Today we’re going to talk about one of Joe Morgan’s favorite topics: consistency. The opposite of “consistent” would be “streaky,” and one Brave in particular has been noted over his career for his streakiness. That’s Kelly Johnson, and he’s the inspiration for today’s chart.
The Braves never seem to know what they’re going to get out of Kelly. One week, he’s Rogers Hornsby, and the next he’s Mario Mendoza; rarely is he anything in between, and that hasn’t seemed to improve over his career. Kelly’s problem now is that the hot stretches are getting shorter, while the cold stretches keep getting longer. He’s having a pretty bad year as a result, and he’s probably hearing Martin Prado’s name everywhere he goes.
The metric for this chart is wOBA, which is explained here. The short version: it’s an improvement on OPS, scaled to on-base percentage. So, .300 is below average, .340 is average, and .380 is quite good. It probably needs a better name, too, because wOBA is a pretty meaningless acronym considering what the number actually means. I guess we’ll work with what we have.
To get each player’s “streakiness” from week to week, I did two things. First, I threw out weeks with fewer than 15 PAs. That’s roughly four starts, which is on the borderline for what I would consider “regular.” I also threw out players who hadn’t accumulated at least five such weeks, so that eliminated David Ross, Omar Infante, and a few others. Then, I compared the standard deviation of wOBA to the player’s overall wOBA (otherwise I’d be punishing players with higher wOBAs). That’s the percentage on the chart, with higher numbers representing greater streakiness.
I realize that there are some caveats with this kind of analysis. Half a season probably isn’t a large enough sample to weed out normal variations in wOBA, and it would be better to look at career numbers. Anyway, here you go. Stats are through Sunday.
You’ll notice that Kelly is actually second to the rookie, Schafer, in overall streakiness, but both of them (and Francoeur) have been all over the place. Escobar and Anderson, on the other hand, have been more consistent with their production. That’s a compliment to Escobar, not so much to Anderson.
I’d say that despite the small sample, this kind of confirms the perception of Johnson that we’ve had for some time. However, it’s likely that no one would care if his wOBA were .393 instead of .293.
jfwiii.net 
This is really interesting stuff to see.
Question 1: Is anything being park-adjusted here? I ask because I’m wondering how much of the variability might be due to playing in different parks from week to week. Then again, this should tend to have about the same effect for everybody.
Question 2: I can’t help but notice that the three hitters with the lowest wOBA ended up as the streakiest here. Would things look substantially different if you didn’t scale things by dividing by wOBA?
Unsolicited Theoretical Statistics Comment: You’re right to be especially worried about the sample size here. Whenever you’re comparing two (or more) standard deviations, conclusions can tend to be notoriously unreliable unless either (a) you have extremely large amounts of data or (b) you know something for sure about the distribution of the data (like that it’s normally distributed, for example).
The answer to question 1 is pretty simple – wOBA is park-adjusted, so that variable has been accounted for.
Question 2 is a little more difficult because I didn’t save the information I used for the calculations. I believe that McCann would have come out second in streakiness without dividing by wOBA, but it seemed like the way I did it was theoretically more fair.
I figured sample sizes would be an issue here. How much data would I need to get something even remotely reliable out of this? A whole season, or even more?