MASTER NOTES: A new metric?

During my interview with the fine sportswriter Michael Salfino on the most recent edition of BaseballHQ Radio, he happened to mention that he pays close attention to relievers’ ISO-against stats.

His reasoning was that to score on a pitcher who is stingy with extra-base hits, his opponents must string together three positive events—singles, walks, steals. By contrast, pitchers who give up Extra-Base Hits (XBH) can be scored on more easily: A double followed by a single will do the job in most cases.

This got me thinking. If it is positive for relievers to avoid XBH, the same might also apply to all pitchers. XBH avoidance might be a useful validator, maybe even a predictor, of pitcher performance. At first, it looked like I was onto something. Later, though, it turned out to be fool’s gold.

I started at baseball-reference.com, and got the stats for all 429 pitchers who had least 20 IP last season. Then I calculated how many Total Bases (TB), Extra Bases (XB) and XBHs each pitcher gave up in three different ratios:

  • Per ball in play;
  • Per nine innings; and
  • Per batter faced

For the purpose of this study, the BIP totals are made up by subtracting Ks, walks and HBPs from Total Batter Faced. So it includes HRs (which are not included in other BIP ratios, like BABIP).

Then I would correlate each combination against ERA, and compare it with the BaseballHQ skills metrics of Dom (K/9), Cmd (K/bb), which have correlations around 35%, which is quite useful.

I was most interested in the BIP ratios, because they take strikeouts out of the calculation. Obviously we want to target high-strikeout pitchers, because there’s nothing better for the ol’ ERA than a whiff or 14. So the idea was to see if we could find the desirable arms when we’re looking at that next tier of pitchers.

Here’s what I found…

All the ERA correlations with the Total Base ratios were much stronger than the BHQ skills metrics. Even the lowest ERA correlation—XB/BIP— was 58%, 23 points better than the skill-alone metrics.

The best correlation was between TB/9 and ERA, a remarkable 88%. But that measure includes strikeouts (innings are made up of all outs, which includes Ks). But not far behind was the TB/BIP ratio, at 75%, which is all the more remarkable in that it is a measure of ERA expectation that explicitly excludes strikeouts.

As if to accent the point, five of the top-10 TB/BIP guys were under the 20% league average K%, as were 18 of the top-50. The best pitcher on the list was Brad Ziegler, who has been a serviceable low-cost closer over the years despite a very low 14% K rate.

The best on the list also included Mark Melancon, who was thought this pre-season to have a tenuous tenure as the PIT closer because of low strikeouts. Other targetable relievers included Will Harris, who captured the closer role in HOU, Sam Dyson, who landed TEX and Ryan Madson (OAK), all of whom have done some very effective closing despite being just a few points above the league K% average.

Among the starters (minimum 50% starts), the top performers looked like the Cy Young ballots in both leagues: Arrieta, Greinke, Kershaw, Keuchel, deGrom, Cole, Bumgarner, Price, Scherzer. While all of these starters also had solid strikeout rates, the low-TB list also had low-K stalwarts like Marco Estrada, Chris Young, Erasmo Ramirez, John Lackey, Rick Porcello and Jordan Zimmermann.

By now, this was getting exciting. But the real test would be checking to see if a pitcher’s TB/BIP performance in a season predicted his performance in the next year.

Uh, not so much.

While the first-year HQ skill metrics maintained a 30%-ish correlation with next-season ERA, the TB/BIP correlation with next-season ERA fell to 11%—so weak that is closer to random than to predictive (this years’ TB/BIP was still 73% correlated to this year’s ERA).

The deal was sealed with one last discovery: last year’s TB/BIP barely predicted this year’s. The correlation between 2015 TB/BIP and 2016 was just 13%.

The idea behind BABIP is that once the ball is struck, the batter is powerless to determine its outcome. The new stats coming out, which let us see bat velocity and launch angles, seem to suggest batters do have some influence on the BIP outcome. And indeed we knew as much when we realized a few years ago that individual batters establish individual BABIPs (which we at BHQ call Hit Rate).

But the same is assuredly not the case for pitchers. Outcomes on BIP remain essentially random—not just hits, but also XBH and TB.

It’s a little frustrating to see something shiny in your pan, only to find it is fool’s gold. We can now only conclude that the connection between TB/BIP and in-year ERA is that both of them are facets of the same events on the field. A pitcher allows lots of TB on his BIP, so his ERA rises. But unlike actual skills, those numbers change in other periods.

But it isn’t all a lost cause. If nothing else, this ultimately futile chase for a new actionable connection between events on the field reminds us of an important rule in analysis: correlation is not always causation.


Click here to subscribe

  For more information about the terms used in this article, see our Glossary Primer.