billforsternz
commented
1 year ago Thank you for this excellent, clearly written problem statement.
It's been a while since I used the hash functions, I originally created them for an old version of Tarrasch when I was trying to use standard database technology to build a Tarrasch database (Basically SQLITE database, all positions in all games stored as hashes - it didn't go well and I abandoned that approach). I don't use these functions any more in any of my projects and am considering removing them; I mean I just generated my own random numbers for these hash functions - maybe the hashes used more widely in chess programming ("Zobrist" hashes I believe they're called) - somehow reduce the random chance of collisions, a real problem with hashes I imagine.
Never-the-less, the hashes should work of course (spoiler alert - I think they do).
I started by reproducing your problem - yes I could do that, thanks again for being so clear.
It didn't take long to see your misunderstanding; The incremental hash algorithm takes a position with a corresponding start hash plus a move and calculates the hash of the new position after the move is played. This is the only way incremental hash calculation can possibly work. Instead you take the position after the move is played, that's not going to work!
Putting it another way, the incremental hash algorithm is a special kind of a make-a-move algorithm - it looks at a position and considers the impact of a move. The position must be the position before the move is made. The comments to the Hash64Update() function itself make this clear.
Hope this helps!
On Sat, 24 Sept 2022 at 23:01, bolomcs50 @.***> wrote:
Hello,
I'm playing around with the Hash functions of the library. I encountered a behavior that I suspect is not the intended one. In short: Hash64Update(hash, move) does not seem to produce the correct Hash. An example: Let's start from the initial position
thc::ChessEvaluation cr; cr.Forsyth("rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1"); thc::Move mv; mv.TerseIn(&cr, "e2e4");
Before playing any move, the position is correct, and it is possible to Hash64Calculate() a Hash:
uint64_t currentHash = cr.Hash64Calculate(); std::cout << "Starting Position:" << cr.ToDebugStr(); std::cout << "Hash64Calculate(): " << currentHash << std::endl;
which yields, as expected
Starting Position: White to move rnbqkbnr pppppppp ........ ........ ........ ........ PPPPPPPP RNBQKBNR HashCalculate(): 1111876492934684975
After playing (or pushing) the move 1. e4, though, calculating the hash again with cr.Hash64Calculate() and updating it with hash = cr.Hash64Update(hash, move), yield different results:
cr.PushMove(mv); // cr.PlayMove(mv); std::cout << "\nPosition after " << mv.TerseOut() << ": " << cr.ToDebugStr(); std::cout << "Hash64Calculate(): " << cr.Hash64Calculate() << std::endl; currentHash = cr.Hash64Update(currentHash, mv); std::cout << "Updated Hash: " << currentHash << std::endl;
yields
Position after e2e4: Black to move rnbqkbnr pppppppp ........ ........ ....P... ........ PPPP.PPP RNBQKBNR Hash64Calculate(): 5618832418565224928 Updated Hash: 6951459208584661089
Shouldn't the two hashes coincide here? Furthermore, if I pop the move, cr.Hash64Calculate() returns the initial hash, as expected, but Hash64Update() does not. In fact:
cr.PopMove(mv); std::cout << "\nPosition after popping " << mv.TerseOut() << ": " <<cr.ToDebugStr(); std::cout << "Hash64Calculate(): " << cr.Hash64Calculate() << std::endl; currentHash = cr.Hash64Update(currentHash, mv); std::cout << "Updated Hash: " << currentHash << std::endl;
yields
Position after popping e2e4: White to move rnbqkbnr pppppppp ........ ........ ........ ........ PPPPPPPP RNBQKBNR Hash64Calculate(): 1111876492934684975 Updated Hash: 2516593853674508462
whereas I would expect that hashing in the same move twice in a row with cr.Hash64Update() would return the initial hash. What am I doing wrong? Am I misunderstanding/misusing the functions?
Thank you!
— Reply to this email directly, view it on GitHub https://github.com/billforsternz/thc-chess-library/issues/15, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABSQYSQUAVMWEOCIKTH7UU3V73NQHANCNFSM6AAAAAAQUTE2PQ . You are receiving this because you are subscribed to this thread.Message ID: @.***>
-- Bill Forster 15A Boundary Road Kelburn Wellington 6012 New Zealand +64 21 357 371 @.***
bolomcs50
Hello,
I'm playing around with the Hash functions of the library. I encountered a behavior that I suspect is not the intended one. In short: Hash64Update(hash, move) does not seem to produce the correct Hash. An example: Let's start from the initial position
Before playing any move, the position is correct, and it is possible to Hash64Calculate() a Hash:
which yields, as expected
After playing (or pushing) the move 1. e4, though, calculating the hash again with cr.Hash64Calculate() and updating it with hash = cr.Hash64Update(hash, move), yield different results:
yields
Shouldn't the two hashes coincide here? Furthermore, if I pop the move, cr.Hash64Calculate() returns the initial hash, as expected, but Hash64Update() does not. In fact:
yields
whereas I would expect that hashing in the same move twice in a row with cr.Hash64Update() would return the initial hash. What am I doing wrong? Am I misunderstanding/misusing the functions?
Thank you!