"Christian Yelich has Clayton Kershaw's number," you might hear in regards to today's Game 5 of the National League Championship Series.
You'll hear that because their head-to-head numbers are gaudy, tilted in Yelich's favor: He owns a .500/.550/.889 line in 20 plate appearances, including the postseason. It's the highest OPS any hitter has off of Kershaw, among the many who have faced him at least 20 times.
:: NLCS schedule and results ::
It's a loud line. It's also all but completely meaningless, as nearly all hitter-versus-pitcher numbers are. The simplest reason for that is that the samples are too small to have any predictive value, small enough in most cases that a lucky bloop here, a fantastic defensive play there, or four at-bats on a very windy day could change the story entirely.
If the hitter happened to be playing through the flu or if the pitcher was trying to hide an injured elbow, that could change everything. There's just not enough data here.
Even in the rare cases where you do have a sizable amount of head-to-head appearances, much of that information is so old as to be useless. Take Felix Hernandez against Ian Kinsler, for example. They've faced each other 105 times, but some of that dates back a dozen years ago to 2006. Clearly, neither player is anywhere near the same now as they were back then.
Let's take this knowledge and apply it to Game 5, by looking at one matchup from both sides. In addition to Yelich and Kershaw, we can take a look at Cody Bellinger against Wade Miley. The outcomes may not be what you expect.
Kershaw vs. Yelich
.600 average / 1.067 slugging since 2015
.298 expected average / .536 expected slugging
Yelich had been in the Majors for about a month when he first faced Kershaw in 2013, and it actually didn't go very well for him: He walked once, then struck out swinging three straight times. 
They've had 16 plate appearances over six games since then, all since 2015, and these are far more interesting to us, because we can apply Statcast™ data to them. With that '13 game out of the way, Yelich's line increases to .600/.625/1.067 since. That's nine hits -- three for extra bases -- out of 12 batted balls. 
Some of those balls have very legitimately been crushed, like this July 21 home run that came off the bat at 105.5 mph. There's no talking around that.

You can say the same for this Aug. 2 blast, which Yelich hit at 109.5 mph for an estimated 381-foot home run.

But you'll notice that Yelich's expected average and slugging, derived from the usual outcomes of exit velocity and launch angle, is a more human .298/.536. That's still very good, it's just not otherworldly. 
How? Because some of those hits were, for lack of a better word, fortunate. Here's Yelich beating out a 2016 single that didn't even make it halfway to the mound.

Here's an infield single right back to Kershaw in 2015, hit at all of 82.4 mph.

And earlier in the same game, here's another grounder right up the middle, this time it is one that Jimmy Rollins was unable to convert into an out.

It's not that those hits don't matter, don't count, can't be turned into runs on the board. Of course they can. It's just that there are a lot of ways to get to a .500 average. Some of them involve actual crushed baseballs. Some of them involve some weakly hit balls. Very little of it is actually predictive.  
After all, if we'd done this entering last Friday's Game 1, Yelich's line against Kershaw would have been .529/.556/.941. It would have been viewed as a massive advantage. He struck out and walked, putting no balls in play.
Bellinger vs. Miley
.200 average / .200 slugging 
.367 expected average / .736 expected slugging
It's the opposite story on the other side, where Miley has limited Bellinger -- note that .200 average matched by a .200 slugging percentage -- while the underlying metrics suggest that Bellinger has had more success than it would appear. Of course, the entire point here is that you shouldn't really put much stock in either set of numbers, because we're talking about six plate appearances (six!) that all took place in the span of 11 days in July. This is why these numbers are generally more trouble than they're worth.
But if you must, let's at least show how Bellinger got to "1-for-6 with a walk and a strikeout," because if the Dodgers repeat their lineup from Game 2, then Chris Taylor is perhaps likely to get the start in center over Bellinger, and this is the kind of data that the Dodgers will evaluate in making their decision. 
Again, we're talking about all of four batted balls, one of which became a ground-ball single. Interestingly, however, that ball (hit at 91.6 mph) was the softest-hit ball Bellinger has had against Miley. Each of the three other times he made contact, the ball was hit hard. It just never turned into success.
The most interesting one of these came on July 31, when Bellinger hit a ball at 102.3 mph for what looked certain to be a home run ... until Lorenzo Cain stepped in. 

Cain took a hit away from Bellinger again that day, though in somewhat less spectacular fashion. Bellinger's 99.4 mph lineout had a 51 percent hit probability.

And on July 20, Bellinger had another hard-hit ball, this time a 101 mph shot up the middle that would have been an easy single, if not for the fact that shortstop Tyler Saladino had been shifted up the middle. That Saladino was unable to collect it smoothly meant that it went into the books as an error, not a hit, but it doesn't change how hard it was squared up.

If this sounds a little familiar for Bellinger, it's because leading into Game 3, we talked about how he'd hit .417/.500/1.083 against Milwaukee starter Jhoulys Chacin. Of course, one of those hits was this "triple," thanks to some ineffective San Diego defense back in 2017.

Despite that flashy line, Bellinger grounded out, twice.
Pitcher vs. hitter stat lines just aren't very predictive or informative. They can't be. There's just not enough information. So in Game 5, Yelich might crush Kershaw, and he may not make solid contact at all. Either outcome is an expected one -- no matter what the narrative might say.