Comments on https://lichess.org/@/lichess/blog/op1-partial-8-piece-tablebase-available/1ptPBDpC
Interesting blog! This certainly brings more strength and abilities to game analysis. Enjoyed reading this. :)
very interesting, great work guys!
Great to hear some progress on tablebases!
Full 8-men TBs are completely feasible with today's technology, but it would probably cost millions, and the resulting TBs would be so huge (> 1 petabyte) that it would be extremely impractical to distribute or use them. As is, 7-men are already rather impractical even with the optimizations that come from the syzygy format. A subset of 8-men that's much easier to compute but with high practical relevance is good to have.
The note at the end about 9-men op2 bases is very interesting, as that could further extend this concept of tablebases that have very high practical relevance while being computationally reasonable.
What share of 9-men endings feature 2 sets of opposing pawns, and how many positions would that represent? Continuing on this line of thinking, what about 10-men op3? 11-men op4? 12-men op5? At this point we just have 2 kings and 10 pawns without any passed pawn or semi-open file, so relevance fades, but it would be interesting to know the trend when it comes to number of positions and practical relevance (share of normal games reaching such positions).
Full 8-men and 9-men op1 probably won't start making sense before we have a tenfold increase in SSD and DRAM density.
OP!
Wonder how long till we have 32 piece TB
Very good! Lichess.
32 piece tablebase when
Very interesting, a leap forward.
8 men is more relevant than 7 men, as 8 men can be materially balanced and 7 men not.