2008 Review, Part 4: Bullpen

by Jim McLennan on Oct 8, 2008 11:54 PM EDT

Much though it may seem difficult to believe, by a lot of measures, the Arizona bullpen was not as horrifically ineffective as it seemed. Their ERA of 4.09 was equal sixth-best in the league, without any park adjustment, and batters' OPS against them was also better than the National League average. They weren't particularly "unlucky" either, with a BABIP exactly on the mean at .298. Yet, somehow, they posted a record of 17-28 over the season: As a comparison, Houston's bullpen had exactly the same ERA as us this year, but they went 26-23 - if the Diamondbacks had done that, we'd have won the league.

Even breaking it down to individual innings doesn't shed any particular light on things. Our worst inning, by quite some distance, was the first, where the opposition scored 92 times. Next was the third, at 86 - neither of these can be blame on the bullpen. The last three innings were down at 79, 77 and 74 runs respectively, and each of them was also almost spot-on the average OPS for the team (.716), with figures of .716, .715 and.718. Yet the team lost fifteen games they were winning after six innings - almost twice as many as the eight dropped last season, when the bullpen ERA was only fractionally better, at 3.95.

Again, that slump in results would be the difference between first and second, right there. It begins to look as if the bullpen was simply remarkably unclutch. They didn't allow a particularly large number of runs, but the ones they did had a nasty habit of proving crucial in deciding the outcome. The 'Late and Close' stat - 7th inning on with the batting team tied, ahead by one, or the tying run at least on deck - does give some hint of that. The team figure for OPS was .729, and it'd be higher if you remove our starters in that situation: Haren (.710), Davis (.675), Johnson (.650) and Webb (a stunning .538), who combined for about 20% of those at-bats. Some of our bullpen did thrive there, led by Qualls (.641) and Cruz (.655). Others didn't.

Here's are the stats for our seven top relievers, by innings pitched, and some comments on each.

         G  W-L  S   IP  ERA  H  R ER HR BB SO
Qualls  77  4-8  9 73.2 2.81 61 29 23  4 18 71
Pena    72  3-2  3 72.2 4.33 80 38 35  5 17 52
Lyon    61  3-5 26 59.1 4.70 75 34 31  7 13 44
Cruz    57  4-0  0 51.2 2.61 34 17 15  5 31 71
Slaten  45  0-3  0 32.1 4.73 33 20 17  4 14 20
Rosales 27  1-1  0 30   4.20 32 15 14  2 15 18
Rauch   26  0-6  1 23.1 6.56 27 18 17  6  9 22

Chad Qualls. Started the season very well, with a scoreless streak of 16.2 innings, and finished it almost as strong, with another one of 14.2 frames. In the middle, however, he had his struggles, most notable a spell where he suddenly couldn't stop inherited runners from scoring. From June 2-July 4, his own ERA was only 4.38, but he allowed eight of ten base-runners who were there when he arrived, to cross home-plate. He also was the victim of the defense: on May 7, he had an 0-3 record, despite an ERA under one, because six of the eight runs he'd allowed to that point were unearned. Probably was our best reliever, keeping opponents to a .601 OPS. Shame it took Melvin until mid-September to make him the closer.

Tony Pena. He was the Mark Reynolds of the bullpen; he would have a streak where he was incredible, then suddenly, would struggle in a way that made you wonder if he'd ever be good again. 40% of those earned runs came in just four outings. He was particularly good with two outs and runners in scoring position, holding opponents to a .184 average, and generally pitched much in close games. When tied or in one-run games, batters hit .203 in 153 at-bats; if the margin was two or more, the figure was .274, and when four or more, he seemed to give up entirely, with the OBA was all the way up to .333.

Brandon Lyon. Poor Brandon. Partly the victim of a .342 BABIP. Was it over-use? Pitching after working the night before, his ERA was 6.92, compared to 4.08 after a rest. We'll never know for sure, but his season imploded on July 19. ERA before that date: 2.37. From that date on: 8.86, and certainly, asking a reliever with known arm issues to throw 27and 28 pitches on consecutive days, seems a questionable decision by Melvin. A small nugget of hope: after being removed from the closer's role, he finished the season with six scoreless innings. I'm just glad it is not my decision whether to offer this free-agent arbitration or not this year, because I have no idea whether he would be worth the risk.

Juan Cruz. We got pretty much what you would expect from Cruz: lots of walks and a phenomenal strikeout rate, 12.37 per nine IP, trailing only Grant Balfour (12.65) among major-league pitchers with 25+ innings. He missed three weeks in July and, despite the best ERA of any of our relievers, seemed to fall out of favor with Melvin - there was a two-week spell in August where he threw just a single inning. That came immediately after the game on the 9th, in which his throw to second led to Hudson's broken wrist. Cruz was particularly tough on left-handed batters, who hit a mere .159, with 35 strikeouts in only 82 at-bats. Like Lyon, he's now a free-agent: estimates suggest, unlike Lyon, he'll be a Type A, so would bring a very nice haul if he signed elsewhere.

Doug Slaten. Did a credible job at what he was supposed to do, retire left-handed batters, keeping them to a .232/.317/.375 line - though as noted above, Juan Cruz was even more effective. Leaving Slaten in there to face any right-handed batters was highly-questionable, as their OPS against him was a healthy .866. He had an odd home/road split: while the OPS was very similar (.801/.778), his ERA was far better away from Chase (6.48/2.87). Slaten saw very little action after the All-Star break, partly due to injury, but was also sent down to Tucson in late August. That all combined to limit him to six innings of work in the second-half of the season.

Leo Rosales. Following his recovery from a broken hand caused by an argument with a wall, Rosales made his debut on June 15 and quietly had himself a decent little season. Though at age 27, he no longer really qualifies to be called a "prospect". Control was probably the main issue, allowing fifteen walks in only thirty innings, but over his first eighteen games and 21.1 innings, his ERA was a respectable 2.53. A couple of bad outing inflated his numbers in the last couple of months, but he is likely among the leading contenders for a full-time bullpen spot from Opening Day in 2009.

Jon Rauch. He picked up six losses for Arizona, in only 23.1 innings of work: as far as I can see, that's an all-time record [the nearest I could find were Jaime Navarro, who went 0-5 in 18.2 IP for the 2000 Brewers, and Ed Farmer, 0-6 in 26.2 IP for the 1983 Phillies]. The main reason is his ineffectiveness when the game was tied. Batters hit .320 (16-for-50), with an OPS of 1.027 - basically, with the scores level, Rauch made opposing hitters look like Manny Ramirez. That losses then ensued, is not really surprising. Small sample size? Probably. But I do not want to see him in any of those "close and late" situations, until he has proven the ability to retire batters when the game is not on the line.

Tweet Comment 9 comments | 0 recs |

Do you like this story?

Comments

Final NL Releiver Elias Projections are out

http://tigers-thoughts.blogspot.com/2008/10/final-elias-projections-nl-relievers.html

He predicts Lyon grabs the last B slot. Still has Cruz as the third best A free agent (second if you DQ Dempster). I guess there goes the theory that saves would secure Lyon an A.

Personally, I’d like to see both of them brought back. I don’t think we have anything better in the system and there is likely not a whole lot better floating in the FA market.

I think the Rauch situation bears more investigation. Simple glances at his BABIP and FIP numbers from years past suggest that he was benefiting from either a better ballpark or better defense before he came here. I don’t think it’s so much that he’s a bad pitcher, but just a bad fit for this team/ballpark. He is still a type A for now, the Diamondbacks may be able to salvage some value in a trade based on that alone.

Fire Bob Melvin

by nihil67 on Oct 9, 2008 10:57 AM EDT reply actions

There's probably some random variation,

or “unclutchness”, responsible for the reliever’s 17-28 record, but let’s not forget that the RPs (like the starters) were also hampered by crummy run support (not relative to each other, but relative to their respective NL peers). After park considerations, both starters and RPs recvd less support than all but three or four teams.

This is more or less what happens when you marry an average (or slightly better than average) bullpen beneath it’s station with a uniformly bad offense – and underscores just how good (and to some extent “clutch”, lucky) our rotation was, finishing 13 g over .500 themselves, despite matchmaker Schu’s ball and chain.

I dont disagree with you that there were some tough losses late, maybe even some random variation sprite hexing the pen. But for the most part, I think that random “bad luck” was counterbalanced by fairly “good luck” enjoyed by the aggregate rotation, and that we shoud expect fairly similar W/L results going forward unless the offense perks up.

by Diamondhacks on Oct 9, 2008 3:31 PM EDT reply actions

Yet, curiously...

Starters: 4th in ERA, 11th in run support, 65-52 record.
Relievers: 7th in ERA, 8th in run support: 17-28 record

So the relievers got better run support, compared to their peers, than the starters, yet had a record which was a great deal worse. Obviously, run support matters, but since the starters performed three notches better, and got three notches worse run support: I’d have said that should be more or less a wash.

Oh, and I found Jon Rauch’s record when appearing for Arizona in tied games. Hitters were 9-for-11 with four home-runs. That’ll get you losses in a hurry.

by Jim McLennan on Oct 9, 2008 5:12 PM EDT reply actions

Uh...

Yikes. That’s even worse than I would have imagined.

by Azreous on Oct 9, 2008 7:03 PM EDT up reply actions

I think we agree on bullpen "leverage"

So the relievers got better run support, compared to their peers, than the starters…

They really didnt.

NL Starter’s mean Run Support: 4.90
AZ Starter’s mean Run Support: 4.81

NL Reliever’s Mean Run Support: 3.97
AZ Reliever’s Mean Run Support: 3.89

If you’re looking for a “wash”, that’s where it is, imo. Run support.

Our starters’ record was a dozen wins better than the pen for two reasons. First, they pitched better (about 50 runs, or six wins, better, relative to respective peer groups). Second, their run distribution assuming equivalent prevention and support was more fortuitous, beneficial – (ie they won more close games & the pen lost more close games – even relative to respective peer groups) – accounting for the additional six win disparity.

The bullpen’s next to worst record in the NL (17-28) was a nexus of three factors:

Poor leverage (or distribution)
Weak run support
They pitched about average, a little above (ie run prevention)

The rotation’s 65-52 record was the nexus of three factors:

They pitched exceedingly well
Above average leverage
Weak run support

by Diamondhacks on Oct 9, 2008 10:54 PM EDT up reply actions

Aargh

I tried to insert the special n into Pena’s name when I was replying and I lost the entire post. Dammit. I can’t be bothered to retype the entire thing, but the main thrust was twofold.

1) Median is better than mean, because it isn’t skewed by freak outliers. That shows that the majority of relievers in the NL got less run support than the Diamondbacks. Though even if we take the mean, the difference of 0.08 fewer runs per nine innings, equates to a whopping four less runs of support over the entire season pitched by the bullpen. Seems barely significant to me.

2) Again, contrast Houston. They had the the same ERA as us, but much less run support – 3.52. So, if that had any effect, you’d expect them to lose more games. Instead, their bullpen were seven games better than Arizona.

I think it’s almost all down to the leverage. From Fangraphs, here are the ERAs and the Leverage Index (LI) for the seven relievers listed above. I’ll sort this list by LI:

Lyon: LI 1.65, ERA 4.70
Qualls: LI 1.43, ERA 2.81
Pena: LI 1.39, ERA 4.33
Rauch: LI 0.96, ERA 6.56
Cruz: LI 0.87, ERA 2.61
Rosales: 0.68, ERA 4.20
Slaten: LI 0.54, ERA 4.73

Basically, Melvin screwed the pooch by sticking with Lyon for about fifty games after it was obvious he sucked, and also continually rolling Rauch out there until he lost us six games in barely a month, while relegating Cruz to garbage innings. And when did Rauch start sucking? Just like Lyon, Immediately after Melvin rolled him out there three days in a row, from August 12-14. ERA before that: 3.02. ERA after that: 9.75. For any’s sake, I hope Max Scherzer is kept out of the bullpen, or else Melvin will likely blow up his arm next season.

by Jim McLennan on Oct 10, 2008 12:51 AM EDT up reply actions

I'm quite surprised

you’re pushing the run support argument so hard, Jim, when we havent even mentioned park considerations. In light of that, the mean-median thing isnt even worth quibbling over.

The fact is most NL relievers enjoyed measurably better run support than dback RPs, regardless of whether they ranked 8th in RS or not. Of the eight teams ranked below Arizona, six of em hopscotch over the dbacks when you adjust for park – including San Diego and LA. Houston RPs didnt receive much less run support than us. Chase runs ran 9-10% higher than Minute Maid and support was very comparable.

In 456 innings (approx 51 full game equivalents), the offense provided our relief pitchers with 197 runs(earned and unearned), while the RPs yielded 232 ( earned plus unearned). The expected W/L associated with 197 RS/ 232 RA is in the neighborhood of 23-28 or 22-29.

I dont want to put words in your mouth, but all through this thread you appear to be advancing the notion that the bullpen should’ve won – I dont know, you tell me – half their decisions? And the overwhelming reason they didnt was because Bob Melvin repeatedly leveraged the wrong guys? But the numbers dont bear that out. Poor leverage, distribution, whatever you want to call it, certainly played a part, when pythag says the pen s/b about 23-28 and they go 17-28 instead. But neither should one assume that a pen yielding considerably more runs than they were furnished with, should somehow magically win half their decisions, were it not for a conveniently scapegoated manager.

by Diamondhacks on Oct 11, 2008 3:46 AM EDT up reply actions

The shape of the season

Through April 30th, the bullpen had the following numbers:

90.2 IP, 2.58 ERA 62 hits, .193 BA against, 5 HR, 35 BB, 71 K’s 7.0 K/9

Stellar number up to that point to be sure. But there were some warning signs. They obviously were getting hit lucky, and that K/9 rate is not really all that great for a bullpen. But most alarming was the AMOUNT of innings the bullpen threw in April. 90 was too much and it was being concentrated to too few guys. The starters were being a little too protected early in the month, as they were not ready to go deep in games yet. I’ve always wondered why the D Backs starters seemed to be about an inning or two behind other teams starters in spring training.

Then towards the end of the month, through some bad luck, there was a spate of games where the bullpen had to pick up a lot of innings due to starters getting knocked out early. This carried over into early May as well.
In fact, right at a time at the end of April and early May when the starters should have been going deeper in games, they went through a 12 game stretch where they barely averaged 5 IP per start.

Well all know what happened to the bullpen next:

May: 4.48 ERA
Jun: 4.22 ERA
Jul: 5.03 ERA
Aug: 4.79 ERA
Sep: 3.70 ERA

From May 1 on: 4.43 ERA 365 IP 381 H, .269 BA , 129 BB, 309 K’s 44 HR

The April workload, where they pitched 1/5th of their seasons innings in 1/6 th of the games was a pretty big factor I believe.

Clearly a better offense, and one that scored more runs late in tie games would have given the bullpen a better won loss record. But the D Backs bullpen struggled quite a bit for 4 /12 months, and I believe that trend started with being overworked early in the season. Of course the team was too slow to remove Lyon and Rauch from high leverage roles when it was obvious to all they were going through more than just mid season slumps. Us message board pundits usually overreact too quickly and are not patient enough to let players work through their troubles, but in this case, the team was simply TOO patient.

by shoewizard on Oct 11, 2008 2:08 PM EDT reply actions

Good stuff

Assuming downward causality from the 90IP is interesting and plausible, but compromised, I think, by the strong likelihood that our post-April pen was due for major performance regression anyway – quite apart from workload.

I dont disagree that 90 IP is on the high side over 28 games, but we’re still talking about a pen that’s thrown fewer innings than any NL club in each of the past two seasons. So, at the end of the day, even after reading your well thought out chronology, I remain kinda predisposed against the notion they were “overworked”. The pen went 80 innings thru 28 games in 2007, for example, and was just as heavily worked as 08 over the first few weeks – and they turned out fine. How out of line is 90 innings over 28 games, really – across the league? And wouldn’t April present the best time to push reliever load, given that everyone should be relatively rested? Also, in April 08, we were winning games, requiring some sort of “A” pen to go out there and secure those wins.

The starters were being a little too protected early in the month…

I appreciate the word ‘little’, but who specifically?

Webb and Haren, arguably the team’s most irreplaceable assets, appeared to get out of the gate eating innings. Were we overprotecting our 44 year old, $26M fireballer smarting from an injury truncated season, by holding him out til April 15th? Maybe, but the fact a guy obsessed with wins didnt go 7IP until mid May suggests , to me, he was genuinely physically not ready. We werent overprotecting the guy making a start with thyroid cancer, were we? Were we babying Micah Owings, who went into the 7th in three of his first 4 starts before hitting a wall, from which he never seemed to recover? Please, help me out here :-)

I guess what I’m suggesting is that, given the pitching resources he had to work with at the time, BoMel’s April decisions on who should take the ball when seem constrained by legitimate, competing concerns and justifiable.

Late season pen leverage legitimately raised some eyebrows, and I myself gagged at more than one appearance of Big John Routs with the game (indeed, the season) on the line. But Melvin also inserted guys all year into hi lev spots that the peanut gallery had conveniently turned on, to methodically record outs and help win games. Qualls. Lyon. Petit. Pena. Not as successfully as 2007, obviously, but he had more limited choices, far more limited late in the season. Let’s face it – only Qualls, Lyon and Cruz were recording outs after Sept 1st. That makes it tough, when you have to win (ie protect) most of those games, and each guy can only throw 10 or 12 innings per month. Other than the curious leverage “mixup” between Cruz and Rauch (still not explained by Melvin to my satisfaction), I’m just not real clear on what better alternatives he had at his day to day disposal.

by Diamondhacks on Oct 11, 2008 7:18 PM EDT up reply actions