The Arizona bullpen allowed 271 runs and finished the season with a 4.61 ERA, both numbers which put them ahead only of Washington among National League relief corps. We've previously discussed the waking nightmare which was the eighth inning, but it's also worth taking a look at the numbers posted by the bullpen in general. How do they stack up in total? What were the issues? How might things have gone, and where do we go from here?
FIP vs. ERA
On straight ERA, as noted above, our relievers were pretty bad, even when you adjust for park factor. However, the underlying performance numbers weren't quite as horrendous. Arizona relievers had a combined line of .264/.338/.395, which is an sOPS+ of only 101, not too much worse than average. Fangraphs.com's FIP stat goes even further: the D-backs FIP was 4.02 - sixth-best in the league - with the biggest gap between that and 'actual' ERA of any team in the National League. This explains why they rated our pen at 36.1 runs above replacement, not just above average, but very close to the number (39.4) posted by the "lights out" 2007 bullpen.
The only two relievers with 10+ innings, to have a FIP higher than their ERA were Clay Zavada (FIP 3.92) and Blaine Boyer (3.34), both of which make sense. Zavada certainly benefited from the lucky scoreless streak which opened his career - his ERA the rest of the way was 5.34. And Boyer is obviously not the 2.68 ERA pitcher he was with Arizona: if he was anything like that, the Cardinals wouldn't have dumped him. Still, even if they deliver their FIP numbers in 2010, I don't think we'll mind too much. The other eight relievers all had FIP better than their ERA, the amount ranging from 0.07 (yeah, Scott Schoeneweis was as bad as he looked) to 1.06 for Juan Gutierrez.
There are various reasons why FIP can differ. It's based purely on home-runs, strikeouts and walks: you might think, "Hang on - the number of hits that stay in the park is irrelevant?" But the problem there, is that those do depend on the defense, so are not a good measure of pitching ability. I'll go into more detail about our glovework in the final part of this series, but Arizona's defensive efficiency - the percentage of balls turned into outs - was below average. If it seems counter-intuitive that you put H, K and HR into a mathematical blender and get a number that means anything, studies have shown that FIP is a better indicator of future performance than ERA.
Win Probability
However, the Hardball Times Win Probability Added is less convinced, saying our relievers cost Arizona 2.81 wins, this year ahead only of Pittsburgh and Washington. This is the same Win Probability we graph after every game; the winning team has 0.5 of a win to divide up, and the losers -0.5 blame. Over the entire season, our hitters "cost" Arizona 6.53 wins, and our starting pitching 1.67. [The total, not by coincidence, is 11, the number we were below .500] However, you should remember the value of a performance is radically dependent on when it happens. As Dave Studeman points out, "If a player hits a home run in the ninth inning of a 1-0 game, he is credited with more WPA points than if he hits a home run in the first inning of a 1-0 game."
The late innings will tend to have more leverage as a result, though given the bullpen threw about half the innings of the starters, to be "responsible" for more losses than the rotation isn't good. Still, the implication is that the offense was a bigger problem for the team in 2009 than the pitching in total. For context, it 2008, the bullpen WPA was -1.16 wins and in 2007, it was a stunning +7.48 victories. This is likely a reflection of the huge number of close contests in which that team was involved. Almost one-third (52) were decided by one-run, and when you get into the late innings of such games, every out has a very significant impact on Win Probability.
Was the bullpen overtaxed?
Not apparently. The average team saw 2,184 PAs dealt with by their relievers last season; Arizona saw fractionally less than that, at 2,122 - it clearly helped to have one of the most durable starters in the league, in Dan Haren, whose IP/start was 6.95, fourth-best in the league. Jon Garland (6.18, 24th) and Doug Davis (5.98, 34th) were also above the NL-average of 5.82. The Diamondbacks were also the least-likely team to use a reliever on back-to-back days: this happened only 80 times, the most common subjects being Gutierrez and Rauch, each eleven times. However, the results of such outings were good, with a 3.67 ERA.
The only reliever who worked more than 55 innings for the Diamondbacks was Gutierrez, who threw 71 innings [Rauch and Tony Peña were both around the same for the year, combining the totals for their two teams], which was not enough to put him among the top thirty 'pure' relievers in the majors. By appearances, his 65 games was only ranked #70, though Rauch ended up =13th by that metric. Arizona were right in the middle of the pack with regard to asking our relievers to get more than three outs too, so overall, workload doesn't appear to have been much of an issue for the bullpen as a whole.
Inherited issues
What did seem to be a problem was the bullpen's problem with inherited runners. Across the league, 30% of inherited runners scored, but for the Diamondbacks' relievers, that number was 38%, the worst figure in the NL. Only two relievers were below league average, and that fractionally: Scott Schoeneweis and Leo Rosales both came in at 29%, and the brevity of the former's appearances were likely a factor. From there, things only get worse, all the way up to Blaine Boyer (8 of 17 inherited runners scored, 47%) and Daniel Schlereth (7 of 12, 58%). Obviously, while we are talking small sample size here, and not all inherited runners are created equal, the overall numbers are significant enough to indicate a problem.
The Arizona bullpen: anti-clutch in action
A major driving factor in a pitcher's ERA will be how they perform with runners in scoring position - as we saw with Doug Davis this year, you can get away with allowing a lot of base-runners, as long as you don't let them get past third-base. This was something our bullpen didn't do last season. The Diamondbacks' OPS for their relievers across all situations in 2009 was .734: they can be divided into three groups by their performance with runners in scoring position (min. 30 such PAs):
Significantly better: Leo Rosales (.653), Tony Peña (.655), Esmerling Vasquez (.686)
There or thereabouts: Chad Qualls (.728), Juan Gutierrez (.730),
Significantly worse: Blaine Boyer (.779), Jon Rauch (.853), Clay Zavada (.959), Scott Schoeneweis (1.078), Daniel Schlereth (1.192)
This would explain the high ERA; not just the raw number of pitchers, but the disparity. We had four relievers whose OPS was 100 points or more worse than bullpen average, when they had runners in scoring position, compared to none who were that much better. Avoiding such a skewed result in 2010 will certainly help. Harder to say whether it's something that can be addressed, or is just a question of waiting for regression to the mean.
What might have been
The front-office got criticism for letting Brandon Lyon and Juan Cruz walk at the end of 2008, and it's fair to say that the arms they brought to Arizona instead, Schoeneweis and Tom Gordon, were dismal failures. Admittedly, the former's problema could in no way have been foreseen, and completely derailed the year. Before his wife's death, Scott had an ERA of 2.53; after, it exploded to 10.80. But in the final analysis, he is just another in a long line of disappointing LOOGY's brought in by the team: see also Randy Choate, Mike Myers, Eddie Oropesa, etc. Since Myers in 2002, no free-agent southpaw reliever has had an ERA+ above 100 for Arizona.
Gordon was a low-risk roll of the die, but even at that, it must have been disappointing to get no more than 1.2 innings and a 21.60 ERA from the veteran. He only cost $600K, but one assumes the budget included the maximum Gordon may have earned, $3m. Between the price of him and Schoeneweis ($2m), the team could certainly have done better. Looing at the players we let go, Juan Cruz signed a two-year, $6m contract with the Royals, but had an ERA of 5.72, with a strikeout rate almost cut in half (from 12.4 K/9 to 6.4). Allowing Cruz to walk now seems wise, and Kansas City will be hoping for much better in 2010.
At first glance, Brandon Lyon was excellent with the Tigers, holding opponents to a .205 average and posting a 2.86 ERA, that seemed to justify fully his $4m deal, and likely sets him up for a big pay-day this winter. However, Lyon did have a freakishly-low BABIP of .229, leading to a FIP above four - the difference put him in the top 20 among major-league relievers. Still, signing him for $4m and getting anything like his actual performance would have been a much better deal for Arizona than Gordon and Schoeneweis.
2010?
From statements so far, it does appear that one or more arms will be added to the bullpen in the off-season. We'll be discussing this in more detail in due course, but I imagine the core is alreay present, in Qualls, Gutierrez, Vasquez and Zavada. Boyer and Schlereth would also seem to have a potential role, though use of the latter in high-leverage situations still concerns me. I still wouldn't mind us spending $5 million or so of the estimated $28m available on a couple of decent arms: I think "reliable" is the key-word there. Rather than us taking a risk on anyone like Gordon, we want someone who we know will contribute.
The last installment of this series will follow, probably next Monday. There, we'll be looking at defense and base-running, and their impact on the team's performance last season.