Home * Search * Path-Dependency

Path-Dependency describes a situation when the same position gets a different value depending on a sequence of moves by which it has been reached. Whilst a few programs do this by design [1], within the standard alpha-beta framework path-dependency is considered undesirable, if impossible to avoid. Typical cases are:
  • the position can be treated differently if there is a chance of repetition or hitting into a fifty-move rule.
  • different move ordering can cause different result (different moves causing a cutoff), when it comes from the hash table [2]
  • if a program uses search extensions, search depth may be different (i.e. 1. d4 Nf6 2.c4 e6 3.Nc3 Bb4 4.Nf3 vs 1.d4 Nf6 2.c4 e6 3.Nf3 Bb4+ 4.Nc3, where only the latter triggers a check extension)
  • history-dependent heuristics, like some versions of LMR may prune different sub-trees.

See also


Forum Posts

2000 ...

2005 ...

2010 ...


External Links


References

  1. ^ Re: Methods to stably evaluate nodes? by Uri Blass, CCC, May 05, 2007
  2. ^ Tony Warnock and Burton Wendroff (1988). Search Tables in Computer Chess. ICCA Journal, Vol. 11, No. 1, pp. 10-13.

What links here?


Up one Level