What is the purpose of the site?
To provide a quick and effective way to train backgammon
play. In our opinion, the only way towards stronger
backgammon play is evaluating one's move selections against computer
Is computer analysis so important for improving at backgammon?
Use of dice in backgammon makes it very difficult,
if not impossible, to tell whether a loss was caused by bad play or
merely by bad luck, unless a computer is employed for analysis.
This makes computer analysis much more vital in backgammon than
in other board games, like chess.
But I've heard that great players can judge a position without the help
Paul Magriel's Backgammon
is considered the bible of the game
and still, there are many positions for which the book-provided analysis
is now seen as wrong, and this is solely because of computer analysis.
Even the ways opening rolls are played
have changed significantly because of
computer analysis; for example, it was only after computer analysis that
it was discovered that playing the opening 6-4 as 8/2 6/2 is very likely
the best play. Before computer analysis this play was dismissed by most experts.
Okay, I am sold. How do I actually use the site for training?
It is highly beneficial to register first so the system
can track your rating. However, registration is not mandatory. Then, just navigate
to the home page and start playing positions. After each submission,
you will get the list of best moves, as evaluated by the computer, together with
their equity (more on equity later). Your move and its equity's difference
from the best move will be highlighted. The smaller the difference, the better
your selection was.
What is this equity you referred to?
In rough terms, the equity is how many points you expect to win from this position,
on average. For example, if in a position you have 60% chance to win one point and
40% chance to lose one point, then your equity would be approximately
0.6 * 1 - 0.4 * 1 =
0.2 (Note that in most cases gammons and backgammons are possible, and
thus the aforementioned equation would become slightly more complicated).
For the purposes of this site the equity would not be exactly 0.2, since the match
score is also taken into account; i.e. the equities presented along with the analysis
are an indication of how likely you are to win the match, not merely the game. Equities
of this type are called normalized.
The details are somewhat involved, but they do not matter
to players who simply seek to improve (and not to write their own analysis software).
What is this ply that appears in the analysis results?
The computer analyses candidate plays by looking ahead to the positions to which they lead.
The number of moves they look ahead is called a ply. So if a play is marked as 3-ply this
means that the computer has looked 3 moves ahead while analysing it. The higher the ply count,
the more accurate the equity calculated.
How is it possible that a move B be ranked lower than move A, even though B's equity is larger than A's?
We can explain this via an example. Suppose that at 2-ply it is:
In view of those results, the computer decides not to evaluate B at the
next ply, because it is clearly worse than A. It then proceeds to the
evaluation at 3-ply and finds that the equity for A is now 0.160. This
doesn't mean that B is a better move. In our example here, if the computer
would continue evaluation of B at 3-ply it would almost certainly find that
its equity would be even less than 0.160. Remember that equities can only
be compared at equal ply. In other words, ply trumps equity.