Introducing Omniperft

[Roman Hartmann]

I didn't note that before but the position in question is not a legal FRC-position. Both side still have their castle rights on the king and queenside although the positions isn't mirrored.

Roman

[ilari]

Wow, didn't expect anyone to still use Omniperft, it was pretty much just a programming exercise and a base for the Board class in Cute Chess.

But thanks for the report. The problem seems to be that you're using outdated binaries, which I've been too lazy to update. The bug should be solved by the latest commit here: http://repo.or.cz/w/omniperft.git

Code: Select all

setboard qrk1b1rn/pppp1ppp/4pbn1/8/8/4PBN1/PPPP1PPP/Q1KRB1RN b gb -
q r k . b . r n 
p p p p . p p p 
. . . . p b n . 
. . . . . . . . 
. . . . . . . . 
. . . . P B N . 
P P P P . P P P 
Q . K R B . R N 
FEN: qrk1b1rn/pppp1ppp/4pbn1/8/8/4PBN1/PPPP1PPP/Q1KRB1RN b kq - 0 1
perft 2
Perft with 2 thread(s)
677
Elapsed time: 0 second(s)

Code: Select all

setboard 1rk3r1/8/8/8/8/8/8/1RK1R3 w EBgb -
. r k . . . r . 
. . . . . . . . 
. . . . . . . . 
. . . . . . . . 
. . . . . . . . 
. . . . . . . . 
. . . . . . . . 
. R K . R . . . 
FEN: 1rk3r1/8/8/8/8/8/8/1RK1R3 w KQkq - 0 1
divide 2
Perft with 2 thread(s)
b1b3: 21
b1b2: 22
b1b4: 20
b1b5: 19
b1b6: 18
b1b7: 16
b1b8: 3
b1a1: 24
c1d2: 23
c1c2: 23
c1d1: 23
e1e2: 23
O-O-O: 22
e1e3: 23
e1e4: 23
e1e5: 23
e1e6: 23
e1e7: 21
e1e8: 4
e1d1: 21
e1f1: 23
e1g1: 23
e1h1: 24
465
Elapsed time: 0 second(s)

There's still a castling bug, because after e1e8 Omniperft allows queenside castling for black. A fix is coming...

[Richard Allbert]

- it's a very helpful tool for debugging variants!

Thanks for the reply.

I'll compile the new source.

Regards

Richard

[Richard Allbert]

Thanks, I didn't notice that.

I was setting some up by hand ..

Thankfully, late yesterday evening, Jabba played a few hundres fast FRC games with no problems. A lot of frustration yesterday!

Richard

[ilari]

- it's a very helpful tool for debugging variants!

Thanks for the reply.

I'll compile the new source.

Regards

Richard

I just committed a couple more fixes, FRC castling should now work properly if you update your sources.

[LiquidNitrogenOverclocker]

If you want to have some more perft results, please suggest some positions, I would calculate on those for comparing purposes then.

Reinhard,

Didn't you have a page with perft() for Gothic Chess computed rather deeply to 9 plies?

I don't think I made it that far, maybe to 8 plies we were in agreement.

Is that correct?

[ilari]

If you want to have some more perft results, please suggest some positions, I would calculate on those for comparing purposes then.

Reinhard,

Didn't you have a page with perft() for Gothic Chess computed rather deeply to 9 plies?

I don't think I made it that far, maybe to 8 plies we were in agreement.

Is that correct?

I still have Reinhard's numbers for the starting position:
1: 28
2: 784
3: 25283
4: 808984
5: 28946187
6: 1025229212
7: 39532257395
8: 1509030960338

Also there are some perft figures floating on the net that I think are wrong. Like these (http://www.xs4all.nl/~mlngveld/gothicchess/index.html):
perft 1= 28
perft 2= 784
perft 3= 25283
perft 4= 809826
perft 5= 28976137
perft 6= 1027599562

[smrf]

Ed, I had calculated such numbers up to depth 8 that time.

I have reduced all those web pages because of lack of interest and support.

Intending to write 8x8 / 10x8 Octopus as a combination of a UCI based GUI and a SMIRF follow-up engine using QT for to support different OSs I still have not found other people interested in that matter. Thus there is no speed in that project.

[LiquidNitrogenOverclocker]

Someone from AT & T's research department published a list of Gothic Chess perft numbers here:

http://akpublic.research.att.com/~njas/ ... es/A114037

You can see the first 8 plies are:

28
784
25283
808984
28946187
1025229212
39532257395
1509030960338

So it looks like Ilari's numbers agrees through ply 6:

>1: 28
>2: 784
>3: 25283
>4: 808984
>5: 28946187
>6: 1025229212

[NaltaP312]

For me

perft 7
Perft with 2 threads...
3195901860
Elapsed time: 153 second(s)

on a mac snow leopard core 2 duo 2.33 with gcc compiler

Regards
NaltaP