The dominant item in the news cycle is that Roger Clemens is
close to signing with the Astros for the remainder of the 2006 season. The
story has gone from “done deal” to something less than that in the last 18
hours, but the sheer force and number of denials are a sign that a conclusion
to this is imminent.
As I’ve said all along, there’s just no way of knowing what Clemens is going
to do, because the decision is an intensely personal one that isn’t as much
about baseball as it is about one man balancing roles as a father, husband,
competitor and teammate. Yesterday’s spate of rumors was the first solid
indication that he would return to the mound, and also the first time I’ve
found myself believing that any particular resolution was likely.
So maybe this happens, maybe it doesn’t, but let’s take a run at it to figure
what the Astros would be getting. This is not as easy as you might think,
because “Roger Clemens” is no stat-generating robot. This is a guy who was the
best pitcher in the National League in 2004, even better than than for four
months in 2005, and then considerably more ordinary down the stretch and into
the 2005 postseason. He’s 43 years old and fought nagging injuries late last
season. Under the best of circumstances, pitchers are unreliable creatures;
guessing what Clemens will bring to the table for four months requires a wide
tolerance for uncertainty.
Just as a start, let’s consider Clemens’ weighted-mean PECOTA projection.
Conveniently, PECOTA projected Clemens to make just 24 starts (and oddly,
seven relief appearances), so the innings difference isn’t great. If Clemens
were to sign this week, make two tune-up starts and join the rotation
the week of June 12, he’d likely get 20, maybe 21 starts. It’s arbitrary, but
let’s say that PECOTA has him throwing 14 relief innings (two per appearance)
and 147 as a starter, for an average of seven innings per start.
Based on his weighted-mean PECOTA projection, in 147 innings, Clemens would
allow 50 earned runs (PECOTA is silent on unearned runs).
Clemens would replace either Fernando Nieve or Taylor
Buchholz in the Astros’ rotation. Neither has pitched consistently;
both pitchers have had excellent outings and complete disasters. If you make
the same calculations for each as you do for Clemens, while weighting their
projection and their 2005 rate of earned runs allowed equally, you find that
Nieve would, if allowed to throw another 147 innings, allow 78 runs, while
Buchholz would allow 87. Now, both pitchers would likely pitch fewer than 147
innings in 21 starts, but because the gap would be filled by the back end of
the Astros’ bullpen, there’s little change in the conclusion.
Adding Clemens to replace Nieve or Buchholz would save the Astros between 25
and 40 runs, if Clemens pitched at the level of his PECOTA projection. That’s
a gain of something between two and four wins.
You can see the range of expectations by looking at the possible range of
Clemens’ performances. For example, if he pitched at his 2004 level, he would
allow 48 runs in 147 innings, not much different from his 2006 projection. If
he showed up as the pitcher who crushed the NL for four months in 2005, then it
looks even better. Clemens allowed just 23 runs in his first 142 innings last
season through the end of July. If he aped that performance, the Astros would
gain five to six wins by having him around. On the other hand, Clemens was a
markedly different pitcher after that, including the postseason. If he allows
earned runs at the rate he did last August, September and October–31 in 85 1/3
innings–the edge over the incumbents slips back into the two-to-three-win
range, and is probably even a little less due to the likelihood that he would
throw fewer innings at that level of effectiveness.
So there’s a fair amount of uncertainty here. Clemens is virtually certain to
be an upgrade over the Astros’ current starters, but that’s not telling anyone
anything. The upper bound of that improvement seems to be about six wins, and
that’s if crazy-low-BABIP Roger from the first four months of 2005 makes a
reappearance. More likely, the gain the Astros get will be four or five wins
over what they would have done without him.
That’s more than worth it, of course. The Astros are probably going to pay
Clemens between $10 and $12 million for his time, and could make that up just
in increased interest due to Clemens’ popularity and a stronger shot at the
postseason. Beyond that, the Astros are likely to be the kind of middling,
somewhat-over-.500 team for which a four-game improvement could be the
difference between being done with ten days to go and adding a lot of home
dates in October. With an aging roster and a weak farm system, they have a
time horizon measured in minutes, and have to leverage any chance they have to
win in the current season.
The real question is whether this is enough for the Astros. They certainly
look like they could use the help, as they’re 15th in the NL in runs allowed.
The rotation, however, has been fairly effective: 10th in the majors in
Support-Neutral Value, sixth in the NL. The bullpen is where they’ve lost
ground: 19th in WXRL, in no small part due to Brad Lidge‘s
struggles. Worse, they’re 13th in the league in EqA, tied for ninth in OBP and
a brutal 14th in SLG despite playing in park that boosts power. Adding Clemens
strengthens the team’s strength without addressing its biggest weaknesses.
That’s not a bad idea–runs are runs, wins are wins–but to make this whole plan
work the Astros are going to have to add a lot of runs. Their corner
outfielders, mostly Preston Wilson and Jason
Lane, have been awful, playing at below-replacement level. There are
organizational commitments to non-hitters Willy Taveras,
Adam Everett and Brad Ausmus, the last of
whom is actually the team’s fourth-best hitter so far.
Roger Clemens’ possible return to the Astros pushes them clearly above .500
and into the thick of the wild-card race. Whether that race turns out well for
them depends less on Clemens’ performance, though, and more on finding
productive major-league outfielders to solve the Astros’ real problem: scoring