clock menu more-arrow no yes mobile

Filed under:

How Many Wins Does a Bullpen Add?

In 2009, the 8 playoff teams were Minnesota, New York, Los Angeles, and Boston for the American League and Philadelphia, St Louis, Los Angeles, and Colorado for the National League. In the end, the New York Yankees won their 27th championship thanks to strong starting pitching, timely hitting, and a lockdown bullpen. Of course, when Mariano Rivera is pitching, everyone and their cat basically stops paying attention. Last year, he was essentially was the last "closer" standing. When your best relief pitcher takes the loss in a playoff game, it becomes difficult to win. That's what happened to Nathan, Papelbon, Fuentes, Lidge, Street, Broxton, and Franklin. Whether or not the losses would've changed the postseason is up for debate, but when there's no tomorrow in October, every win is critical.

We've seen how much locking down the end of the game has affected the Rays over the last few years. In 2007, the then Devil Rays fielded an underwhelming team, but even with a bottom-feeding defense and some underwhelming players, they compiled 17.0 WAR for players and a 12.1 WAR for the pitchers; neither was last place in the league and should have prevented a worst record overall in baseball for the fourth time.  While they did have a weak team in many areas, none was as glaring as the bullpen. FanGraphs has the bullpen valued at -23.8 runs over 497.0 IP. That's more than 20 runs worse than the next lowest ranked bullpen of Pittsburgh (-2.2 in 514.2 IP). That'll lead to a lot of losses, and also helps explain the charge to an improbable pennant in 2008 and why there were no playoffs for the Rays in 2009 despite another strong roster.

WAR (Wins Above Replacement) is one of the best measures of player value, but it just never seems to translate correctly to relief pitchers. Granted, you have players only in games for 60 innings a year and the fluky nature of a bullpen makes for low replacement value. And in a vacuum, all the innings are generally the same. After all, the game can be "saved" in any inning, not just the ninth.

Rather than use WAR to try and place a value (in wins) on how much the different bullpens have added to the Rays, let's look at something a bit different: WPA (Win Probability Added).

WPA -WPA +WPA pLI WPA/pLI Clutch WAR(pen)
2007-Dohmann 1.07 -1.81 2.88 1.11 0.74 0.22 0.1
2007-Reyes 1 -6.97 7.97 1.66 0.8 -0.2 0
2007-Camp -1.37 -4.66 3.3 1.18 -0.78 -0.39 -0.2
2007-Stokes -2.08 -6.26 4.18 0.93 -0.76 -1.47 -0.6
2008-Percival 0.42 -5.78 6.2 1.79 0.26 -0.03 -0.7
2008-Wheeler 1.03 -7.51 8.54 1.72 0.74 -0.14 0.2
2008-Howell 3.33 -5.33 8.65 1.26 1.49 1.16 1.5
2008-Balfour 2.69 -3.12 5.81 1.26 1.79 0.34 2.2
2009-Wheeler 1.36 -3.67 5.03 1.15 0.7 0.48 0.2
2009-Nelson -0.34 -3.11 0.277 1 0.05 -0.39 -0.4
2009-Balfour 0.44 -5.59 6.03 1.11 0.98 -0.58 0.9
2009-Howell 0.5 -8.17 8.66 1.82 0.74 -0.46 1.1
2010-Wheeler 0.17 -2.95 3.13 1.17 0.05 0.09 0.4
2010-Balfour 1.5 -2.66 4.16 1.26 1.02 0.17 1.2
2010-Benoit 1.8 -2.15 3.95 1.14 2.01 -43 1.4
2010-Soriano 4.2 -3.84 8.04 1.84 1.76 0.52 1.4

Okay, first off, those are some really depressing numbers put up by the 2007 bullpen.  Dohmann might have had the best season and that's not really a big deal.  Stokes and Camp are the two listed because they were probably the two to have failed the most, but it could've been anyone.

2008.  That Magical Year.  That is some nice pitching by the Rays bullpen...well, except Troy Percival.  There were noticeable jumps by Wheeler, Howell, and Balfour, contributing around 4 WAR and 7+ WPA to help lead the team to the World Series.

2009 was a season of expectation. Lots of times, the outcomes failed to live up to it. Despite a winning season at 84-78, the bullpen proved to have been a liability at -0.19 WPA (see below). But that should not come as a surprise. While the bullpen suffered from a postseason hangover during the beginning of the season, by the beginning of May it was operating as expected and helped the Rays stay in the pennant race until a fateful night in Seattle, which seemed to the end the hope in the Rays' season. Especially after an 11-game losing streak in which almost every night was a palooza of disaster.

Of course, the statistic that's the most pleasant to see for this season is Rafael Soriano's numbers. He's been money all season and has been worth the investment made for him. After enduring years when Troy Percival was expected to finish games off, it's nice to not have to restock on blood pressure medication after each save. Soriano picked up save number 38 on Monday in dominating fashion and there are very few others picking up saves that I'd rather have right now.  Balfour seems to have returned to the 2008 version which has been a boost along with Benoit picking up where the injured-JP was in 2008.  Wheeler's not contributed as much to the bullpen as he did in 2008, but his situations have changed, and we're thankful for that.

And now for the season totals:

WPA -WPA +WPA pLI WPA/pLI Clutch WAR(pen) WAR(no pen)
2007 Devil Rays -7.83 -43.15 35.31 0.95 -4.65 -3.63 -2.4 31.5
2008 Rays 9.3 -33.72 43.02 1.14 4.65 3.49 3.5 45.6
2009 Rays -0.19 -34.95 34.76 1.03 2.43 -2.62 2.1 47.3
2010 Rays(Current) 8.21 -21.04 29.25 1.09 4.11 3.43 3.9 33.4

Those are some huge swings. But it just re-enforces the fact that the it's almost impossible to have a steady relief corps even with similar pitchers. The team made huge strides in 2008 by solidifying the bullpen and having a better defense. When your bullpen generates a swing of -7.83 WPA to 9.3 WPA (+17.13 change) that'll win you a lot more games. Since the Rays had a 31-win increase that season, we can theoretically translate that 17.13 WPA swing into a 17.13 win increase, using a conversion of 1.0 WPA to 1.0 win.

This season from 2009 to 2010 is looking at a potential 15 win increase, going from 84 to 99-100. Extrapolating the Rays' WPA and the WAR (assuming they increase linearly) will give a WPA of 10.56 and a WAR of 43. While a decrease of 4 WAR and a 10.5 Win increase from the 2009 numbers doesn't fit well with the win total of 84 to give the 15 win increase, rather more like a 7 win increase for a total of about 92 wins.  But using the conversion, the 2009 season should've had a decrease of about 8 wins for a total of 89 rather than 84.  Those 5 wins that aren't there fit the model better to give an expected 2010 win total of 97 about.

This is an experimental correlation that I hope proves accurate in the future.  Right now only a Small Sampling to go on. If this works, I want the copywrite and patent on the statistic.