You are right that my comparison with the 5 piece tablebase was a little glib.
But for practical purposes, the number of files is pretty much irrelevant.The number of positions is what matters to the calculation. The size on disk is not important, because the Syzergy tablebase has to be used online anyway.
Of course, the tablebase files are all very much smaller than 7 piece ones because two pieces are in one of 2 configurations rather than ~4032 (actually symmetry adjusts that, but still a big ratio.
I see now the number is different to the 5 piece tablebase but only because one essential piece - a king - is accounted for. There is no such effect on other rooks.The extra freedom is to have one of 10 types of piece in place of the king already accounted for, But the ratio is reduced by an increased chance for repeated other pieces (eg a position with two queens has just under about half the positions of one with queen and king. And one with 3 pawns has about 3 times fewer positions than one with king and 2 pawns. (And the pieces have somewhat fewer square to go on due to the king and rook. Illegal checks have a less clear effect).
Overall, this is not going to make it more than 10 times bigger than the 5 piece tablebase, IMO, but bigger it will be. But still more than 1000 times smaller than Syzygy (which is 16,000 times more positions than a 5 piece tablebase).
Stockfish blunders in 6 and 7 piece table base positions
then SF shouldnt be calcing out 6 to 7 piece TB positions. it should turn it to a rigid solutionist like sygyzy right ?
so. does it ?
It can't calc out to 6 to 7 man tablebases using Syzygy because 7 man is the highest you get in Syzygy.
It will use the 6 and 7 man Syzygy tables once it gets there if the SyzygyPath option is set by the GUI in use to a URL or sequence of URLs containing the relevant 6 and 7 man tablebases, otherwise not.
If the the 5 man tablebases are not included it may proceed to blunder in 5 man positions, but I think you're pretty safe once you get down to 4 (but, of course, by that time the damage may already be done).
On an only loosely related matter, the Syzygy 7 piece tablebase could be complemented with a 7 piece tablebase that only contains positions where castling rights exist. There is an oddity with the online version that a FEN with castling rights is displayed/entered, but not respected.
This is not such a huge undertaking - of similar difficulty to creating a 5 piece tablebase because there are only 2 configurations for a king and a rook, leaving 5 other pieces to place elsewhere.
Of course the idea is that if castling rights disappear during analysis, you access Syzergy, and if a capture takes place you access a similar tablebase allowing castling with 6 pieces (very much smaller), with other (relatively) titchy ones needed for the whole job.
Commensurate with size, in total this would take about 0.01% as much computing as creating Syzergy.
The obvious way to do it would be to have a set of tablebases with single castling rights which are children of tablebases with two similarly with 3 and 4 all much smaller than the corresponding 0 castling rights child, but a fair few thousand of them. I think there are about 2000 7 man tablebases with rooks - that scheme would produce about another 30000 plus the rest for 3-6 men. I think that's why it's not been done.