According to our quantitative experiments with
razor margin = queen val,
limited razoring at pre-pre-frontier nodes exhibits no measurable loss in
tactical strength as compared with extended futility pruning alone. Yet
it shrinks the search trees by additional 5%-15% on average at
fixed search depths > 10 plies. Table 1.2 lists the
according performance data as obtained by searching all 2180 positions of
the ECM, WAC, and WCS test suites to fixed depths of 8, 10, and 12 plies
with the 1997 WMCC version of DARKTHOUGHT (see
Appendix 1.3.8 for a detailed description of the exact
experimental setup). The data shows that limited razoring hardly
influences searches of fixed depths <= 10 plies but then really takes
off at fixed depth of 12 plies. We observe even greater savings
(10%-15%) in other test positions with deep solutions which need
search depths >= 13 plies to be found.
Last but not least, we like to highlight the synergetic performance of
both extended futility pruning and limited razoring combined.
Table 1.3 presents the according data by comparing the
normal version of DARKTHOUGHT (see Section 1.3.5) with
one that is identical except for the addition of extended futility
pruning and limited razoring.
|
Nodes