Game Development Reference
In-Depth Information
A similar reason for the 0 to 10 range of the pain scale is the usability of the
information. While a doctor may respond differently if you report your pain at 7
rather than 6, he is unlikely to react much differently if your pain is a 61 rather than
60. This is the portion that is similar to
significant digits
in science and math calcu-
lations. Certainly, there is a numerical difference between 61 and 60, but what does
it really mean? If the granularity at which you are tracking a value is finer than the
granularity at which you are able to measure the value in the first place, the extra
data is
superfluous precision
.
Not Enough
On the other hand, a pain scale whose range was less than 10 may possibly not be
expressive enough. For example, if we change the scale so that “extreme pain� is 3
rather than 10, it is more difficult to express exactly how we feel. (“Well, doc… I'm
not a 1, and 2 seems a little high. I'm kinda somewhere between 1 and 2 but closer
to 2 than to 1.�) It also means that the doctor doesn't get as accurate of a picture of
our condition from which to work.
Interestingly, most of the pain charts show only 6 face icons as I did in Figure
13.4. The reason for this becomes startlingly clear when we attempt to determine
how we would “tweak� the faces to create ones in between the existing set. (There's
only so much expression that you can present with smiley faces!) This leads to a dif-
ferent granularity for the “smiley face� version of the pain chart than the number
line version.
However, as we will discuss later, it isn't necessarily a correct assumption that
we should limit ourselves to only the outputs that we can display such as in the case
of the faces. Using response curves, for example, we can always combine portions
of the range into one displayed behavior. We can still benefit from tracking the un-
derlying data without being concerned with what we can exhibit externally. This is
especially important in the game development world where we are typically limited
on which animations and actions we can display.
An extreme example would be a behavior that had only two outwardly visible
signs such as the “fight or flee� scenario we have referenced a few times. The response
curve could continue to track a more granular “degree of fear� (or some other such
value) and only trigger the change to “flee� when “degree of fear� reaches a partic-
ular threshold.
The essential secret to constructing a weighting scale, therefore, is to arrange it
so that there is enough granularity to express the detail you need to differentiate
meaningful differences—
and no more!
