"Alexander Belov" wrote in message
...
What about programs playing nearly all moves the same? Is it impossible
for programs that playing the same variant of opening with very open
and mutually terminating continuation where almost each move has
decisive meaning and ends up with a draw. If programs refuse to fail
in each move, there should be a suspicion that they are clones of
perfect engine.

Concerning List, what if to make List to play with several versions of
Crafty
from some test set of positions and see the difference with Crafty playing
against
itself from the same positions?

I think that the most secret place of the chess engine is its evaluation
function.
The rest of the code uses almost the same algorithms, so this shouldn't be
the secret.
The demands for the source code of the chess engines should be limited to
the
subset of source code that does not include evaluation (which should be
replaced with
dummy code). After investigation of this less secret code it can be
compiled
and binary
compared with the real executable image. They must be different only in
some
service
information and the part that corresponds to evaluation function. After
that
if program
does not have violations in the presented source code and plays different
game
of the suspected parent program, it MUSTN'T be treated as a clone.
Also if evaluation binary machine code can be extracted from the suspected
parent program, it can be compared against suspected clone evaluation
binary
code.

As a clear addition, the source code have to be compiled with the same
version
of compiler and its settings as a binary from tournament.
Comparison of evaluation code is possible under the same condition for both
programs.