Spending $50 Million on Two Very Different Pitchers

Early in the off-season, Ricky Nolasco signed with the Minnesota Twins for $49 million over four years. Over the weekend, Matt Garza signed with the Milwaukee Brewers for $50 million over four years. While these contracts are nearly identical, the two pitchers could hardly be more different.

Over the last three seasons, Nolasco has averaged 199 innings per year, while Garza’s averaged just 152 innings per year. Nolasco has been reliably durable, avoiding the disabled list entirely for each of the last three seasons, while Garza has had three separate stints on the DL since the start of the 2011 season. Nolasco’s strongest selling point is his health track record; health is Garza’s weak point.

Nolasco has his own weak points, however. He’s consistently underperformed his FIP for nearly his entire career, as his 108 ERA-/92 FIP- is the largest spread of any active starting pitcher in baseball at the moment. While Nolasco’s BB/K/HR rates are all solid, he has a long history of giving up hits on balls in play and failing to strand runners, so his run prevention has never matched up with the estimators. Garza, on the other hand, has a below average BABIP for his career, and has a slightly positive ERA/FIP differential, though that hasn’t held true over the last three years.

On a per-innings basis, Garza is pretty clearly the better pitcher, but Nolasco has traditionally given his teams more innings. The choice between the two could be generally described as the age-old trade-off between quality and quantity. And the market has apparently determined that the difference in durability exactly offsets Garza’s difference in per-innings performance, as they signed the same basic contract in the same off-season with neither having free agent compensation attached.

That brings up the obvious question: is the market right?

Let’s start with what we know about predicting pitcher injuries. Jeff Zimmermann has done a lot of work on disabled list forecasts, and last year, the model he created to forecast DL stints did astonishingly well. If you look at the data he’s produced for 2014 DL predictions, you’ll see that, on average, starting pitchers are forecast to have a 38% chance of landing on the disabled list, with the spread ranging from roughly 30% — guys like Madison Bumgarner, Clayton Kershaw, and Mike Leake — up to around 50% — Tim Hudson, Brandon McCarthy, and yes, Matt Garza — at the higher end of things. Bartolo Colon stands out as something of an outlier, getting a 64% chance of landing on the DL, but the range is mostly 30%-50%, as 113 of the 127 pitchers fall within that span.

Zimmermann’s model definitely labels Garza as high-risk and Nolasco as low-risk, as Garza is forecast for a 51% chance of landing on the DL at some point in 2014, while Nolasco is just at 34%. This is a big gap, certainly, but what does it translate into in terms of expected number of starts going forward? Well, pulling Zimmermann’s data for disabled list stints by pitchers in 2013, I found that the average DL stint for a hurler last year was 69 days, though that is significantly inflated by pitchers who missed the entire season after recovering from surgery. The median is probably a better number to use here, since most pitchers who land on the DL don’t end up having season-ending surgery, so we’ll use the 51 day median instead of the 69 day average.

Taking Zimmermann’s DL forecast percentages and applying them to the 51 day median length of a DL stint for a pitcher, we’d find that his model would suggest Garza may be in line for something like 26 days on the DL next year. That’s basically a whole month of the season, and would cost Garza roughly five starts, plus some diminished performance on either side of the DL stay, so maybe the real cost is a loss of six or seven starts. That’s 20% of the season, and with that kind of risk, it’s easy to see why teams weren’t exactly bidding through the roof to land Garza.

However, Nolasco is still a pitcher, and even previously durable pitchers get injured too, so the difference isn’t five extra starts for Nolasco. The model projected him for a 34% chance of landing on the DL, after all, so applying the median DL stint of 51 days, it would suggest 17 DL days for Nolasco in 2014. Instead of a 26 day difference, we’re actually looking at a nine day difference in DL forecasts, or essentially two starts of the season.

Two extra starts. That’s roughly 12 extra innings. Even if you used the average DL stint instead of the median (to account for the higher chance of Garza actually needing surgery and missing the whole year), you’re still only looking at a 12 day DL difference, which is still only 2-3 extra starts over the course of the year. We’re talking about a forecast health difference of less than 20 innings, based on Zimmermann’s model at least.

If we’re really only looking at a 20 inning difference between a very high risk and a very low risk starting pitcher, then it seems that perhaps the price discount that injury prone starting pitchers are taking might perhaps be too high. After all, over 160 innings at something close to their career norms, Garza should be expected to allow roughly 70 runs, while Nolasco would be expected to give up something closer to 85 runs in 180 innings. Getting 20 extra innings and giving up 15 additional runs in the process isn’t so beneficial to winning games; any generic Triple-A arm can make up 20 innings while giving up fewer than 15 runs in the process.

Of course, these numbers are all estimates, and teams have far better access to medical data than we do. It is probably telling that a bunch of teams who have had Matt Garza aren’t interested in having him back, and they’re the ones who know the most about the red flags that doctors have raised. However, projecting future health is still mostly guesswork, and it seems like it’s possible that teams are putting too much of an emphasis on formerly durable pitchers and not enough on the quality of the pitcher when he is on the mound.