Flu-Like Symptoms: Giving ‘Em LIP

On June 13, Andrew Miller relieved Trevor Bauer with the bases loaded and two outs in the sixth inning of the Indians' game against the Dodgers, tied at 2-2. He struck out Yasiel Puig to end the inning and retired all three batters he faced in the seventh. In the top of the eighth, though, he allowed a home run to Cody Bellinger and a single to Yasmani Grandal. He was pulled in favor of Bryan Shaw. Miller took the loss.

The next day, again in a 2-2 game against the Dodgers, Miller entered in the top of the eighth inning. He got Puig on a fly ball but then allowed home run, a single, two walks, and an RBI grounder before getting pulled. He again was charged with the loss. His manager, Terry Francona, in a display of candor, later admitted, “I think I pitched him too much.”

Now, Andrew Miller has pitched 35 2/3 innings this year. (All 2017 figures are through Tuesday, June 20.) That’s not that many. He’s currently tied for 170th among major-league pitchers for innings. Among relievers, he’s tied for 16th. His workload, contrary to Francona’s mea culpa, doesn’t seem to be excessive, although certainly the manager worked his ace reliever very hard down the stretch and throughout October last year.

Of course, even focusing on this season alone, it’s not just the innings, or the pitches, we increasingly hear on broadcasts. It’s the high-leverage innings. Pitching in garbage time doesn’t take as much out of a hurler as pitching with the game on the line. Yusmeiro Petit of the Angels, for example, has pitched 42 2/3 innings—more than Miller. But in his 27 games, Petit has appeared in 17 (63.0 percent) with Los Angeles trailing. Miller has pitched in 31 games, entering only three (9.7 percent) with Cleveland trailing. That’s more pressure.

We can take a crack at calculating this. We know how many innings pitchers pitch, obviously. Noted sabermetrician Tom Tango created the Leverage Index, which quantifies the leverage of each situation using win expectancy tables. The average Leverage Index, or LI, is 1.0. An LI less than 1.0 is a lower-leverage situation; one greater than 1.0 is higher leverage. In the example above, Petit’s average Leverage Index, throughout all his appearances this year, is 0.85. Miller’s is 1.60. Miller’s been used in high-leverage situations. Petit hasn’t.

By using average LI in relief appearances (available from our friends at FanGraphs) and reliever innings pitched, I calculated a new metric with the stupid acronym of LIP, which stands for Leverage Index times Innings Pitched. Bill James once said something to the effect that a new metric is useful if it mostly confirms what we already know. Here are the 10 relievers who had the highest LIP Index in 2016. See what you think of them.

That works for me. It’s a combination of closers and setup guys. The common denominator is that they were all very effective, so their managers didn’t hesitate to bring them in with the game on the line.

That last sentences touches on a key objection to LI: It’s not under the pitcher’s control. That’s a fair criticism. Dave Robertson and Nate Jones were both in the top 10 in large part because of the type of games the White Sox played (i.e., close) and the way manager Robin Ventura deployed them (in tight situations), not because of anything they did. All true. But that doesn’t minimize the fact that both found themselves in a lot of high-leverage situations.

So I’m willing to go the LIP Index for now. Which leads to three questions.

Are pitchers today really "softer"?

Back when Goose Gossage was getting 25 multi-inning saves every year (actually, the most he ever got was 21, but never mind), relievers threw more innings, so even if the leverage wasn’t always high, the LIP Index was higher, right? It turns out this is true. Here’s a table of the 15 highest LIP Indices since 1974, the first year for which LIs are available.

Well, yeah, OK, so Gossage had a point. Those leverage indices aren’t meaningfully different from the ones in 2016. The innings are, though. The highest-ranked pitcher from the 21st century is Billy Koch of the 2002 A’s: 2.09 LI, 93 2/3 IP, 195.8 LIP, 43rd since 1974. The highest-ranked pitcher from the current decade is Dellin "Not A Closer" Betances of the 2015 Yankees: 1.87 LI, 84 IP, 157.1 LIP, 114th since 1974.

So let’s accept that relievers today don’t compile the same leveraged workloads as their predecessors. That’s interesting, but there’s nowhere we can go with it. Nothing’s going to change.

Here’s the same list, just for the 30-team era since 1998:

Which 2017 teams and relievers may be overworked?

How about the current game, though? Which bullpens may be overtaxed? I’m going to skip the LI and IP details and just list team LIPs so far this season.

And here are the highest-LIP pitchers so far in 2017:

What does a high LIP index portend?

Well and good, but what does it all mean? Is Francona right that too many high-leverage innings affect a pitcher? Are the Rays, Brewers, and Blue Jays wearing down their bullpens, or are their high LIP Indices just a product of near-.500 teams playing a lot of close games? Are Cleveland and Pittsburgh—both with two relievers in the top 15—wearing its best relievers to a nub?

To check, below is a table that shows the 15 pitchers with the highest LIP Index at the All-Star break last year. In the following three columns, I’ve listed their change in ERA, strikeout percentage, and walk percentage after the break. Positive changes are good, negative changes are bad.

For example, Jeurys Familia had a 2.55 ERA, 24.0 percent strikeout rate, and 8.8 percent walk rate at the All-Star break. After the break, he had a 2.55 ERA, 28.7 percent strikeout rate, and 10.7 percent walk rate. That works out to a +0.00 change in ERA, a +4.7 percent (favorable) change in strikeout rate, and a -1.9 percent (unfavorable) change in walk rate.

Here’s how it worked out last year for the LIP leaders at the break.

That’s not what I was expecting, either. Of the 15 pitchers with the highest LIP Index at last year’s All-Star break, all but five had a better ERA and a higher strikeout rate after the break. All but five had a higher walk rate, but that didn’t do much damage. (And yes, those figures all easily exceeded the league averages of a 0.12 improvement in ERA, 0.3 percent increase in strikeout rate, and 0.1 percent increase in walk rate in the second half.) Overall, working a reliever hard, at least given last year’s results, does not appear to be well-correlated to a breakdown during that same season.

Unless, as in the cases of Jeanmar Gomez, Santiago Casilla, and, yes, Betances, last year, it does.