PS zStats for Pitchers at the Midpoint

The emergence of Statcast (and similar types of tracking data) over the past decade has revolutionized many aspects of baseball analysis. The biggest category that was really missing was the concept of “expected” statistics. Until then, the numbers were all results statistics, and proto-expected metrics, like Bill James’ Component ERA, were taken from the old stats list. But tracking data has opened up new opportunities in this area, allowing us to look more closely at home runs and hits, and see the underlying processes and skills that produce those results. Although the past is always in the past, expected statistics are useful when talking about the future.

As someone who made the odd decision to work with baseball for part of his life, I am very interested in finding the best use of this type of information when predicting the future. Like Statcast ratings (prefixed by x, as in xBA, xSLG, etc.), ZiPS has its own version, cleverly using az instead. zStats has some correlation to xStats, but not perfect, as ZiPS uses things like spray data, pitching speed, and plate orientation metrics in its calculations.

It is important to remember that these are not predictions in themselves. ZiPS doesn’t just look at last year’s pitcher’s zSO and say, “Cool, brah, we’ll go with that.” But the data covers how events unfold and is more stable than the actual statistics of each player. That allows the model to shade the projection to one side or the other. Sometimes that matters a lot, like in homers allowed to pitchers. In neutral statistics, homers can be highly degraded, and pitchers’ home run ratings are more predictive of future homers than actual homers allowed. Also, the longer a player “underachieves” or “overachieves” in a certain statistic, the more likely ZiPS believes the actual performance is higher than expected. Call this the Isaac Paredes Rule, in honor of the player who always tormented zHR. In a way, we are revealing just how brutal the reversal will be.

More information about accuracy and design can be found here.

As a hitter, the best place to start is to check out some of last year’s high achievers and underachievers.

Of the top 20 hitters in zFIP, meaning the model had better FIP-based performance than one would expect from tracking data, 17 saw their FIPs drop in the second quarter. As a team, they had a true first-half FIP of 3.12, compared to a zFIP of 4.02. Together they posted a 3.90 FIP over the course of the season.

Compared to overachievers, only three of these underachievers failed to improve on their FIP in the second half, although two (Tanner Houck, Bowden Francis) didn’t make it as they were ruled out with major arm injuries. Despite a combined 5.19 FIP in the first half, pitchers on this team posted a starting zFIP of 3.90. They had a 4.11 FIP over the course of the season.

OK, on ​​to the current business.

Joe Ryan tops this year’s FIP list for overachievers, but you’ll find that it has very little impact on his actual projection because he does this so often. Remember, this is not a guess, and if a player ignores what his tracking stats suggest for too long, ZiPS will care less about its expected stats cabinet. While ZiPS doesn’t really buy what we’re seeing from Michael Soroka, I think he’ll still be happy with his zFIP since most of his past six seasons have been affected by injuries.

Juan Mejia is perhaps the most intriguing of the minors, a smooth, hard-throwing youngster who had little success for the Rockies last year. His 3.98 FIP is already below his 5.79 ERA, so by dropping the FIP even lower, ZiPS cuts his ERA roughly in half. There is a real strike at the top; a hard-throwing reliever with a contact rate of less than 75% must have two K/9 numbers. I don’t – about 80% of shortstops with a fastball velocity of at least 96 mph and a contact rate of less than 75% have a K/9 of at least 10 hitters, while only 42% of other relievers do. I also like that ZiPS thinks Adrian Morejon was better than his 2.47 FIP.

Homers for pitchers are a terrible statistic. The simplest explanation for this may be that xFIP actually works, in the sense that while xFIP isn’t a perfect statistic, pitchers’ homers are so variable that even the opposite assumption — that all pitchers should allow homers at a league-average rate — gives you more predictive value than FIP. So expected homers statistics tend to be very useful, even though they are on the variable side. The most interesting one here is Jacob Misiorowski, as ZiPS thinks he should have the lowest home run rate allowed among starting pitchers, at 1.0%. (He relieves Mason Miller and Adrian Morejon on the Miz fringes.) Misiorowski wasn’t far from making the FIP list for the underdogs, either.

ZiPS is also more optimistic about the Reds’ rotation than I am.

Surprisingly, Skenes’ zStats haven’t been up to their usual level this season, and this was before his recent slump. Not that his numbers are bad here, it’s just that they show him a little short of his Cy Young form from last year. Kiri Oler has more on Skenes.

Skenes also appears at the top of the board he shouldn’t want to be on. I should note at this point that a zFIP of 3.40 is still an ace field. ZiPS sees a slight decline for Cam Schlitter, but like Skenes, if Schlittler reaches his 3.29 zFIP for the season, he will be an elite starting pitcher.

ZiPS thinks Sandy Alcantara has a lot to do in his second year back from injury, and while Zac Gallen has been bad, there’s at least hope that he’s not actually a bottom-six pitcher forever. This isn’t the first time ZiPS has misunderstood why Tim Herrin isn’t a hitter, though not to the level of early Nathan Eovaldi.