Learning to love the K% stat

How many batters did Sean Doolittle strike out in his All-Star 2014 season? There are many ways you can answer that question. You could say he struck out "a lot" of hitters, if you just want to keep things simple. If you wanted to put a number on it, you could say he fanned 89, which is a wonderful quantity for a reliever. Perhaps you could add context to that number, and mention that he accrued that total in 62⅔ innings. Grab your calculator and take it one step further, and you'll find he finished with a 12.8 K/9 (strikeouts per 9 innings).

However, there is an even better measure than K/9, and the only thing stopping it from being the norm is that, well, it's not yet normal. That metric is called K%, which stands for strikeout percentage (or strikeout rate), and it cuts to the chase by seeing what percentage of total batters a pitcher struck out. The calculation: strikeouts divided by batters faced, which seems simple enough. Doolittle struck out 37.7% of the batters he faced in 2014, which was the seventh-best K% among qualified relievers that year.

Why use K%?

Why is K% a better strikeout metric? To answer that, I'll paraphrase the example given in the FanGraphs glossary. Let's imagine that Pitcher A has the following inning: groundout, single, walk, walk, strikeout, strikeout. In other words, he loaded the bases but got out of it with a pair of Ks. Pitcher B does this: groundout, strikeout, strikeout. A 1-2-3 inning! Each pitcher tossed a scoreless frame, and each fanned a pair of hitters for an 18.0 K/9 rate. But one pitcher faced six batters (for a 33 percent K%), while the other only faced three (for a 67 percent K%). As you can see, there is a big difference in the strikeout frequencies of these two pitchers, and K/9 misses it entirely.

Now, that is an extreme example based on the tiniest of sample sizes. When you stretch things out to a full season, you won't get discrepancies like that, and K/9 will still work for the most part. But the fact remains that, if your goal is to find out how often a hurler strikes out batters, you will get a more accurate picture from K%, and you may not immediately know the instances in which K/9 is telling you a slightly skewed story. And once you've come to that realization, why would you go back to the old way?

The obvious answer is that you might simply be used to the old way, and that's a great reason. You know that 200 strikeouts is a great number for a starter and that 100 is mighty impressive for a reliever, and you know that a strikeout per inning (or a 9.0 K/9) is a wonderful mark for anybody. The only thing stopping you from switching to K% is not really knowing what a good or bad one looks like. So, let's clear that up right now.

What does a good K% look like?

To begin, the 2015 MLB average K% was 20.4%. That's the league-average mark for all pitchers who appeared last season. If you want to stop right there, then that's fine: 20% is average, 15% is about as low as you want to see (a contact-oriented guy like Graveman), and 25% is wonderful (a strikeout-per-inning guy like Lester or Price). Sonny Gray was almost the definition of average in this sense, at 20.3% (his greatness comes from other things besides just strikeout totals). Clayton Kershaw led the league, at 33.8%.

You can also take it a step further. Relievers strike out more batters than starters do, since starters have to think about their efficiency so they can go deep into games. The league averages last year: relievers were at 22.1%, and starters were at 19.5%. Now Sonny is a bit above-average for a starter! The best among relievers is Aroldis Chapman, who led the league in 2014 with an insane 52.5% and then did so again last year at "only" 41.7%.

What about the different leagues? NL starters get to face the opposing pitcher a couple times, which helps their rates: AL starters were at 19.1%, with NL starters at 19.9%. That difference almost entirely disappears for relievers, though, since by the late innings the pitcher's spot is usually replaced by a pinch-hitter (the difference was AL 22.0% vs. NL 22.2%). Here's a handy reference table for you:

* Dan Otero is an example of a guy who went into the danger zone but still succeeded, at 12.9% in 2014. But then, it sure caught up to him in 2015, which makes him an exception that proves the rule.

If you want to compare across eras, you also have to consider that, like any other stat, this one changes over time. Hitting 40 homers in 2000 wasn't the same as hitting 40 homers in 2015, because offense has gone down quite a bit. Likewise, strikeout rates have gone up over time, both because offense has gone down in the last decade and also because offensive strategies have changed over the last several decades. Pitchers throw harder than they used to, and more batters are willing to trade some chance at making contact for a greater chance of hitting a dinger. Go back to 2010, and the league average K% was only 18.5%, two full percentage points below 2015. In 2000 it was 16.5%, in 1990 it was 14.9%, and in 1980 it was 12.5%. (It was up over 15% in the 1960s, before the mound was lowered.)

What do the A's look like?

One great real-world example of the difference between K% and K/9 is Oakland's R.J. Alvarez. The righty is known for being a hard thrower with huge strikeout ability, and he lived up to that name by racking up 10.35 K/9 last year. However, he was terrible in almost every other way, allowing tons of walks and hits, and so his innings tended to extend for lots of total batters. Because of this, his K% was only 23.0%, barely above league average and far from the team lead.

The top of the charts in Oakland last year was essentially a tie between Evan Scribner and Fernando Rodriguez at 26.9%, with Doolittle just behind at 26.3% in his injury-shortened campaign and Ryan Dull at 24.2%. However, several of the team's new additions should join that conversation, including Rich Hill (34.0% in a tiny sample last year), Liam Hendriks (27.2%), Marc Rzepczynski (26.0%), and John Axford (24.8%).

***

I'm bringing this up now because it's a metric we use with increasing frequency on AN. As time goes on, you're going to keep seeing it referred to here and elsewhere, so you may as well learn how to interpret it.

What do you think? Are you convinced to give K% a try if you haven't already, or are you going to stick with K/9 or some other measure of strikeouts? Share any remaining questions you might have in the comments!

All numbers courtesy of FanGraphs