Mediating projections, Part II: CC Sabathia

Welcome back to Mediating Projections, wherein I look at interesting conflicts between forecasting systems and attempt to split the difference. Read Part I for more background (or the original installment for a full explanation of my reasoning). Today, we'll dive into CC Sabathia's projection. After another rough year in 2015, he has an uncertain future — Steamer and ZiPS don't know where he'll go from here:

System	IP	K%	BB%	HR%	BABIP	LOB%*	ERA	FIP	fWAR	fWAR/200
Steamer	141.0	18.8%	6.8%	3.0%	.291	71.5%	4.13	4.19	1.7	2.4
ZiPS	124.2	19.0%	6.8%	3.7%	.304	69.1%	4.98	4.46	0.2	0.3

*ZiPS doesn't project strand rate directly; I created this from the numbers it does project, using Sabathia's career hit-by-pitch rate.

The former thinks he could make a quasi-comeback, whereas the latter just sees a further plunge. In which system should we trust? Let's find out.

Home Runs

Even in his decline years, Sabathia's always fanned a fair amount of batters while keeping them from stealing free passes, so ZiPS and Steamer accede in those regards. They diverge, however, when it comes to long balls, the area where he's most regressed. Since 2013 — the year that his waning velocity really began to take effect — he's allowed 3.6 percent of opponents to smack a four-bagger, the main reason why he's posted an ERA- of 119 and FIP- of 107 in that span. Will Sabathia bounce back from that gopheritis, or will it continue to plague him?

Well, it all comes back to that velocity. Simply saying it's fallen really doesn't do it justice:

This is an incredibly massive dropoff. From 2007 to 2012, the six years in which Sabathia truly dominated, his four-seam fastballs went an average of 94.6 MPH. That plummeted to 91.8 MPH from 2013 to 2015, a two-and-a-half-MPH difference. Similar disparities arose on his sinker (93.8 to 90.6 MPH), changeup (87.2 to 84.9 MPH), and slider (81.6 to 80.6 MPH). Given more time to react to his offerings, hitters have smacked them out of the ballpark with gusto: His average fly ball distance has shot up from 276.8 feet in the former period to 286.1 feet in the latter, causing his home run per fly ball rates to swell from 8.7 percent to 15.5 percent.

Of course, not all pitchers need velocity to suppress fly ball impact. For the first several years of their careers, Matt Cain and Jered Weaver limited round-trippers despite tossing relatively softly. Perhaps, as age forces Sabathia to evolve from a thrower into a pitcher, he will follow that same path.

Let's put aside the fact that Cain and Weaver played half their games in voluminous stadiums, while Sabathia occupies the American League's most homer-friendly location. Last year, Jeff Zimmerman analyzed the consequences of pitchers losing velocity, discovering that their xFIPs wouldn't decline to the extent that their ERAs and FIPs would. This means that they'd consistently have a higher ERA and FIP than xFIP; that means, well, a lot of home runs.

To find another example of this, we need look no further than Cain and Weaver. They've seen their clout evaporate (albeit to differing degrees), and with it has gone their home run avoidance. Remember, too, that they've had their friendly fences to aid them. Sabathia, on the other hand, has had no such advantage to cushion the velocity-less nosedive, nor will he receive anything in the coming season. When it comes to his home run allowance, ZiPS has the correct mindset.

Batting average on balls in play

As with home runs, Sabathia has fared much worse recently when it comes to balls in play. His BABIP has jumped from .296 between 2007 and 2012 to .316 between 2013 and 2015. And as with home runs, ZiPS and Steamer don't know if it will continue or halt. So what are we to believe?

Sabathia's batted-ball profile reveals an interesting nugget of information. Across the second span, opposing hitters have made hard contact 31.0 percent of the time. (Because the accurate version of this metric dates back only to 2010, we'll focus on these years.) That mark is about five percent worse than the 29.4 percent MLB average — a significant difference, surely, but not one befitting such an inflated BABIP.

Let's take this a step further. Hitters have cracked 66 home runs off Sabathia from 2013 until now. Of those, we'll assume 50 — roughly three out of every four — came on a hard-hit ball. That would lower his hard-hit rate on balls in play to 28.2 percent, in line with the MLB mark of 28.1 percent. This would make him an average contact pitcher, whose BABIP we should thus expect to regress.

But can a home run-prone pitcher really limit solid hits elsewhere? Look at one of my favorite hurlers, Miguel Gonzalez. He's earned a sub-average BABIP in each of his four seasons at the show, adding up to a .272 figure over 580.1 innings. Simultaneously, he's sacrificed a ton of long balls — doing so 3.5 percent of the time, well above the major-league standard. Gonzalez's 29.2 percent overall hard-hit rate might seem high for someone who prevents hits as often as he does; with the balls that leave the yard taken into account, it makes more sense.

Or, as a BtBS reader, you may prefer something a little more empirical. I'd thus direct you to Steve Staude's pitching metric correlation tool, which shows no relationship whatsoever between a pitcher's BABIP and home run rate. In other words, there's a precedent for the model in which Sabathia could find himself — even when taters rain down, they don't necessarily make it pour.

If Sabathia hasn't conceded hard contact on balls in play, why has his BABIP suffered so much? The mediocre Bronx defense hasn't lent him a hand recently, placing 15th in the majors in UZR (and 21st in DRS.) That could change next year, though. FanGraphs' Depth Charts assert that the Yankees should have positive or neutral gloves at every position save right field, where the husk of Carlos Beltran roams. Overall, the site thinks they'll save 9.2 runs above average, some of which should trickle down to Sabathia.

Now, Sabathia has had some issues with line drives recently. His clip of those has gone from 19.8 percent in the good times to 22.0 percent in the bad ones, which never bodes well for anyone's BABIP. With that said, liners fluctuate a great deal year-to-year, and they are a pretty imperfect statistic in general. (Plus, Gonzalez has given up a lot of liners too, and it's never bugged him.) The fact that hitters have knocked line drives off a pitcher in the past doesn't mean they'll continue to do so in the future.

The disappearance of Sabathia's velocity automatically puts him at greater risk of ineffectiveness, so optimism here could backfire. Moreover, BABIP itself has a lot to do with luck — whether Sabathia succeeds or fails may not be up to him. Nevertheless, the batted-ball characteristics Sabathia has shown of late, combined with his newly competent defense and surprisingly beneficial home ballpark, make him a solid candidate to lower his BABIP. While I wouldn't expect Sabathia to completely quell hits, overall Steamer strikes me as the most realistic projection here.

Strand rate

Stop me if you've heard this one before: Sabathia used to leave runners on base (74.7 percent strand rate from 2007 to 2012), but he's struggled in that respect as of late (70.5 percent strand rate from 2013 to 2015). ZiPS predicts an even greater decline from his recent numbers, and Steamer foresees him recovering somewhat. Can Sabathia improve himself in this regard, or will he invariably falter in the clutch?

It's important to grasp the extent of the disagreement here. Across the past three years, a league-average pitcher has stopped 73.1 percent of his batters from scoring. Neither Steamer nor ZiPS anticipate that level of production from Sabathia, which makes sense — subpar hurlers will usually allow more runs as a whole (obviously), which means they'll leave fewer men on base. We just want to know whether Sabathia will trend up or down from the 27th-worst strand rate in the majors.

Sabathia possesses one key advantage that should continue to aid him here. Very few runners have sought to move up against him, especially as of late: Since 2013, he's permitted 16 stolen-base tries in 646 opportunities, giving him an attempt rate of 2.5 percent — less than half the 5.6 percent major-league average during that span. Part of this stems from his left-handed throwing ability, which grants him the power to stare down the runner at first base. Brian McCann, his primary catcher, has helped too, catching 36.5 percent of attempts since coming to New York. Because Sabathia presumably won't switch hands in 2016, and because McCann and his sizable contract won't go anywhere, this element of his game will probably persist.

Most of a pitcher's stranding ability, though, comes from his splits. How much worse does he perform when the opponents have reached? In Sabathia's case, much worse: Over the previous three years, his .331 bases-empty wOBA has given way to a .361 mark with runners on. This has negated any gains he'd made by holding runners in place, so this should warrant the majority of our focus. Diving into his situational splits, we can discern a clear trend:

Situation	K%	K%+	uBB%	uBB%+	HR%	HR%+	BABIP	BABIP+
BE	20.6%	99	6.3%	89	3.7%	137	.304	103
MoB	18.0%	94	6.5%	85	3.5%	144	.332	112

Sabathia has lowered his strikeout and walk rates by a bit, suggesting he pounds the zone more often with men on base. That could theoretically lead to hitters better squaring the ball up in those scenarios, which would explain the rises in long balls and hits. But Sabathia's actually done a better job of avoiding good contact — his 30.4 percent hard-hit rate with runners on trails his 31.3 percent mark with the bases unoccupied. If the home run and BABIP differences regress away, he could leave on a few more runners.

Strand rate is even flukier than BABIP, with an incredibly weak year-to-year correlation. Sabathia could allow everyone to score next season; he could allow no one to score next season. Based on the evidence here, however — Sabathia's (and McCann's) ability to kill steals, and Sabathia's probable improvement with runners on — I'd take the over. Keep in mind that all of these numbers came with a 70.5 percent strand rate, and even a marginal upgrade from that would put him in Steamer territory.

Conclusion

Of the three conflicts, one goes to ZiPS and two go to Steamer. By this shoddy logic, Sabathia has a decent shot at some level of improvement. With his contract possibly expiring after the 2016 season, his play over the next several months could make a big difference for his future job security — not to mention bolster his case for the Hall of Fame. Although the past three years haven't gone well for Sabathia, perhaps the fourth one will be the charm.

. . .

Ryan Romano is an editor for Beyond the Box Score. He also writes about the Orioles on Camden Depot (and on Camden Chat that one time) and about the Brewers on BP Milwaukee.