates: 0
abulmo2
Posts: 433
Joined: Fri Dec 16, 2016 11:04 am
Location: France
Full name: Richard Delorme

Re: Level 11 Perft statistics

Post by abulmo2 »

What does Disc and Dbl mean?
Richard Delorme
andreasoverland
Posts: 3
Joined: Sun Jan 27, 2019 10:28 am
Full name: Andreas Øverland

Re: Level 11 Perft statistics

Post by andreasoverland »

Discovered and Double check.

Andreas.
User avatar
Ajedrecista
Posts: 1968
Joined: Wed Jul 13, 2011 9:04 pm
Location: Madrid, Spain.

Re: Level 11 Perft statistics.

Post by Ajedrecista »

Hello Andreas:

Welcome to TalkChess! Gerd writes in this forum, so you will probably reach him there.

I guess that you know about François Labelle's page about statistics on chess games. You could check your results with his, despite he has got some fields that you did not post and vice versa. A quick glance on your Perft(7) statistics reveals that both of you agree in checks (32668081 + 435767 = 33103848) and checkmates (435767), as well as in discovered checks (18026 + 1628 = 19654) and double checks (1628) if I understand well these breakdowns.

Regards from Spain.

Ajedrecista.
andreasoverland
Posts: 3
Joined: Sun Jan 27, 2019 10:28 am
Full name: Andreas Øverland

Re: Level 11 Perft statistics.

Post by andreasoverland »

Ajedrecista wrote: Tue Sep 17, 2019 6:52 pm Welcome to TalkChess!
..
Ajedrecista.
Thank you Ajedrecista!

And no, I did not know about the statistics produced by François Labelle! Those are some impressive numbers.
Thank you for pointing me in a very interesting direction :)

Your understanding of the breakdowns are correct.

Andreas.
alessandro
Posts: 49
Joined: Tue Aug 12, 2014 11:21 am
Location: Lund
Full name: Alessandro Iavicoli

Re: Level 11 Perft statistics

Post by alessandro »

Hi Andreas!

Good job! I'm happy to hear that AdaChess was in some way useful for you!
--
AdaChess - Smart Chess Engine - https://github.com/adachess/AdaChess

Image