... by^1.1

The program also contains code written by Peter W. Gillgasch (see ``Acknowledgments'' for details).

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... & ~y),^1.2

ANSI-C syntax is used for all examples throughout the article.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... lookups.^1.3

A clear explanation of how this bit-fiddling trick works lies beyond the scope of the article.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... practice.^1.4

Although Buro developed and tested PROBCUT within his master-strength othello program LOGISTELLO it should work in computer chess as well.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... time-controls.^1.5

Private and public discussions with authors of other chess programs confirm this notion. During the last three WMCCs DARKTHOUGHT seemed to always search among the smallest trees of all participating programs.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... recognized.^1.6

Please note that mates (as read from distance-to-mate endgame databases, for example) require special care, too, because standard quiescence searches (whether captures-only or checks included) can normally not return any real mate scores.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... Pawns.^1.7

The only valid yet practically irrelevant counterexamples that we know of are positions with more than two equally coloured bishops for any side.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... search.^1.8

We estimated this number because the original text only gives the winning percentages but not the absolute rating differences.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... with.^1.9

Hyatt and Newborn will most probably correct their publicly available files, too.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... subsets.^1.10

Of course, the lower numbers of ``(I - 2) Best'' moves caused the ``Fresh Best'' rates to increase and the ``(I - 2) Best'' rates to decrease w.r.t. the overall ``Best Change'' amounts.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...THOUGHT.^1.11

The complete yet totally raw result files of the experiments with both CRAFTY and DARKTHOUGHT as well as our short Perl script for the automatic analysis of the result files of CRAFTY are publicly available in gzip'ed TAR format at URL <ftp://i41s10.ira.uka.de/pub/ernst/deep14.tgz>.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.