LeelaZero Go selfplay games are split into single sgf files and classified in weight numbers/hashes.

• using batch.py

  • ($ head batch.py) to see instructions
  • main function
    • unzip xz files
    • remove ^M line ending
    • check integrity of sgf game

• Manually fix integrity issues which are list in batch.log in repo or produced by script

• using sgfcount.py

  • ($ head sgfcount.py) to see instructions
  • main function
    • prepare for splitting in line with sed command
    • split sgf file into small files(5 for now)

• using split2hash.py

  • ($ head split2hash.py) to see instructions
  • main function
    • put each sgf game into its hash named folder (8chars)
    • PB/PW in self-plays are the same
    • PW is more reliable except one missing case in all_2M.sgf

• shell scripts
  • stat_opening_hoshi
  • ...

• opening style
• find joseki
• as you like ...
• ...