A Graphical Look at Pitcher Types

This afternoon, Mike Petriello put up a really good post on Dallas Keuchel’s breakout season. Included in that post was a graph plotting every 2014 qualified starter based on two variables: their groundball rate, and their strikeout rate minus their walk rate. Basically, the point of the graph was to show not just how extreme Keuchel’s groundball tendencies have been, but how rare it is for a pitcher to get that many groundballs while also getting strikeouts and limiting walks.

I gave Mike the graph after seeing that he had beaten me to writing a post about Keuchel’s emergence, but that’s not the only interesting data point on the chart, so I’m giving that chart its own post, highlighting some of the more interesting pieces of information that we can gain from plotting pitchers based on those variables.

Before we get to the graph, however, a quick explanation about why the vertical axis is strikeout rate minus walk rate, rather than the traditional K/BB ratio. As others have previously pointed out, K/BB ratio can often be problematic when a pitcher posts a very low number of walks, driving down the denominator and thus creating an artificial sense of distance between similar performances. For instance, David Price currently has a K/BB ratio of nearly 13-to-1, with Tim Hudson coming in second at 10-to-1. Both pitchers are walking fewer than one batter per inning faced, as is Bartolo Colon, who ranks 4th in MLB in K/BB ratio.

But while limiting walks is great, it isn’t the sole purpose of pitching, and many great pitchers trade a few extra walks for a lot of extra dominance. No one actually thinks Colon is dominating the strike zone more than Jose Fernandez was, after all, even though Fernandez’s K/BB ratio is “just” 5-to-1, ranking 7th overall in MLB. Fernandez’s strikeout rate was twice that of Colon’s, but because the low walk rate is the denominator, Colon’s K/BB comes out ahead when dividing them.

If you instead subtract walk rate from strikeout rate, you get something that more realistically aligns with what we know about pitchers dominating the strike zone. Fernandez is now #1, with a strikeout rate that is 28% higher than his walk rate. Masahiro Tanaka comes in second at 26%, with Price not far behind at 25%. The other leaders in K% minus BB%? Jon Lester, Max Scherzer, Zack Greinke, Yu Darvish, Stephen Strasburg, Johnny Cueto, and Michael Wacha. These are great pitchers, and they’re great pitchers because they dominate the strike zone. Subtracting walk rate instead of dividing by walk rate gives us a better list of who is actually owning the strike zone, rather than simply exalting those who avoid walks altogether.

While K% minus BB% isn’t going to be as immediately familiar as K/BB ratio, it’s pretty easy to adjust to the scale. Because league average K% is 20.4% and average BB% is 8.2%, the league average K%-BB% is 12.2% so far this year. Anything over a 15% difference is pretty good, and getting over 20% is reaching ace-type levels.

But dominating the strike zone isn’t the only way to succeed, and a pitcher who avoids walks and gets a ton of strikeouts is probably making a trade-off of allowing more fly balls, because in-zone contact rates are lowest on pitches at the top of the strike zone. So that’s where GB% on the horizontal axis comes in, and why we’re plotting the metrics that make up xFIP on a graph. By putting the strike-zone dominance on one axis and groundball rate on the other, we can see not just a pitcher’s overall result, but how they are getting those results.

So, here’s that same graph from Mike’s piece, only with the gridlines adjusted so that you can see the four quadrants.

You’ll note that the graph isn’t centered around the average or median points, as the midpoint of GB% is set to 50% and the midpoint of K%-BB% is set to 15%, with both marks representing points above the league average. The point isn’t really to show the big cluster of guys in the middle, but to demonstrate the buckets by which we often describe pitchers. Guys with a groundball rate north of 50% are often referred to as groundball pitchers, and as you can see from the few number of data points to the right of the vertical axis, there aren’t actually that many “groundball pitchers” among MLB starters. Even more rare are pitchers who get a lot of groundballs and rack up strikeouts.

That upper right quadrant is basically the nirvana of pitching. The perfect pitcher would live as far up and as far right as humanly possible, though clearly, being at the top of the chart is more important than being at the far right. As a pitcher moves right, he also moves downward — to get a lot of groundballs, you have to pitch in higher contact parts of the zone — so it’s always a question of making the most of what each individual pitcher has, and maximizing the total value of all three events.

But a pitcher who can miss bats while also generating a lot of groundballs can be quite good, and that’s why I’ve highlighted all the names of the data points in that upper right quadrant. Keuchel is on the island by himself on the far right side, with Tim Hudson being a less extreme example of both groundballs and strikeouts, and then the rest of them clumping closer towards the 50% GB% mark. A few of those names probably won’t surprise you, as guys like Cueto, Yordano Ventura, and Nathan Eovaldi have generated a lot of headlines this year. But don’t overlook what C.J. Wilson has done as a reason why the Angels are contending again, as he continues to look like the Angels best free agent signing in recent years.

And then there’s Brandon McCarthy, perhaps the biggest outlier of them all. You can see the names of the guys he’s pitched like, on a K%, BB%, and GB% basis. Here’s the ERA- for each of the pitchers in that upper right quadrant.

Cueto: 34
Ventura: 59
Hudson: 61
Keuchel: 74
Wilson: 82
Eovaldi: 99
McCarthy: 132

Of the seven pitchers in that bucket, only McCarthy has a below average ERA, and he’s not even close to the average. The problem, as has been pointed out, has been balls flying over the fence; McCarthy’s HR/FB ratio is 21.4%. Getting groundballs is primarily good because it limits the amount of home runs you can allow, but if you’re getting groundballs and still allowing home runs in the process, it’s not all that helpful. The good news, of course, is that HR/FB ratio is barely predictive in large samples and basically not predictive at all in small samples, so there’s no reason to think McCarthy is going to keep giving up home runs at this rate. If he keeps getting groundballs and strikeouts while limiting his walks, positive results will follow, as they have for every other pitcher in that area.

The problem is that what he’s doing is very hard to sustain. What all of them are doing, actually. For reference, here’s the same graph, but with the numbers from last year instead of this year.

Note that there are exactly three pitchers in that upper-right quadrant, with a fourth just sneaking in below the K%-BB% line. Felix Hernandez, Stephen Strasburg, and A.J. Burnett were three of the best pitchers in baseball last year, but it’s a lot easier to tell someone to just pitch like those guys than it is to actually pitch like those guys. It takes pretty incredible stuff, and sustained command of that stuff, to get batters to swing and miss while also pitching in the parts of the zone that generate grounders.

So while McCarthy is in line for some positive HR/FB regression, he’s probably also in line for some negative K%, BB%, and GB% regression, because generating these kinds of numbers over 200 innings is just very difficult. It’s good to remember that regression goes both directions, and a pitcher with a huge gap in their ERA and xFIP will probably see improvement in their ERA and and decline in their xFIP. The numbers will get closer together because both numbers will move back towards the middle.

As this post is already pushing 1,500 words, we’ll save the breakdowns for the other quadrants for another day. One quick note, though: Kyle Gibson, you might want to start getting guys to strikeout occasionally.