Zobrist Hashing

[kuket15]

Thanks for sharing this interesting project with the community.

I just wonder why you opted for zobrist hashing. Isn't perfect hash achievable with an unsigned long?

In my mind something like the following could work:

• 5 bits for each piece, to store the current position on the board (including the removed status). With 10 pieces -> 50 bits total.
• 1 bit to store the active player.
• 7 bits to store the current permutation of the 5 cards out of 120 possible permutations.

This leads to a perfect hash using 58 bytes and avoids any related hash collision issue.

Of course I might be wrong and that's why I'm asking.

Regards

Luca

[maxbennedich]

Hey Luca, thanks for your note, you're absolutely right, and in fact I have a C++ version which I'm not quite ready to publish yet, but which uses a scheme similar to what you suggest, with a 57 bit perfect hash (I believe 5 bits are enough to store the cards, assuming 30 permutations).

I think I was too influenced by chess engines when I originally wrote this that I didn't stop to think if it could be done in a better way :)