StupendousMan said:
There's quite a range of speeds in the pitches thrown during any one game. Ian's graphics show a thick box, with thin little whiskers sticking out above and below. The vertical extent of the box shows the range which encompasses 2/3 of the pitches, which is a pretty standard way to describe the distribution. A good, simple rule is to compare these thick boxes from one game to another. If the boxes overlap each other vertically, then the two games have pitch speeds which are effectively the same. If the boxes don't overlap, then you can make a strong claim that there is a significant difference from one game to the next.
Yeeesss ... but we need to think about what we mean by "significance". There's statistical significance, which is an arbitrary concept often (though not always) set at p < 0.05, and there's functional significance, which can be very different. For example, we could have a hypothetical pitcher whose fastball ranged from 88 to 98 mph, averaging 93, compared to one whose fastball ranged from 92.5 to 93.5. They'd be statistically the same -- there would be no statistical significance to the difference -- but there could be a huge functional difference in how hard their pitches are to hit. Conversely, we could have two pitchers with almost no variation in their fastball speed, averaging out to 94 mph and 94.1 mph, who would achieve statistical significance because they have almost no variation in velocity, but there would be little or no functional significance to the difference.
(This is all super-simplified -- statisticians aren't that naive -- and I know Stups is already familiar with this, but it might help for background. )
We also need to remember the difference between descriptive and predictive stats. The charts above are descriptive -- they take phenomena that have been measured and counted more or less precisely. They are not a sample of a larger population, they are the entire population we're looking at. That means we can think about them in different ways than if we were sampling from a larger population, and we can interpret them in different ways than if we were using them to predict the future.
For example, the common and simple statistics that are usually used help correct for errors in sampling, and errors in measurement. We might be willing to argue that in this case neither is an issue.
(Again, super-simplified, we're ignoring many things here.)
Because these are descriptive and comprise the entire population, it's not unreasonable to just look at the mean and variation and use your intuition to tell you whether they're different or not. Is the average faster? Are the peaks higher? Are the valleys higher? Then it's not incorrect to say that he was faster in this game, even though a Student t test might say the difference is not significant.
But then if we start using these for the purposes of prediction, everything changes. Now, for example, we're drawing samples from a larger hypothetical sample of all Masterson's potential pitches, so we do have to worry about sampling error. We could look at today's game and made predictions as to the average and range of his pitches in next game, and we might say that his results would be statistically identical, and we still wouldn't be surprised to see boxes and whiskers that looked somewhat different.
So in conclusion: Yes.