MASTER NOTES: 2019 Pitches per out

We all have a different way of watching a game on TV when one of our fantasy pitchers is on the mound. My approach is to curl up in the fetal position and peer at the screen through my fingers.

Ha ha! Just kidding! I can’t look through my fingers because they’re holding onto a tall glass of Maker’s Mark, especially when Mike Leake is pitching. Actually, I sit upright and pay attention. And I pay particular attention to how many pitches my guy is using to get his outs.

Over time, based on this informal estimating tool, I had come to the conclusion that a pitcher getting around one out per every five pitches was going well and had a pretty good chance at a good game—he’d get deeper into the game, improving his chances for a win; he wouldn’t be allowing many baserunners, improving his chances for a WHIP-positive outing; and those fewer baserunners would mean fewer runs against, improving his chances for an ERA-positive performance.

But we all know that this kind of informal, anecdotal, observational info is not the stuff of which decision should be made. So I thought I’d look into it. And you know what? I was pretty close!

I used BHQ’s PQS logs (from the Leading Indicators page) to get the data from all the MLB starts through Tuesday. A bit of jiggery-pokery with Excel, including filtering for non-“opener” starters with 10+ starts, and I had a pretty interesting info set.

First, a couple of overview bits. Of the 2,187 starts by those qualified starters, the average pitches per out (P/O) was 5.5. But in wins, the average was 5.1 P/O, and in losses it was 6.4 P/O. Put in reverse, 50% of starts with 5.1 P/O or lower were wins, versus 26% of starts over 5.1 P/O.

As well, P/O correlates well with both ERA and WHIP. For pitchers with 10+ starts, the season-long correlation is 0.43 (on a scale where 1.0 is a perfect correlation and 0.0 is no correlation), while WHIP correlates at 0.53.

The P/O data next made me wonder which pitchers this season have the best ratios. So I stacked ’em all up, from lowest to highest (again, only pitchers with 10+ starts, although I’m now thinking of looking at relievers as well).

Seventeen pitchers were at or below that 5.1 P/O cutoff mentioned earlier, led by Hyun-Jin Ryu of LA, at 4.8 P/O, and the list contains many of the names you’d expect: Kershaw, Tanaka, Greinke, Hendricks, Berrios, Verlander.

There are also a few names that might raise an eyebrow but not come as a shock: Yonny Chirinos of TAM (only a couple of hundredths behind Ryu at the top of the table), Michael Soroka of ATL, German Marquez of COL, Ross Stripling of LA, Joe Musgrove of PIT, Brett Anderson of OAK, Miles Mikolas of STL, and C.C. Sabathia and Domingo German of NYY.

One name really might shock you. Glenn Sparkman of KC is seventh in the rankings at 5.0 P/O. This might be a sample artifact—Sparkman barely scraped into qualifying with exactly 10 starts, and his 842 pitches so far this year are only about half of the others on the list.

But Sparkman makes a more important point by showing the dangers of a metric like this one: He has an anomalously good ratio, but he's not a good pitcher. Sparkman is a very low strikeout pitcher, 5.2 K/9 this season and 5.6 K/9 in 13 total starts over two seasons in KC. Low strikeouts tend to reduce pitch counts and, therefore, P/O rates. But Sparkman's ERA this season is 4.54, which is pretty meh, and looks overcooked considering his xERA is well over five. For his career, his ERA is over five, and his xERA higher still.

So don’t rush out to grab Glenn Sparkman. Take the P/O metric as a first indicator, or maybe as a tiebreaker. And even if you don’t have moves to make, keep track of how many pitches your pitchers this weekend are taking to get their outs. You’ll have some idea of how likely you are to get a win, if nothing else.

* * *  

While I was organizing the PQS logs in the spreadsheet for this week’s edition of Master Notes, I also ran a quick study to see how pitchers’ ERAs and WHIPs might have been affected by the worst starts of their seasons.

In a nutshell, I logged every pitcher with 10+ starts, calculated a score for each start using the formula ERA + (3 x WHIP), and then subtracted out his worst (highest) score from each pitcher's overall record.

Many, even most, of the biggest impacts among starters’ ERAs involved pitchers whose ERA and WHIPs are pretty bad even without their one bad outing, and just slightly less horrible when that outing is removed. I saw a lot of 6.00-plus ERAs dropping into 5.00-plus territory, and a lot of 1.50 WHIPs dipping to 1.40.

But some bad outings affected good pitchers. The first of these is Pablo Lopez of MIA. His decimals at the time of the study were 4.23/1.12, which is pretty good. But his May 10 outing at Citi Field, was a catastrophe: 4 innings, 10 ER, 12 baserunners. Take that outing off Lopez’s record, and his decimals improve to 3.18/1.00, and he looks much more like one of the best pitchers in baseball.

Another pitcher really blasted by a bad start was Domingo German of NYY (also a low-P/O guy from above), whose already impressive 3.49/1.07 would fall to 2.88/1.04 had it not been for a 5-inning, 7-ER/9-BR stinker at KC on May 26.

David Price of BOS: His 3.16/1.15 is impressive, but knock off the 6 and 6 he gave up June 13 at home versus TEX, and he’s 2.59/1.10.

And Hyun-Jin Ryu: Already otherworldly at 1.78/0.97, he starts bordering on intergalactic at 1.29/0.87 if we wipe off his June 28 start in COL, 4 IP, 7 ER, 10 BR. (Not surprisingly, @COL starts are by far the most common among these duds, along with @LA, @MIL, and @PHI.)

The obligatory caution is this: What has happened is not reliably predictive of what is going to happen. And naysayers about this kind of study argue, persuasively, that the bad start did happen and should not be discounted. I’m not so sure I agree. There’s a reason we call outliers “outliers,” and it is in large part because we believe they can be discounted as being so anomalous as to be non-indicative of a player’s true potential.


Click here to subscribe

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