This is the third installment of our series projecting the 2014 Yankees by comparing it to 2013 team, position by position. To see our take on the offense’s remodeling, click here. For the starting rotation, clizzick. Today, we look at the bullpen.

If predicting starting pitching is a fool’s errand, trying to predict relief pitching is Eduardo Nunez pushing Sisyphus’ rock up the hill without dropping it. Pitching stats take many innings before they become meaningfully predictive, and relief pitchers only accumulate small sample sizes in one season. On top of that, reliever roles vary highly as managers tend to fiddle with the pen throughout the year.

However, for this analysis, we’ll give it a try and compare the current stable to that of the 2013 campaign. Our stat of choice in this exercise is Win Probability Added (WPA). WPA is the difference between the team’s win probability when the reliever enters the game and when the reliever leaves the game (or when the game ends). So

For example, if the Yankees went into the 9th with a 2-run lead, they have something like a 95% chance of winning the game. If Mariano Rivera came in and struck out the side, he moved the team win probability to 100% and gets credited with a +.05 WPA. If Joba Chamberlain came into the bottom of the 7th with a one-run cushion and immediately gives up a two-run jack, the team’s win probability falls from 80% to 35%, and Joba is credited with a -.45 WPA. WPA provides a pretty good picture of what a player actually contributed to his team’s wins and losses, but it is a poor predictor of what he will do in the future.

Here’s how every Yankee reliever with a minimum of 10 innings performed in 2013, courtesy of Fangraphs:

nomaas yankees bullpen wpa

Last season was the Mo & DRob show, and Joe Girardi effectively utilized his best two bullpen weapons by employing them in the highest leverage situations (they ranked 12th and 29th in the Leverage Index, respectively). In fact, the two had remarkably similar and remarkably superb seasons. Rivera pitched 64 innings with a 2.11 ERA. Robertson pitched 66 innings with a 2.04 ERA. Both surrendered 15 earned runs, but Robertson was actually the more clutch pitcher. He tended to give up his runs in less important situations. Rivera’s runs happened to come in more crucial situations, resulting in 7 blown saves. We wouldn’t make much out of that, except to say that there’s no reason to think Robertson can’t handle pressure situations.

Adam Warren pitched the most relief innings for the Yankees and pitched fairly well (3.52 ERA in 69 IP), although it was mostly mop up work (.52 Leverage Index). As his slightly negative WPA indicates, he had a few costly blow ups.

After those three pitchers, Shawn Kelley contributed the 4th most innings (53). His 4.39 ERA wasn’t good, but he came up with enough key strikeouts for Girardi to keep sending him out there. The only other RP with 50 IP was Preston Claiborne, who was unremarkable (4.11 ERA), but effective enough so as to not cost the Yankees games.

Girardi went the recently-departed Joba Chamberlain quite often (42 IP), and it didn’t work out, as the former phenom pitched to a 4.93 ERA. The result was a -0.6 WAR, worst on the pitching staff. Fortunately, Joe was at least judicious enough to limit Joba’s exposure by using him in fairly meaningless situations (.67 Leverage Index), which limited his damage to the team.

As the team’s lefty specialist, Boone Logan was utilized in many crucial situations, with a Leverage Index of 1.10. He pitched well overall, finishing with a 3.23 ERA. However, when he puked up runs, they tended to be in late inning situations when the score was close. Girardi misused Logan, having the LOOGY face as almost as many righties (74) as lefties (85). That’s been a consistent blind spot in Joe’s managing since he took over the helm.

Overall, this cast of relievers was excellent. The Yankees’ pen finished 4th in MLB in WPA (6.14), and was behind only the Rangers among AL teams.

The outlook for next year is far more cloudy. Obviously, the biggest change is the loss of the GOAT. The Yankees are extremely fortunate to have one of the game’s very best relievers to replace Mariano as closer. Robertson has proven that he is an elite pitcher, and the numbers indicate that he won’t melt just because he will appear in a game 20 minutes later.

After that, things are largely unsettled. Look for Shawn Kelley to take over for DRob as the 8th inning guy. His poor ERA hid a ridiculous K/9 rate of 12 and a 3.24 xFIP. He’s a good pitcher, but his flyball tendencies mean a step down in in 8th inning reliability.

Matt Thornton takes over Logan’s spot as LOOGY. Thornton was a dominant pitcher for several years, but his strikeout rate has been falling precipitously and bottomed out last year. He was still very effective against lefty batters, so it will be important for Girardi not to get caught up in Thorton’s reputation as a lefty who can pitch to both LHBs and RHBs. He’s not that guy anymore.

A bullpen is almost as much about management as it is about personnel. This is strong point for the Yankees, as Joe Girardi seems to understand leverage better than most managers. Of course, his decisions were easier than most managers since he had Rivera and Robertson. Yet, he’s consistently used his most skilled relievers in the highest leverage spots, and kept his least skilled guys away from the important moments. On the other hand, he is oddly blind to LOOGY strategy. He costs the Yankees games by over-relying on his lefties to get outs against RHBs. It’s an easy fix, but don’t expect him to change now.

This will be Girardi’s most challenging year in managing the pen. This is a very top heavy bullpen with a lot of young, unproven guys behind Robertson. The relief corps is sure to be worse than it was in 2013, and they will almost certainly need to add someone at some point in the season. Continuing with a team theme, if Robertson gets hurt it could get very ugly.