ollowed by a question as to what is the best thing
to do, becomes a problem, to which the answer is--_the best thing
to do_. Of course, in most strategic problems, there are so many
factors almost unknown, and so many factors only imperfectly known,
that we can rarely ascertain mathematically what is the best thing
to do. Nevertheless, there must be a best thing to do, even if we
never ascertain exactly what it is. Now in arriving at the decision
as to the best thing to do, one estimates the weight of each factor
and its bearing on the whole. If one estimates each factor correctly,
that is, if he makes no errors in any estimate, and if he makes no
error in summing up, he will make an absolutely correct decision;
and any departure from correctness in decision can result from
no other cause than from errors in his various estimates and in
their final summation. In other words, skill in strategy is to
be attained by the same process as is skill in other arts: by
eliminating errors.

So, when we take the decisions of the game-board and the war problem,
we must not allow ourselves to forget that there has been a tacit
assumption that the numbers and the skill of the personnel have
been equal on the two sides; and we must supplement our decision
as to the best material to be employed by another decision as to
how we shall see to it that the assumption of equality of personnel
shall be realized in fact--or rather that it shall be realized in
fact that our personnel shall get the maximum of effectiveness
out of the material.

In designing the machine, therefore, we are confronted with the
curious fact that, in general, we must design the various material
parts before designing the personnel parts that are to operate
them.

The most obvious characteristic of the personnel parts is that
the number of personnel parts shall be sufficient to operate the
material parts.

To ascertain the number of personnel parts, the only means is actual
trial; though naturally, if we have previously ascertained the
number of men needed to operate any kind of mechanism, say a certain
kind and size of gun, we can estimate quite accurately the number
needed to operate a similar gun, even if it differ somewhat from
the other gun. After the gun is tried, however, we may have to
change our original estimate, not only because the estimate may
have been in error, but because the requirement of operating the
gun may have changed. For instance, the re