gordonr wrote:I'm currently downloading a partial set of 6 man TBs. I will have at most 0.5 Terabyte of space to use, so can't download the full set.
However, maybe I'm underestimating the need for all derivatives before downloading those containing pawns?! For example:
[d]8/8/P7/B3nk2/8/8/P7/K7 w - - 0 1
Here, I've downloaded BPPvsN but I noticed that this caused issues while analysing with Zappa. So I then downloaded QBPvN but I still see issues (Zappa doesn't play a7 and allows a draw).
Rybka handles this ok. I realise that some engines have minimum search prior to using TBs. This defaults to 4 for Zappa. Why is Zappa avoiding a7? The best line stays within my downloaded TBs.
I also plan to start checksumming my downloads for corrupt files, etc. Not yet done.
thanks
Gordon
This is a known problem, has been discussed several times in the past, and some programs do not have the problem (Crafty is an example). The problem is that if you have a pawn on the 7th, and you don't have the promotion EGTBs handy, then if your program can not see a forced mate after promotion, and does see a forced mate (due to EGTB hits) with the pawn sitting on the 7th, it will simply fail to promote.
The solution is to do a search when you reach a position at the root where egtbs say "mate in N" and there is a promotion possible, But you turn off EGTBs for that case so that the pawn on the 7th is no longer a mate in N...
That's a trivial case, you see the promotion in your search (without an instant EGTB mate score) but then see a capture that takes you to the 3-piece tables also.
That's easy.
The hard case is where you promote and there is no quick capture to take you to the n-1 piece tables. Then you have a problem.
