My original goal is this post was to combine two thoughts that have been rattling around my brain for a while and that I thought were ultimately linked together. However, once I fleshed both out, they were really too long so I decided to split it out. However, in my mind the two thoughts remain inextricably linked, and to reflect this I'm treating it as a two part series.
PART I: HOW CASEY JANSSEN BECAME AN ELITE RELIEVER
It's no secret that I (among many other Blue Jay fans) was not a big fan of John Farrell's management, and in particular his bullpen management. I was constantly confounded by the situations in which he used his best and worst relievers, and how many games were lost with his best relievers left in the pen to protect leads that never came while the game was lost with lesser pitchers on the mound. But that's only one part of bullpen management among many, and in particular leveraging platoon advantages through playing match-ups is an important one. Though I didn't think Farrell was very good at this either (see: Dotel, Octavio; early 2011 usage thereof), I had never really looked into this aspect. So in the run-up to Farrell bolting for Boston, I dug into it (more on this later perhaps in the future, it was actually pretty interesting).
One of the things that surprised me the most was the nature of Casey Janssen's improvement over the past couple seasons. We all know that he emerged from being a decent middle reliever to successfully holding the closer's role this past season, but what's really interesting is exactly how he improved. In my initial screens, I was looking platoon usage in 2011-12 under Farrell, compared to the previous three years. Here is what it looked with for Janssen
Two quick notes on methodology. First, a positive number indicates a typical platoon split given the reliever's handedness, a negative number is a reverse split. Second, to analyse platoon splits, you first have to quantify the platoon split, which in turns depends on how you measure pitching. I have determined platoon splits in terms of FIP (measuring only the three true outcomes) and in terms of wOBA allowed (measuring results on a linear weights basis). These represent the two ends of the spectrum, and I simply split the difference and take a simple average in determining the overall split.
Looking at that chart, from 2008-10, Janssen had a massive platoon split in terms of FIP and a pronounced albeit muted split in terms of wOBA. Overall, he was a quite decent option against RHP, but someone to keep away from LHB as he had an above average platoon split at around 26%. In other words, a run of the mill middle reliever. On the other hand, in 2011-12, things were extremely different. First, he was better, perhaps even signficantly better, against RHB, both in terms of FIP and wOBA allowed. But he was even better against lefties, holding them to a 2.55 FIP and .225 wOBA. That's LOOGY quality performance against LHB right there.
So why did that happen? What changed? Did he get better? Before turning to these questions, there was an issue I noticed. The data for 2008-10 could be not be very reflective of Janssen's actual talent. He entirely missed 2008, so that's reducing the sample size. But in 2009 after missing time at the beginning of the year, he made 5 starts before another DL stint and then coming back for the last two months as a reliever. So not only is the sample smaller than usual due to injuries, but the 2009 data is almost two-thirds starting pitching. And unfortunately, there's no data source that will give split splits (you can have L/R splits, or starter/reliever splits, but not both). So I was unsure if there was actually an improvement to explain in the first place.
Fortunately, Fangraphs does have a play log. So I copied the 2009 play log data into Excel, and built a sheet that would parse the data, separate out SP innings from RP innings, and calculate splits to that I could look at his splits as a reliever only
Even removing the starting pitching innings, the same type of split persists (big FIP split, small wOBA split) and of essentially the same size at 29% versus 26%. As a further improvement, I decided to extend the data to include Janssen's 2007 campaign, his first season as a reliever (in the middle column). I think this is the best basis for comparison since it considers entire sample of relief innings, and what we see is that the FIP split is much smaller, such that the overall split is about 7%. This is much closer to normal for a typical right-handed reliever.
But the broader story is still the same: early in his career is a reliever, Janssen was good against RHB, but only okay at best against LHB. It was fine to have him face a lefty, but you wouldn't want to use him if a couple LHB were due up in the first 3 or 4 batters. But since 2011, that's completely changed. He's been better against righties, but remarkably tough against lefties to the extent of having pretty large reverse splits. So it is just a small sample, or can we expect him to carry this performance level forward*? I turned to the Pitchf/x data at Brooks Baseball to answer the question.
*Whenever a pitcher has reverse splits, especially with a smallish sample size, regression towards normal splits is expected. I refer to him continuing to be a much better overall pitcher.
What Is Janssen Doing Differently?
At the outset, it's worth noting that Janssen has a more diverse arsenal than the average pitcher, as he uses or has used six pitches: four seam fastball, two seam fastball, cutter, slider, curveball, and change-up. Even as a reliever, he has used all six at various times.
In 2007 against LHB, he used his array of fastballs about two-thirds of the time, favouring a cutter that averaged around 90 MPH and mixing in two seamers early in the count as well as a few four seamers, both averaging 93 MPH. Against RHB, that mix was similar, with the two- and four- seam use reversed. In terms of off speed pitches, against LHB he favoured his curveball, while occasionally using change-ups and sliders (7% each). One interesting tendency was using the curveball to pitch "backwards", as he used it in the first pitch 23% of the time, but ditched it when behind in the count. Sliders, on the other hand, were used mostly ahead or with two strikes, indicating it was his put away pitch. Against RHB, he preferred the slider (23%) and mixed in the curve (11%) mostly with two strikes or ahead, completely ditching the change-up.
If we skip ahead to 2010, the pitch usage was almost the same, with the major difference being that he used the cutter a little less against LHP, and increased the use of his four seamer. The velocity on his fastballs was down about 0.5 MPH. But by and large, Janssen was using a similar arsenal of similar quality, and a similar approach in 2010 as in 2007.
In 2011 however, Janssen's pitch mix changed quite significantly, changes largely carried through to 2012 and which were especially pronounced against LHB. In previous years, Janssen used a change-up against them, if sparingly. It was gone entirely. Likewise, he had used his slider against LHB just under 10% of the time, mostly when ahead or with two strikes (15-20% in these situations). It too was gone entirely. He had used the sinker almost 20% of the time against LHB, mostly early in the count. It's use was shaved down closer to 10%. Instead, Janssen relied increasingly on his four seam fastball, its use vs. LHB rising from 13% in 2007 to 23% in 2010 to 31% in 2011, and finally to 40% in 2012. He also used his cutter a little more.
Against RHB, the changes were not quite so dramatic, but still significant. His slider use fell steeply from 23% to just 9% in 2011-12. In its place, used his curveball a little more, 14% instead of 11%, with a particularly large increase in its use with two strikes. Instead, he used his cutter a lot more, from around 30% to close to 40%, but that fell back to 26% in 2012. As against LHB, four seam fastball use increased significantly, from 27% in 2010-11 to 48% in 2012. In particular, he used it almost 60% of the time on the first pitch.
What was the effect of these changes? First, scaling back the mix had positive effects. Janssen's change-up was frankly just not very good. It's a small sample of only 107 pitches, but career against LHB it ended up being a ball over 50% of the time, while inducing whiffs at a below average rate of 18% and giving up a lot of fly balls. Change-ups are usually effective weapons against opposite handed batters, but for Janssen it was simply not a good pitch and counterproductive to use. Likewise, his slider was just not very effective against LHB, only generating 14% whiffs and resulting in a 40% line drive rate when put in play (again, SSS alert with only 20 balls in play). This is especially bad when considering that he used it mostly when ahead in the count.
Next, one of the most important skills for a high leverage reliever is the ability to generate swings and misses, particularly if there are runners on base. Below is a chart showing Janssen's ability to miss bats:
As context, the MLB average for relievers is about 21.5%, last year it spiked to 22.5%. What we see is that on three of Janssen's pitches, his whiff rate has increased in a significant manner: his four seamer, his slider and his curveball. The increase on the slider is due mostly to not using it against lefties, as that pitch had a large platoon split. Having reviewed the data, the modest bump in the whiff ratio on the curveball is largely due to it being used more late in counts when batters have to protect rather than early in the counts when batters can be more selective about choosing to swing or not.
That leaves the four seam fastball, and it seems a rather curious case to me. It's not about velocity, as his average four seam velocity has stayed between 92 and 93 MPH. It doesn't appear to be about how he's using it. But it does appear that it did take a step forward.
Finally, by going with more four seam fastballs, especially early in the count, Janssen is getting ahead in the count more often. When the first pitch is not put in play, Janssen gets an 0-1 count roughly 60% of the time off his four seamer. This compares to the two seamer in the low 50% range and the cutter in the high 50% range. There is a tradeoff, however, in that four seamers generate significantly fewer ground balls. Over Janssen's career, his four seamer has a ground ball rate of 37%, versus over 50% for both the two seamer and cutter. In fact, the increased use of the four seamer in 2012 may reflect having assumed the closer's role: a middle reliever is more likely to come in with runners on, and a ground ball can generate multiple outs; whereas in the closer's role the pitcher typically enters with no runners on and there's less of a premium on ground balls.
Casey Janssen's emergence over the past couple years is largely due to him erasing and reversing significant platoon splits he had earlier in his career as a reliever (in addition to improvement versus RHB as well). This in turn was driven by a more effective four seam fastball, the increased use of that four seam fastball vis-a-vis other types of fastballs, and the refinement of his pitch mix to eliminate change-ups and sliders against LHB. Finally, these improvements appear to be sustainable in that they are based on real changes and not smoke and mirrors.
PART II: CAN CASEY JANSSEN TELL US ANYTHING ABOUT BRETT CECIL'S FUTURE?
Whenever I think about Casey Janssen, and his conversion from mediocre starter to middle reliever to effective high leverage reliever, I inevitably come around to thinking about Scott Downs. Both were initially HR-prone starters who struggled to get opposite handed batters out and had significant platoon splits. Both became very effective relievers meriting high leverage innings, but without the knock-out raw stuff typically associated with closers and set-up men.
After Brett Cecil lost velocity in 2011 and it became apparent that he was not an effective starter with reduced velocity and the thought occurred to me that if the velocity didn't come back, Cecil could potentially fit a similar profile. HR-prone starter? Big check. Platoon splits? Check. Lacking knockout raw stuff? Check. Cecil has been a target of wrath for Jays fans, but convert him to relief, get a tick up in velocity in shorter outings, thin out the arsenal, and maybe you end up with a valuable piece. In the next part (much shorter), I will flesh out this comparison in more detail and see what inferences we might be able to draw about Cecil's potential future value.