[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [gnugo-devel] Computer Go Test Collection 2.0

From: Gunnar Farnebäck
Subject: Re: [gnugo-devel] Computer Go Test Collection 2.0
Date: Sat, 22 May 2004 21:39:25 +0200
User-agent: EMH/1.14.1 SEMI/1.14.3 (Ushinoya) FLIM/1.14.2 (Yagi-Nishiguchi) APEL/10.3 Emacs/21.3 (sparc-sun-solaris2.9) MULE/5.0 (SAKAKI)

Markus wrote:
> Version 2.0 of the Computer Go Test Collection has been released.
> The original Computer Go Test Collection version 1.0 was developed by Martin 
> Müller in 1995. The current collection is based on the "specialized 
> collections" subset of the CGTC 1.0. It has been greatly improved in 
> content and restructured to take advantage of current regression testing 
> tools based on the Go Text Protocol (GTP).

This is nice. I tested with current CVS and got the following results
compared to GNU Go 3.4:

blunder:470     PASS F14 [F14]
blunder:1170    PASS E8 [E8]
blunder:1290    PASS T10 [T10]
blunder:1300    PASS T10 [T10]
cut_connect:30  PASS Q12 [Q12|Q15]
cut_connect:71  PASS C4 [C4]
cut_connect:270 FAIL B3 [R10]
cut_connect:300 PASS P17 [P17]
cut_connect:301 PASS P17 [!P18]
cut_connect:340 PASS F14 [F14]
cut_connect:581 PASS J6 [J6]
cut_connect:670 PASS J5 [J5]
cut_connect:820 FAIL L2 [L6]
cut_connect:1031 FAIL G3 [Q6]
enclose_breakthrough:210 FAIL F3 [E2]
enclose_breakthrough:360 PASS E8 [E8]
enclose_breakthrough:410 PASS Q9 [Q9|Q10]
endgame:51      PASS K2 [D2|K2]
endgame:60      FAIL D1 [G4|A11]
escape_capture:230 FAIL O12 [L11]
ko:91           PASS T10 [T10]
life_death:120  PASS A18 [A18]
low_liberties:300 FAIL N9 [E2]
low_liberties:650 PASS Q6 [Q6]
territory:100   FAIL E13 [C4]
territory:171   FAIL M17 [D2|C4]
territory:460   PASS C4 [C4]
territory:480   FAIL Q12 [R17|R7|H17|J3]
territory:860   PASS L13 [L13]
territory:880   PASS D3 [D3|C2]
territory:1220  FAIL R7 [C12|G3|H3|Q3|Q6|Q11]

The node counters went from 488989176 611355 3210745 to
434911984 636177 3572554. The total cpu time was reduced by 2% but due
to varying load this is more uncertain.


reply via email to

[Prev in Thread] Current Thread [Next in Thread]