WCC has two kinds of databases. The first is
commonly referred to as a win/loss/draw or game theoretical value database
(GTV for short). A GTV database tells a program whether or not a position
is a win, loss, or draw. Unfortunately, this information is only "theoretical,"
and provides no method for actually playing a position to its final conclusion.
GTV information is compact, although files
are still extremely large for the 8-, 9- and 10-piece databases. Because
of a special technique known as run-length encoding it is possible
to "stuff" the results of many different positions into a single byte. WCC
contains an average of 30.741 positions per byte in its 4x4 subset of the
8-piece database. As a result, all 111,387,534,401 positions only require
3,623,115,148 bytes, or 3.37 GB. For more information about how this compares
with other programs, click here.
The position above demonstrates how even a
6-piece database program can win a 12-piece position without any difficulty.
With black to move, a program will generate not only moves like 6-10 and
2-7, but those that seem to toss away material as well, such as 20-24 and
14-17. Eventually it will "stumble into" a 6-piece position after generating
the sequence of moves 14-17 21x14, 20-24 28x19, 6-9 13x6, 2x9x18x27. This
position is shown below on the left. With white to move now, it is a draw.
This result means that the black-to-move position shown above will draw
if that sequence of jumps is executed.
Next, the sequence 14-17 21x14, 20-24 28x19,
6-9 13x6, 1x10x17x26 is generated. That position is shown below on the
right. In this position, the 6-piece database says white to move loses.
Therefore, the black-to-move result from the parent position is a win!
For this reason, a GTV database, which is consulted as moves are generated
during the search, is a tremendous asset to a program. You can see from the
positions below that it is not immediately obvious that one position draws
while the other loses.
From these examples you can see how a GTV database is of great
value, even when the parent position contains many more pieces than the
database being probed. The irony is, once you enter a position that has
the same number of pieces as the GTV database, there is no information
as to how to win the game! Remember, the only information contained in
this kind of database is win, loss, or draw. This does not tell you which
move leads to the quickest win, nor does it help at all if every move in
the parent position is a win.
This is why Gil Dodgen and Ed Trice invented the perfect play
lookup (PPL) database for the game of checkers. The PPL database is
unique to WCC. No other program has it! (NOTE: Do not confuse the PERFECT
PLAY LOOKUP database with the vastly inferior CONVERSION database. Click
here to see why conversion databases
are of very limited value.)
The PPL database allows WCC to announce the distance to the
actual end of the game (assuming both sides play flawlessly), from any
position within the database, without having to conduct a search. Currently,
you can buy the 7-piece perfect play database set, which has the capability
of announcing a win from a distance as great as 253 plies (a total of 253
moves for both sides).
Are PPL databases really valuable? Could it be possible for
a program to be in a won position, and not win it? The answer to this question
is a resounding "Yes!" WCC was able to hold both the Wyllie program and
the Kingsrow program to a draw, even though WCC allowed them to play from
a won position. When playing the Kingsrow program, WCC took over from the
point where Kingsrow conceded the draw, and subsequently won the game with
no trouble at all.