[computer-go] 50, 576 pro/dan games without repetitions nor easily detectable problems

Jacques Basaldúa jacques at dybot.com
Tue Sep 19 03:36:48 PDT 2006


I have finished my "game hunting" for the moment. I feel that 50K is
as good as 60K and, unless someone could get a fresh collection,
(e.g. games played at IGS. Is that available?) it is not worth working
a lot just to get a dozen new entries. Maybe in one or two years the
collection will grow easily when 2006 (or 2007) at KGS is available.

Please note that this collection is the filtered result of studying
141,890 SGF files. It contains 10,670,608 moves. All files have over
100 moves and there are no other sequences than B,W,B,W,.. with pass
moves only at the end if any. See the details here:

    http://www.dybot.com/masters/masters.log

The collection is in a single binary file at :

    http://www.dybot.com/masters/masters.zip

The file is an array of 50576 structures called BinGame of fixed 640
byte length. The Pascal (Delphi) definition is:

TySource    = (srAngYue, sr40K_v1, sr40K_v2, srKGS, srMFOG, srJIGO, 
srEuroAm);

Pos16       = packed record                 // 16 bit
    x, y        : byte
end;

pBinGame    = ^BinGame;
BinGame     = packed record                 // 640 bytes in all
    Hash100,                                // 32 bit unsigned hash of 
the first 100 moves
    Hash200,                                // 32 bit unsigned hash of 
the first 200 moves
    HashLast    : Unsigned32;               // 32 bit unsigned hash of 
the whole game
    NMoves      : longint;                  // 32 bit signed number of 
moves in mov[]
    SecPMove    : integer;                  // 16 bit signed. this is 
min (TM, 32767)
                                            // where TM is the SGF property.
    MinPlLev    : byte;                     // 8 bit 0 = unk; 1..9 = 
dan; > 10 = pro
                                            // (19 = 9p)
    source      : TySource;                 // 8 bit 0 = srAngYue, 1 = 
sr40K_v1, ...
    mov         : array [0..309] of Pos16   // 0-based Cartesian coords. 
(0,0) is the
                                            // lower left corner (255, 
255 is pass)
end;

Since this is used for research, has no remarks or other
"copyrightable" material, I understand that it can be considered
public domain. If it has owners, then it has thousands of owners ;-)

Jacques.


More information about the computer-go mailing list