Chess Stack Exchange is a question and answer site for serious players and enthusiasts of chess. It only takes a minute to sign up.
Sign up to join this community
This is a follow-up question to Is Houdini 6 a Stockfish clone?
Houdini 6.03 included a bugfix: "Correction for incorrect detection of stalemate in positions with white pawn capture moves."
If Houdini is a Stockfish clone, then this bug should also have been in Stockfish. Was it?
Reviewing the Stockfish Issues/Bugs and commits & Releases on GitHub
would indicate that this has not been logged as an issue or a bug. In addition I am unable to locate any instances of users reporting the same faults that Houdini had with Stockfish.
In addition if you review the commit difference between version 8 and 9 it is also not addressed in that release (or any other future releases to my knowledge)
Potential Resons why the bug exists in Houdini and not Stockfish
The question implies that there are no code changes between Stockfish and Houdini.
This is largely the case. The code more or less went through a google translate function. The algorithms with how it calculates the strength of positions and moves remains entirely unchanged. They added a expletitive multiplier on the output for the calculations / second which buffed the number by 8% or something crazy. Whole bunch of dodgy stuff. The program was basically identical but Dutch.
One thing I did notice that was different when all this came out and the source code was flying around; Was the weighting of pieces in stages of the game. I am not going to go through it programmatically (partially because I can't seem to find the code anymore) but a simplification on what is occurring is the following;
StockFish gives pieces a value based of a theoretical weighting depending on the position of the game (or more specifically how it interprets the stage of the game);
Lets say hypothetically stockfish evaluates it as;
Houdini does the exact same thing, but someone didn't agree with the overall evaluation that the folks behind stockfish thinks, they do it hypothetically like the following.
Say on the board White has 1 Bishop 2 Pawns , Black has 5 Pawns. Effectively the difference is;
Now fundamentally the code is identical but slight variable changes can have an impact on the output. Now this is the next thing; who's weighting is correct? Considering Stockfish (to my knowledge happy to be proven wrong) have not released a version update indicating a "fix" for this issue that Houdini had, I would say ... Stockfish probably evaluated it right.
To summarise it would appear , that Houdini is largely a dutch translation of source code of stockfish. It would make sense that Stockfish would be bound by similar issues ...however the Houdini team also did change some minor things that does affect the position evaluation.
