Here’s a little bit more explanatory background regarding a central concept of my statistical analysis: The Random Ball Machine.
There are ball machines used in baseball, but in terms of the “randomness” concept that I’m employing, it’s probably more helpful to visualize the kind of ball machines used in tennis practice.
The more advanced tennis ball machines have a “random” setting, so that they can spray the ball back and forth and up and down — giving the practicing player a variety of balls to which to respond. (Of course, you’d always set the tennis ball machine to hit the ball over the net, so you have to imagine it set to shoot the ball on the ground randomly as well.)
And here’s a short video of one in action:
Now that you have that in your head, consider that our “baseline” for hitting performance is the “Random Ball Machine.”
The Random Ball Machine has a batting average on balls in play (BABIP) of .298 (which is the 10-year MLB average). So if every plate appearance was handled by the Random Ball Machine, every ball would be put in play and the result would be 29.8% singles and 70.2% outs. Every player would have a batting average, on-base percentage and slugging percentage of .298. No walks, no strikeouts, no extra-base hits. A slash line of .298/.298/.298.
In my approach, rather than the mythical “replacement-level player,” the baseline is the mythical “Random Ball Machine.”
If you recall, we identified the key elements of baseball as (1) avoiding outs (or getting outs for pitchers); (2) producing offense (or denying offense for pitchers); and (3) doing both at the same time.
So we have two hierarchies to consider.
For avoiding outs:
- Home run or walk (in this regard they are equal) (100% safe; 0% out)
- Non-HR well-hit ball (some percentage higher than 29.8% safe; remaining percent out)
- Random Ball Machine (29.8% safe; 70.2% out)
- Strikeout (0% safe; 100% out)
For producing offense:
- Home run (avoids out; avoids reliance on random-y ball-in-play; produces lots of offense)
- Non-HR well-hit ball (reduces chance of out; likely to produce considerable offense)
- Walk (avoids out; avoids reliance on random-y ball-in-play; produces minimal offense)
- Random Ball Machine
- Strikeout (guarantees out; eliminates balls-in-play)
In both cases, the strikeout is worse than the Random Ball Machine, but every other outcome is better than the Random Ball Machine.
Our stats look at the world like this: what if you examined that certain portion of a player’s plate appearances in which he is eliminating the Ball Machine? How much does he “move the needle” one way or the other from what the Ball Machine would do?
For each category, we looked at the 10-year MLB average. How much better is the average MLB player than the Random Ball Machine? That’s the value that we set at “100” for our stats.
Needless to say, we do not preclude the second-stage argument: for some players the Ball Machine is not Random.
In fact, going in, we acknowledged that Ichiro’s Ball Machine was not Random at all, and that he would be systematically under-valued by this approach.
We now acknowledge, as well, that there players for whom the Ball Machine is non-random in the other direction, and, perhaps, Edwin Encarnacion is an example. His career BABIP is .275. Perhaps he is systematically over-valued by this approach, and bsf makes a good argument that he is.
Again, however, the argument is a second-order one: there is a specific reason why this player shouldn’t be compared to the Random Ball Machine.
More to come.