Chess programs often look up the positions at the beginning of the game in an Opening Book. The opening database can then be used as long as the opponent plays a new move from the database, so usually more common opening lines will be stored to a much higher depth than the uncommon ones. As soon as the program is "out of book" it has to continue using the normal search routines.
The most advantage by using text form is that they are readable, understandable, and editable by humans and can be viewed, edited with normal text editors. However, the main disadvantage is that they may take too much space as well as time for searching since an item may take from 50 bytes (for FEN strings) to hundreds of bytes (for PGN text). If we store openings "continuously" (all possible opening positions) both data size and searching time become too huge and unacceptable for playing. They are usually missing weight values (to compare between items to know which ones are better) either. In practice, those opening books are usually very small with some lines or positions only, they are hard to be used for real game playing but for testing.
Stockfish Opening Book Ctg 16
Binary books have some important advantages, especially for computer chess engines: space-efficient, fast on accessing and searching, more useful information. At the most simple form, they are just a conversion of above text-opening books into binary, in the ready form for software and save text-binary converting steps. However, the major number of top books is built and worked with hash values from transpositions. Often each position is stored as a hash value (8 bytes) and some extra information such as the number of times occurred, number of games won by white/black / drawn with this position, average/maximum Elo of players playing to this opening position, chess program's success with the position. To save space those extra information typically about 2-8 bytes. For example, PolyGlot uses 16 bytes for each item, including hash, move, weight, learning values. Items of a book should be sorted by their hash. So when a chess program has to decide on a move, it will look up (by using binary-search) using the hash value of the current position to collect all information come with that hash value and then make a decision upon it.
All in all the information stored in an opening database should guide the chess game into a type of middle game in which the chess program proves itself most successful. But to achieve this there are two main ways.
If the opening database designer has some special preference of a certain opening line, he might create it fully manually. This is often the case if a chess program is about to play an important game against a grandmaster. Then the opening style will probably be chosen to be difficult and dangerous for the opponent. Another option is to first automatically generate a database and then fine-tune it manually.
Another option is to select a number of games (the type of which will affect the openings played: e.g. drawn games only, GM games only, etc) and then up to a certain depth store all positions in the opening database. It is true that by using this way of database generation some severe errors might by copied as well, but usually the chess programs are set to stop playing according to the book if a position occurred less than a certain number times before. 2ff7e9595c
Comments