egtb tree

[jshriver]

Does anyone know if there is a tree or document describing the requirements of each egtbs against others? I might be wrong, but I thought some required subset egtbs to work properly. So knowing what require what would be nice.
But maybe that's just for generation and not for general use.

-Josh

[guyhaw]

The EGTs are self-standing when generated but require all the 'successor subgame EGTs' for fully accurate generation. If P=Q is assumed for conversion, somewhat less accurate EGTs may be generated more conveniently.

The 'tree' of endgames is in fact a lattice because, e.g., KPKP can segue into KQKP and KPKR, which can both segue into KQKR.

[Kirill Kryukov]

Does anyone know if there is a tree or document describing the requirements of each egtbs against others? I might be wrong, but I thought some required subset egtbs to work properly. So knowing what require what would be nice.
But maybe that's just for generation and not for general use.

-Josh

EGTB sets don't require other sets to be used by an engine. The problem is that some engines have difficulties converting from a egtb position into position for which they don't have egtb. In such situations some engines can make suboptimal moves, or miss a win. As far as I know all recent engines don't have this problem, but I am not sure. I use only sets for which I also have all sets reachable from them, so I don't observe this problem anyway.