Today I uploaded new versions of the 9x9 opening books at https://katagobooks.org/
A bunch of the expansion was based on a dataset of GoQuest games (one of the ones at SGF library or package of 9x9 games? - #11 by aheavyburden), where every variation that occurred in at least 5 games was given at least 1 million visits out to 24 ply.
Also I manually upweighted the book expansion to explore a ton more variations for the 3-3 invasion under the 4-4 point since I’ve heard people are sometimes interested in that. Of course, the 3-3 invasion is bad for white, but did you know that if black follows hane-at-the-head-of-two into the natural line for 19x19 then actually the position is favorable for white at top levels of play? https://katagobooks.org/book9x9jp/43/007FA3A343C5072FC64E152F872F9CD4EBD805B8398C94BD65FE6E639F00FA33.html?symmetry=0
And generally I just let it explore more in a variety of variations, with some manual biasing and focusing of some sharper lines. Most evals are very similar as one would expect, with some side lines that got more visits being more a bit more converged towards 0% or 100% winrate than before, but if you dig you can also find places where there are some interesting changes to some evaluations.
New: https://katagobooks.org/book9x9jp/37/07DF2B2E3704A92EADE812D9DD3E1FA160399FB7E19517F17478754B2AA74C0A.html?symmetry=5
Old: https://katagobooks.org/book9x9jp-20231027/37/07DF2B2E3704A92EADE812D9DD3E1FA160399FB7E19517F17478754B2AA74C0A.html?symmetry=5
Here’s a particularly big difference in evals for a move, which was already a bit suspicious due to bad shape:
New: https://katagobooks.org/book9x9jp/71/74774EE071B392276F773464906B1EE8C21844FACAF49F62D416B62B0E7CB271.html?symmetry=6
Old: https://katagobooks.org/book9x9jp-20231027/71/74774EE071B392276F773464906B1EE8C21844FACAF49F62D416B62B0E7CB271.html?symmetry=6
Also just for fun (not a lot of visits, also not really advertised anywhere) here’s a short test of a 19x19 book https://katagobooks.org/book19x19jptest/root/root.html