If you fed Kershaw's numbers into a random generator and set the parameters to create 1,000 sample sizes of 189 IP (his actual amount), you'd get a range of results that would include some sample numbers that match nearly identically, what Kershaw actually produced.
Of course, many more would show better and much better results- some even better than his regular season results, but you would find some samples like the one he actually had.
Now, of course, there are not 1,000 Kershaw-type pitchers that have 189+ IP in the playoffs, so you cannot prove the actual results match evenly with computer generated results, but studies have shown when you enter all the regular season data and generate a random sampling of expected results, it matches up with reality.
The actual amount of pitchers who far exceed their regular season numbers in reality match up with the same amount in the randomly generated sample size, and same with barely better, the same, barely worse and much worse.
The random samples mirror the reality samples.
Did I explain it well enough?