Table 1.14 lists the ``Best Change'' rates as originally
published by Hyatt and Newborn in 1997 [114]. Because
Hyatt and Newborn did not provide any explicit numbers for the 107
middlegame positions alone, we calculated them as described in
Appendix 1.5.11 below. The numbers in parantheses represent
the rates that our automatic Perl script computed when we applied it to
Hyatt and Newborn's original result file of CRAFTY for the full set
of 347 test positions. The slight differences do not affect the overall
interpretation of the results and we regard them as constituting some
omnipresent imprecisions of measurement. Please note that we debugged
our Perl script thoroughly, performed many cross-checks during its
execution and really trust its automatically generated output.
Results of Crafty for the Middlegame Positions in 1997
Unfortunately, Hyatt and Newborn did not present the experimental
results of CRAFTY for the 107 middlegame positions alone in their
article. Therefore, we derived the ``Best Change'' rates of CRAFTY
for these positions by ourselves from the available data of the other
positions. The absolute number of new best moves as chosen by CRAFTY
in the middlegame positions equals the number of new best moves for all
test positions minus the new best moves for the opening positions minus
the new best moves of the remaining positions.
Table 1.15 shows our according calculations in detail
and lists the resulting ``Best Change'' rates of CRAFTY for the 107
middlegame positions from iteration #2 to iteration #14.
|