![]() In the Cerebellum book the earlier position would have the +0.3 evaluation rather than the +0.9. A number of move later, due for example to transpositions this may now be 55% and the evaluation has dropped to +0.3. How is this different from any other opening book? In a typical opening book we mat have a position that appears to be scoring 75% and have an engine evaluation of +0.9. This book is engine generated and fully consistent. The engine itself is a copy of the latest Development Build with one important modification: the addition of the Cerebellum opening book. So moving onto versions which do more than speed up Stockfish.Ī very interesting project. ![]() This means the engine should be optimised for your hardware and environment. This is an utility which will complie a copy of the lastest Stockfish development versions on your machine. These can be around 10% faster than the basic versions. A version called Goby has been coded for the Mac -although versions of this also work on Windows! Different CompilationsĪ number of helpful people compile their own versions of Stockfish Development versions. Personally I have found the engines to work better in different positions. What is interesting is that some users find CFish to be faster than AsmFish whilst others find AsmFish to be faster. ![]() Is a very similar idea to AsmFish, except the C language is utilised. For those who want a 100% congruence with Stockfish a Pedantfish version exists. This does not mean that the engine searches 20-30% deeper but conducts the same search 20-30% faster which is very useful in analysis! Asmfish gives the same result a Stockfish in 99% of situations due to slight coding differences. (If this is gobbledygook to you dont worry!) The outcome of this is that AsmFish runs around 20-30% faster than Stockfish or Stockfish Development Versions. This is a rewrite of Stockish in Assembler, rather than the C++ Stockfish is coded in. Most of the Clones/Copies of Stockfish fall into two areas - either ones which try and increase its speed or ones which change either its evaluation or search functions. Common opinion is that we should wait a week or two until after release as very occasionally new versions are rolled back. Typically each improvement results in a tiny elo gain but when added together these become significant. Each ones undergoes rigerous self testing where the Develop version plays potentially tens of throusands of games against the current version - to have been released it needs to have shown statistically better results. These are small improvements which are released every few days. This is a solid place to start and all other versions should be compared to this. This is the "vanilla" version of Stockfish, of the shelf with releases every 6-9 months. You would not compare 10,000m runners by looking at their 100m times and the same is true here - be very wary of rating lists. One word of caution, there are a number of rating lists which rate these versions almost all have a common failing - the test games are run at very short time controls. A side effect of this development is that a number of different versions, clones and copies exist some of which may be of interest to the aspiring correspondence chess player. Its ongoing open-source communal development continues to make it stronger. The Stockfish Chess Engine is undoubtedly one of the strongest available at the moment. Russell Sherwood Tuesday, February 21, 2017
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |