However, let's pause for a moment and zoom in on what I would argue is the most interesting success of the 2016 M's campaign so far - the bullpen. Consider where this bullpen was at before the season started:
- The closer was Steve Cishek, coming off a year where he got jettisoned by the Marlins from closing.
- The setup man was an aging Joaquin Benoit.
- Jerry Dipoto traded away the best reliever from last year, Carson Smith.
- The whole bullpen was going to be managed by Scott Servais, a man with no managing experience.
- Half the bullpen hit the DL in spring training, leaving the group perilously thin going into the season.
- Also, lest we forget the M's bullpen was pretty awful last year, so improvement was somehow needed out of this concoction of factors.
Somehow the mix has worked this year. Actually, it has more than worked; it has been flat-out good. It has been so good that it has led me to wonder more about bullpens and how to analyze them nicely.
Evaluating bullpens is trickier than other exercises in baseball because they are so context-driven. Relievers, unlike starters, come in with a relatively known game state. A ballgame has developed before they come in, meaning a manager's choices of who to pitch when actually says something about their managing acumen. I wondered if the M's bullpen was doing so well because of some wizard-like managing from Scott Servais. Was it possible that he was lining up the relievers in some crazy optimal way that made the group look better than it actually is?
I broke bullpens into the following components:
- Throwing hand - matchups matter, though they tend to be overstated
- Innings pitched - this measures who gets used the most
- Runs above replacement (RAR) - This measures the overall production of a pitcher. The more runs above replacement, the better they are.
- Cumulative leverage (pLI) - This measures the context a pitcher pitches in. A reliever pitching the 9th inning in a 10-1 rout faces virtually no leverage, whereas a reliever pitching the 9th inning a 2-1 nail-biter faces very high leverage. The leverage index (LI) measures how much the odds of winning change depending on the outcome of a play. Each play has a leverage index and pLI adds all of these plays together that the reliever is involved in. So, pitchers with higher cumulative leverages were used in more crucial situations.
The result is a scatter plot, with RAR on the x-axis, pLI on the y-axis, the size of points proportional to innings pitched, and the color of a point (really a bubble) determined by the hand a pitcher throws with. Here is what the 2015 Mariners bullpen looks like with this visualization:
Not much in the way of surprises in the 2015 bullpen. Most of the players live to the left of the black vertical line, which denotes exactly replacement level. No matter whom the M's threw at opponents in 2015 they weren't particularly good. What was Lloyd McClendon to do?
Carson Smith was far and away the best reliever on the team, and by the end of the season he was being pitched as such. McClendon stuck with Rodney too long, as evidenced by his large bubble floating above and to the left of everyone else. He was the worst possible combination - a bad reliever in high leverage situations used very often. Joe Beimel was also quietly a questionable choice by McClendon over and over again.
Still, overall, it's hard to fault Lloyd McClendon and the 2015 coaching staff for last year's bullpen woes. The overall trend of the graph is a line from bottom left to top right, which suggests better pitchers tended to pitch more crucial innings. McClendon did what he could with what he had.
I present last year mostly as a contrast to this season though. Behold, the 2016 Mariners bullpen of awe and wonder:
If anything, the Mariners bullpen is doing well despite Scott Servais! Steve Cishek has been solid in the closer's role, but he is only the M's fourth-best reliever. Vidal Nuno and Mike Montgomery are better than typical bullpen lefties and would benefit from expanded roles. Joel Peralta should be in middle relief at this point in his career. There is no linear trend at all in this data like we see with McClendon and the 2015 Mariners 'pen.
It will be interesting to see how the bullpen evolves during the year. It will get healthier, and Scott Servais is likely to rearrange roles to maximize the best performers in the most crucial spots. Maybe some regression happens, but it could be offset by better players coming off the DL and a bit savvier managing. There is no reason to expect this Mariners bullpen to suddenly implode. In fact, there are good reasons to think it could do even better as the year goes on.