close 965025 thanks upstream clarification from Daigo:
> We (as the upstream) created an opening book file and > evaluation files in a statistical / machine learning > way from a large amount of data with lots of machine > power and tuning parameters. It was not our priority > to develop an out-of-box environment of tools and data > for exactly mirroring what we produced. However, it > would be capable of power programmers to customize the > binary files by reading and using APIs defined in OSL. > Say, they could put more weight on Bishop or ban some > moves in an opening position.