Is there a hash or string that represents a stockfish evaluation of a position to a certain depth? For example for a given FEN, can you analyze to a depth of say 10 and get a hash to represent that position AND depth of analysis? That would allow you to come back later and, if wanted, continue analysis to a greater depth at a later point.
2 Answers
I doubt that can be possible.
An analysis to a certain depth contains the generated position tree and the calculated values, which is quickly becoming a very large amount of data.
For depth 10, you probably need several megabytes of memory to represent the analysis result in a way that it could be used to continue searching deeper. There is simply no way to compress this into a small hash or anything else.
-
Then how does lichess store their analysis on the cloud so that one can revisit it at a later time and immediately have depth X? May 25, 2020 at 4:28
-
1Because you cannot continue to evaluate to a further depth. The analysis you see on lichess is pre-computed and then displayed to you immediately when you reach the analyzed position, but you cannot pick up where it left off.– causaSuiMay 25, 2020 at 5:26
-
Well, storing several megabytes of data is definitely possible with the current technology... And compressing the data to some extent is likely to also be possible. A different story is whether it is not done because it can reach much larger sizes than that at larger depths.– wimiMay 25, 2020 at 6:01
It is definitely impossible. Stockfish uses a game-tree search with pruning and evaluation heuristics. If it has searched to a certain 'depth', it means that it has built (in memory) a game tree to that depth excluding many pruned branches. To deepen a search, Stockfish would not only have to continue the search to subsequent moves past the end of each originally searched path, but also have to check many heuristically pruned branches (which may include those that were pruned as per alpha-beta pruning, since the evaluations may change). Doing so would require knowing practically the entire game-tree. There is thus absolutely no way to compress this into a short string, short of knowing a simple and optimal strategy for playing chess!