Over the course of yesterday and today, I uploaded a big update to the 9x9 opening books on katagobooks.org!
The raw size of the zipped book data is about 70% larger than before, and more than double the very first release of the 9x9 books.
The great majority of opening lines considered draws has remained stable, while lots and lots of deviations and clearly dispreferred moves have incrementally sharpened yet closer towards 0% or 100% and gotten deeper and more confident.
But also, if you browse the new version and the 2024 version side by side, you’ll quite readily also be able to find variations whose evals have swung by a lot more. A lot of these are peppered among the moves that were nominally showed close-to-draw winrates, but were still noticeably multiple percent worse than the preferred moves, or had score estimates that looked much worse and suggested the move could be suspect or fragile.
For example here in JP rules two moves:
(old) https://katagobooks.org/book9x9jp-20241105/A6/24217201A680AAEB6C993141F83752C70640ACB894EEAAA74E3180DACED14CC0.html?symmetry=1
(new) https://katagobooks.org/book9x9jp/A6/24217201A680AAEB6C993141F83752C70640ACB894EEAAA74E3180DACED14CC0.html?symmetry=1
Also, the shoulder hit here against tengen in area rules:
(old) https://katagobooks.org/book9x9tt-20241105/94/70BF9CFE9471DA25F556E35887E3886AE05A21AC3AF6D61C6364C94693032B7D.html?symmetry=0
(new) https://katagobooks.org/book9x9tt/94/70BF9CFE9471DA25F556E35887E3886AE05A21AC3AF6D61C6364C94693032B7D.html?symmetry=0
And in some cases it went the other way! Here, the D6 move has been rescued from being a move with a far-from-solved but noticeably worse winrate, to being perhaps a suspect/fragile draw (estimated score still suggests something may be wrong with the move though and it might turn out to be losing again were the book expanded even more but it’s not as clear as the prior book thought):
(old) https://katagobooks.org/book9x9tt-20241105/57/18CC9B6357C0389B42CF5E03C521DA5A2800C28B8E687EDB455F277457105C88.html?symmetry=6
(new) https://katagobooks.org/book9x9tt/57/18CC9B6357C0389B42CF5E03C521DA5A2800C28B8E687EDB455F277457105C88.html?symmetry=6
There are many cases (just numerically speaking, among all the millions of nodes) where the eval has swung even more, entirely flipped from winning to losing, but all the positions where that happened seem to be relatively deep. I’m not aware of any major opening line that has a variation that went entirely “180” on its evaluation, or where a major opening line that was a confident draw and/or among the top preferred moves, got refuted.
If you’re a 9x9 opening researcher/enthusiast, it’s actually kind of fun to find points where the books do differ noticeably though, and play down both side by side having the book that has a higher winrate for black play black’s moves, and the book that has a higher winrate for white play white’s moves, until you reach the ultimate point where they disagreed, where you get down to the actual tactic.
Hope some of this is interesting and useful for those who study 9x9 openings. Enjoy!