Somewhat unexpectedly, the Astros will turn toLance McCullers to start Game 4 of the American League Championship Series today -- rather than Brad Peacock -- hoping their 24-year-old righty will look more like the All-Star he was in the first half and less like the pitcher who struggled with injury and inconsistency in the second half.
Undoubtedly, Astros fans are looking at McCullers' recent success against the Yankees and hoping it carries over here. After all, in his three career starts against New York, he has a 2.08 ERA, allowing four earned runs in 17 1/3 innings. McCullers has struck out 23, allowing only four walks. It's enough to give a pitcher confidence heading into a huge game.
:: ALCS schedule and coverage ::
Of course, you can only put so much emphasis on what happened in the past. The July 27, 2016, game had Mark Teixeira, Brian McCann and Carlos Beltran starting for the Yankees -- not Aaron Judge, Gary Sanchez and Todd Frazier. On June 30 of this year, the Yanks had Austin Romine, Tyler Wade, Ronald Torreyes and Chris Carter in the starting lineup. And, of course, we've seen very different versions of McCullers over the past 18 months. Which is to say what happened then, under dissimilar circumstances, may not tell you much about what may happen now.
While you can glean some small insights from past matchups -- there has to be something about McCullers that Chase Headley doesn't like, having whiffed in five of seven plate appearances -- McCullers hasn't faced any current Yankees player more than nine times, which means that there's relatively limited utility to be had from pitcher-vs.-batter numbers.
In fact, sometimes, they can be downright misleading. Let's focus on an example of that here, and call it our "matchup to watch" headed into Game 4. In eight career plate appearances against McCullers, Yanks shortstop Didi Gregorius has five hits. He's hitting .625, a huge number. Gregorius has a massive Weighted On-Base Average of .606, which is nearly double the 2017 Major League average of .327. (wOBA is like OBP, except it gives more credit for extra-base hits rather than treating each time on base equally. Michael Trout had 2017's best wOBA, at .451.)

When you look at those numbers, you'll surely think that Gregorius is a trouble spot for McCullers. You'll probably hear that Gregorius has "owned" McCullers; after all, a .625 batting average is huge. But is that a true indicator of skill? No stat, no matter how advanced or traditional, can really give you a strong answer to that over just nine plate appearances, but we can at least take a deeper look using our latest Statcast™ metrics.

Yes, Gregorius has a .625 average against McCullers. But his Expected Batting Average (xbA) is just .179. Sure, Gregorius has a .606 wOBA against McCullers. But his Expected wOBA is just .168.
Those are tremendous differences. They're the difference between the story being that Gregorius has been fantastic against McCullers and that he's been dominated by him. So what's the truth -- where do those numbers come from?
Both xBA and xwOBA are calculated the same way. They look at a hitter's quality of contact, based on exit velocity and launch angle, and then express the likely outcomes based on all the other similar batted balls. For example, in Game 2, Greg Bird doubled to left off Charlie Morton. But it wasn't hit hard at all, just 78.3 mph, and it was hit very high, at a 43-degree launch angle. It was a weak popup, really. Since left fielder Cameron Maybin for some reason wasn't able to get close to a ball that's often caught, it went as a double against Morton, despite having an expected average of just .040. That's how often that ball goes for a hit, four percent of the time. Morton did his job, but he wasn't rewarded.
By looking at the seven batted balls Gregorius has hit off McCullers (he struck out once, on May 12 of this year), we can do the same. While the outcomes have been great, the process hasn't exactly looked the same.
Just look at how McCullers has managed to allow those six hits:

Of those seven batted balls, only one has qualified as "hard-hit," which we define as being 95 mph or harder, a 102.1-mph single this May. Three of the singles came off contact that, based on exit velocity and launch angle, turns into hits fewer than 20 percent of the time. (Put another way, a ball with an expected average of .138 is a hit 13.8 percent of the time.)
Only one of those batted balls was really a scorched ball that's usually a hit -- that 102.1-mph single from May -- and even that was merely a hard-hit grounder to the right side between first and second. Only one ball even traveled more than 93 feet, and that was a flyout to left.
On one of those singles, the one with the .138 expected average from July 27, Gregorius looked as surprised as anyone that he'd made contact. Just look at the moment of contact, where he is completely flat-footed, basically having just thrown his bat into the zone. It was a hit only because it happened to go right down the third-base line, against a slightly shifted infield.

When McCullers and Gregorius met in June, Gregorius managed a hit with an exit velocity of just 68 mph, putting a swing on a ball well outside the strike zone -- look at the moment of contact below -- and slapping a grounder to shortstop Carlos Correa that Gregorius managed to beat out by a whisker.

These are hits, but they're not really showing a hitter that has a pitcher completely figured out, either.
It's not that those hits don't matter or count; of course, they do. If Gregorius gets three hits in Game 4 and they're all weakly hit but they help the Yankees win anyway, he'll surely be far happier than he'd be if he scorched baseballs into gloves left and right, making outstanding contact but going 0-for-4 in a Yanks loss. It's just that while it's accurate to say that he has seen success against McCullers, none of this predicts that he will see success again tonight.
We don't know enough yet to say that the advanced stats are predictive, either. Ultimately, this all comes down to the fact that we have such limited sample sizes on hitter-vs.-pitcher matchups as to render them mostly meaningless. After all, nothing will matter more tonight than whether we see first-half McCullers or second-half McCullers. That's the only question Houston's starter really needs to worry about.