vittyvirus wrote:I've lately started to take interests in Stockfish code. My Questions are:
1. Can Stockfish improve?
2. How much can Stockfish improve (I've heard sky is the limit)?
3. WHERE can it improve? (Opening, Midgame, or Endgame, or evaluation, or search, or move generation, or code optimization?)
I think the more interesting question is, will anyone be able to keep up with Stockfish? Being open-source, you would think SF could never stay on top since others could borrow ideas, but no one else has an army of developers working for free.
I think that the answer is yes because I see lack of interest in developing stockfish that cause the framwork to be often empty.
There is a way to prevent it by having many tests with low priority or having tests with longer time control but it seems that it is more important for the leaders to save testing time and note that there is at least a small chance that testing at longer time control or random change of some weight may help to find an improvement.
If you go through SF code, you'll be amazed to see how succinctly its coded. If I want to borrow ideas from sf, except evaluation and search techniques, I'd end up in rewriting my engine.
I'm just giving an example, not that I've a working engine.
