*** Short-term do-able ***

- Study Richard Delorme's end game solver (and maybe also its references).

- Lookup edge pattern from config. index rather than searching
- Use potiential mobility to replace simple disc counting (maybe)
- Add no-corner stable edge pattern (XOOOOOOX).


*** Major Weakness - Listed in the order of importance ***

- Lack of fine knowledge of corner (pattern) configuration. The constant corner value approach is a much too coarse approximation, making it totally imcompetent at a competitive level. This shortcoming along limits its strengh to no more than mediocre level.

- Certain unstable positions are not accounted in quiescence search. These situations, however, are not too frequent. Thus the program could still be 'good/OK' even with this weakness, should the first one problem be addressed.

- Other weakness/missing features in evaluation function: edge, parity, potential mobility, etc. (Need learning -- currently absent.)

- Search speed related, e.g. transposition table, better ordering, optimization, etc.


Conclusion:
Must have learning abililty if it is going to get anywhere beyond mediocrity, since the #1 weakness can only be addressed by gaining knowledge of numerous corner pattern, which practically can't be manually tuned.

