hgm wrote:The FxR+H_ prefix is only used when a board size is used that is not standard for the variant of the given name (so 16x16+0_normal would be normal Chess on a 16x16 board, and 8x8+5_knightmate would be a crazyhouse version of knightmate, because it specifies holdings>0,and holdings is the number of piece types you can hold in hand after capturing them to drop back onto the board). As we are interested only in standard Capablanca Chess, (i.e. on 10x8 board) feature variant="capablanca" would be fine.
During the tourney I will probably play gothic as capablanca with a setup position, because not all engines support "gothic" (it used to be patented, but the patent expired now). For the other initial setups (Carrera, Bird reverse Embassy) I would have to do that anyway.
WinBoard does handle Capablanca FENs (A for ArchBishop, C for Chancellor, "10"for 10 empty squares, KQkq for castling rights). So you can support setboard to receive them (the obsolete edit command would work as well; Fairy-Max still uses that). It is important to be able to handle setup positions, as some of the games will be played from positions other than the standard Capablanca opening. (Like last time we will probably play 10 games per pairing, 5 different opening setups with both colors. This is a good way to guarantee different games even for engines that have no book.)
Castling is done by moving the King 3 steps towards the Rook, and in XBoard protocol is denoted as f1i1 (king side) or f1c1 (queen side). WinBoard will also understand O-O and O-O-O (and any SAN, in fact), but it will send long algebraic unless the engine requested feature san=1 (which I would not recommended). I will only use starting positions where the Kings are on f1/f8.
Thanks for all the tips.
I got here an initial version doing some perft on variant capablanca.
8 r n a b q k b c n r
7 o o o o o o o o o o
6 . . . . . . . . . .
5 . . . . . . . . . .
4 . . . . . . . . . .
3 . . . . . . . . . .
2 P P P P P P P P P P
1 R N A B Q K B C N R
a b c d e f g h i j
Perft 1 Summary
Count: 28
Time: 0.01 sec
Perft 2 Summary
Count: 784
Time: 0.01 sec
Perft 3 Summary
Count: 25228
Time: 0.01 sec
Perft 4 Summary
Count: 805128
Time: 0.30 sec
Perft 5 Summary
Count: 28741319
Time: 10.30 sec
Perft 6 Summary
Count: 1015802437
Time: 364.97 sec
I am testing this for actual game play now and hope to release within this month. Here is a sample search from starting position in 20 sec.
Wow, you are fast! I think TJchess10x8 has a perft function (none of my engines do, and they would in general not be correct anyway, ad they don't handle all under-promotions).
hgm wrote:Wow, you are fast! I think TJchess10x8 has a perft function (none of my engines do, and they would in general not be correct anyway, ad they don't handle all under-promotions).
OmniPerft 1.0 by Ilari Pihlajisto
setboard rnabqkbcnr/pppppppppp/10/10/10/10/PPPPPPPPPP/RNABQKBCNR w KQkq - 0 1
r n a b q k b c n r
p p p p p p p p p p
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
P P P P P P P P P P
R N A B Q K B C N R
FEN: rnabqkbcnr/pppppppppp/10/10/10/10/PPPPPPPPPP/RNABQKBCNR w KQkq - 0 1
divide 6
Perft with 4 thread(s)
a2a3: 28701556
b2b3: 28126504
a2a4: 32556619
b2b4: 30369092
c2c3: 30363782
c2c4: 32155480
d2d3: 46397175
d2d4: 48325852
e2e3: 43107759
g2g3: 27723215
e2e4: 46818840
f2f4: 52678872
f2f3: 52623654
g2g4: 29988375
i2i3: 24802715
h2h3: 31717791
i2i4: 26996704
h2h4: 35870856
j2j3: 28588868
j2j4: 32465690
b1a3: 28723639
b1c3: 32058743
c1b3: 35660984
c1d3: 39217127
i1j3: 30250326
i1h3: 33345105
h1g3: 53499031
h1i3: 52668083
1015802437
Elapsed time: 24 second(s)
It matches mine,
Perft 6 Summary
Count: 1015802437
And for perft 7, I got from OmniPerft using also variant capablanca just posting, I still have to run mine on this.
OmniPerft 1.0 by Ilari Pihlajisto
setboard rnabqkbcnr/pppppppppp/10/10/10/10/PPPPPPPPPP/RNABQKBCNR w KQkq - 0 1
r n a b q k b c n r
p p p p p p p p p p
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
P P P P P P P P P P
R N A B Q K B C N R
FEN: rnabqkbcnr/pppppppppp/10/10/10/10/PPPPPPPPPP/RNABQKBCNR w KQkq - 0 1
divide 7
Perft with 4 thread(s)
b2b4: 1085757096
a2a3: 1013134867
b2b3: 985712154
a2a4: 1203959097
c2c3: 1080543017
c2c4: 1179776149
d2d3: 1964205201
e2e3: 1685542966
e2e4: 1953509849
d2d4: 2066731682
g2g3: 934224360
g2g4: 1070733443
f2f3: 2291524622
h2h3: 1150855234
i2i3: 824679846
f2f4: 2308252221
h2h4: 1365929557
i2i4: 939842622
j2j3: 1002558686
b1a3: 1020793907
j2j4: 1194950806
b1c3: 1168594586
c1b3: 1379774760
c1d3: 1571743310
i1h3: 1228072384
h1g3: 2277097554
h1i3: 2206626689
i1j3: 1085978499
39241105164
Elapsed time: 576 second(s)
And for perft 7, I got from OmniPerft using also variant capablanca just posting, I still have to run mine on this.
OmniPerft 1.0 by Ilari Pihlajisto
setboard rnabqkbcnr/pppppppppp/10/10/10/10/PPPPPPPPPP/RNABQKBCNR w KQkq - 0 1
r n a b q k b c n r
p p p p p p p p p p
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
P P P P P P P P P P
R N A B Q K B C N R
FEN: rnabqkbcnr/pppppppppp/10/10/10/10/PPPPPPPPPP/RNABQKBCNR w KQkq - 0 1
divide 7
Perft with 4 thread(s)
b2b4: 1085757096
a2a3: 1013134867
b2b3: 985712154
a2a4: 1203959097
c2c3: 1080543017
c2c4: 1179776149
d2d3: 1964205201
e2e3: 1685542966
e2e4: 1953509849
d2d4: 2066731682
g2g3: 934224360
g2g4: 1070733443
f2f3: 2291524622
h2h3: 1150855234
i2i3: 824679846
f2f4: 2308252221
h2h4: 1365929557
i2i4: 939842622
j2j3: 1002558686
b1a3: 1020793907
j2j4: 1194950806
b1c3: 1168594586
c1b3: 1379774760
c1d3: 1571743310
i1h3: 1228072384
h1g3: 2277097554
h1i3: 2206626689
i1j3: 1085978499
39241105164
Elapsed time: 576 second(s)
Variant capablanca to perft 7 is fine now it matches with OmniPerft results.
8 r n a b q k b c n r
7 o o o o o o o o o o
6 . . . . . . . . . .
5 . . . . . . . . . .
4 . . . . . . . . . .
3 . . . . . . . . . .
2 P P P P P P P P P P
1 R N A B Q K B C N R
a b c d e f g h i j
Perft 7 Summary
Count: 39241105164
Time: 2829.85 sec
Well, it seems you are already quite close to something that could actually play games. If you have anything you want me to test, I could set up some games that could be viewed over the internet.
hgm wrote:Well, it seems you are already quite close to something that could actually play games. If you have anything you want me to test, I could set up some games that could be viewed over the internet.
Still fixing some minor bugs, hope to release within the next 24 hours.
hgm wrote:Well, it seems you are already quite close to something that could actually play games. If you have anything you want me to test, I could set up some games that could be viewed over the internet.
I got one here that is ready, look for Bihasa v1.0, this is the engine name that also supports gothic aside from capablanca. Setboard command is supported, it also can ponder. https://sites.google.com/site/deuterium ... ownloads-1
Bihasa seems to work quite well, and also seems quite strong. Good job!(And in record time! )
I have set up a gauntlet for Bihasa, (40 moves/10 min), against the 10 other engines, weak to strong. Not sure if I am runing the latest version of everything; I just used what I had and could easily find (I am also using a new WinBoard with the internal tournament manager, so I had to re-install all engines).
After 2 games per pairing against all 10 opponents, I will repeat the cycle. This might cause duplicate games against some engine if Bisaha is deterministic.
On move 49 Bihasa is at +12.70, Nebiyu is comparatively unworried (-1.65). On move 50 Bihasa's score sudddenly collapses to +1.71. It must have overlooked something (could be a perpetual).