clock menu more-arrow no yes mobile

Filed under:

The 2015 Bullpen: When things go wrong at the worst time

Denny Medley-USA TODAY Sports

The Blue Jays wrapped up the first half yesterday afternoon by tying together a number of elements all too common in 2015 into an epic multi-faceted meltdown: a one-run loss with a poor start, questionable bullpen management, defensive miscues, base-running blunders and relievers ultimately just not being able to hold a lead. And it's that latter point that is worthy of further elucidation.

Going into last offseason, the overwhelming consensus - including from the highest reaches of the front office - was that the bullpen was a key area to target and prioritize improving. And thus when the season began with no additions other than some waiver claims and rookies from A-ball, the same wisdom said that as in 2014, the bullpen stood to be the team's biggest weakness in 2015.

But a funny thing happened along the way from Opening Day to the All-Star: amidst a great deal of roster shuffling, a number of quality arms have emerged, popping 95s and higher on the radar gun like it was nothing special. Far from being bad, by a number of metrics they've been decent-to-very good:

Raw Adjusted Rank
ERA 3.62 91 14
FIP 3.56 90 13
xFIP 3.49 90 4

In terms of actual runs allowed, the Jays rank in the middle of the pack. The peripherals suggest the Jays better than that, if allowing more home runs than would be expected given their batted balls. Another metric, RE24, also puts the Jays in the middle of the pack, at 15th in MLB. So in terms of allowing runs, the bullpen has been basically average, while potentially demonstrating an even better underlying talent level.

Of course, when it comes to bullpen all innings are not created equal: five scoreless innings in a blowout and two runs with a one run lead in the 9th might work out to a dandy 3.00 ERA overall, but it's distributed very sub-optimally. Runs are merely a means to the end of a win, so context and leverage matter, greatly. And so WPA is a necessary tool to account for leverage, even though it measures run prevention rather than trying to disentangle pitching and fielding  (so it's not directly comparable to any of the above metrics, though very close to ERA).

Here, the Jays' bullpen has been disastrous, their -3.38 WPA ranking the second worst in baseball. The only worse team is Oakland (-4.22), whose bullpen has been a disaster with a 115 ERA- (second worst in baseball). The only other team team remotely close to that WPA is Atlanta (-2.44), and they have the worst ERA- in baseball at 118. So intuitively, it makes sense that those teams would have horrible WPAs - but what's going on with the Jays?

I wanted to look at the exact relationship between WPA and adjusted metrics (ERA-, FIP-, xFIP-) since despite the caveats above, I would expect that broadly WPA (wins) should tend to track runs allowed. I pulled the team level data for 2002-14 (as far back as xFIP is available) from FanGraphs and charted them, with the ERA-WPA chart below:


As expected, the relationship is very strong, with a strong negative correlation: the lower a bullpen's era relative to league average, the higher the seasonal WPA, with ERA explaining almost 60% of the variation in WPA. For real-world data, this is very high. FIP- and xFIP- also had negative correlations, though somewhat less strong explaining only 45% and 32% of the variation in WPA respectively. This too makes sense, as they're more predictive in nature versus actual runs allowed. So we'll focus just on ERA.

Using this model, we can estimate what WPA is expected to be for a given ERA-. A team with an ERA- of 90 as the Jays have would be expected to have a full season WPA of +2.38 (+1.33 over 91 games), or almost five wins better than the actual -3.38. That's the difference between being in fourth place 45-46 and leading the division at something like 49-41 (technically tied, but I assume in this alternate universe that the Jays wouldn't have blown the second game of the season to the Yankees putting them at 47-41)

Historically, this is one of the biggest outliers. On the above chart, I added the 2015 Blue Jays as the enlarged bright blue data point. It's obvious that they're one of the most extreme outliers on the negative side, the 7th worse under-performance of 391 team seasons, two standard deviations below average.

But it's actually worst than that: WPA is a counting stat, and the 2015 Jays have only played 91 games whereas all other teams in the sample played full seasons. Extrapolating the Jays year to date results to 162 games gives a WPA of just over -6.00, which is the enlarged red data point. Visually, it can be seen that this would be the furthest from expected line, worse than the 2004 Royals who managed a -5.83 WPA despite a 93 ERA-. That said, much like Chris Colabello's .400+ BABIP, it's very unlikely this continues in the second half.

Why has this happened? It's basically comes down to bad things happening at really bad times. It's not just one or two pitchers, pretty much every reliever who's pitched more than 10 innings has had a couple bad blowups, and at certain times the defenders have had critical lapses. Some might use the term choking, some might use bad luck instead. I'll go with the more neutral snake-bitten. Basically, there's a certain amount of (random) variation in converting runs to wins, and in 2015 the Jays have drawn the very short end of the stick. And it couldn't have been a worse season for it to happen.

On the plus side, there's basically nowhere to go but up in the second half.