Starting Kershaw in Game 4 is the wrong call

— -- After dropping two games to the Washington Nationals, the Los Angeles Dodgers are fighting for their playoff lives. That, combined with this weekend's postponed game, left the Dodgers with an unexpected rotation crunch. Before the delay of Game 2, the Dodgers had a very clear opportunity, if they so chose, to pitch their top two starters twice each, Clayton Kershaw on short rest in Game 4 and then Rich Hill on four days' rest in the possible Game 5.

The rain threw that proverbial wrench in the works, meaning that now if Clayton Kershaw pitched on short rest in Game 4, that Rich Hill would also have to pitch on short rest in Game 5, or the Dodgers would have to pitch Julio Urias or some combination of the two in Game 5. With Game 4 and Game 5 both being necessary wins, the Dodgers were forced to decide whether to go with Julio Urias in Game 4 and Clayton Kershaw with five days' rest in Game 5 or short-rested Kershaw in Game 4 and that aforementioned combo in a possible elimination game.

In the end, the Dodgers went with Clayton Kershaw in today's Game 4. The comparison that some are trying to make is the Orioles going with Ubaldo Jimenez instead of Zach Britton. But that doesn't work logically. You use your best relievers early in extra-inning games because of the uncertainty of future leverage compared to current leverage. The benefit of using Zach Britton in the 11th is because it increases the chances the team wins and never has to turn to the much inferior Ubaldo Jimenez in the 12th. The Dodgers are not in that situation. Kershaw could throw a perfect game, striking out 27 batters on three pitches, and it doesn't prevent the Dodgers from having to win with an inferior pitcher on Thursday (not to bash Urias or Hill, but everyone's pretty much inferior to Kershaw). They have to win two games. Kershaw's excellence can't make them only have to win one.

In the days of a five-man rotation, something which has now been around for about 40 years, there isn't a lot of truly significant data on how an individual pitcher fares with three days' rest versus four or five days. But we do have aggregate data for pitchers as a group, whether all pitchers or top pitchers.

Let's start by looking at the postseason, starting with the beginning of the wild-card playoff era in 1995. Ninety-one pitchers have started on three days' rest in the postseason over that time frame, and overall, their teams did not fare well. The teams that started pitchers on three days' rest went 38-53, a .417 winning percentage. For reference, only a single team in 2016 was worse than .417 during the regular season (the Minnesota Twins).

Nor was it simply happenstance. The pitchers starting on three days' rest did in fact pitch worse. Collectively, those 91 short-rest pitchers put up a 4.53 ERA in those games. And it's not as if these pitchers were simply worse. Teams aren't likely to use pitchers on short rest whom they don't actually want pitching at all. Going through the games one-by-one, I calculated the expected ERA for the pitchers in those games based on their regular season ERA. If the short-rested pitchers pitched as well on short rest as they did on normal rest, you'd have expected those 91 pitchers to combine for a 3.57 ERA, nearly a run -- or 27 percent -- better than they actually did.

The same story was true when we look at only truly elite players in these postseason games. Limiting the games to only pitchers with an ERA under three during the regular season gives us a list of elite starters (Maddux, Kershaw, Randy Johnson, Brown, Santana, Hudson, Lackey, Pettite, etc). If those pitchers pitched as well on short-rest, you'd have expected a 2.58 ERA in those games. They put up a 3.32 ERA, or 29 percent worse.

Or in a shorter number of words, playoff pitchers on short rest have allowed nearly a third-more runs than you'd expect from them on normal rest. The short rest and the fact they generally pitched worse resulted in them also pitching fewer innings, 5.3 per start instead of their normal 6.5. For the elites, it was 5.8 innings per start as opposed to their regular-season 6.9 average.

So how does this change the probabilities? I projected the outcomes of the final two Nationals/Dodgers games based on two scenarios. For the Kershaw Game 4 scenario, I project with Kershaw pitching 27 percent worse than normally projected, over one fewer inning than usual, and Urias and Hill each pitching four innings in Game 5. For the Kershaw Game 5 scenario, I project with a normal Urias start and a normal Kershaw start. I'm assuming Joe Ross and Max Scherzer pitching for Washington in both scenarios.

The Dodgers have to win both games, and the two scenarios listed don't look promising for their chosen path. And there are a few intangible factors that might even make the idea worse.

First, Kershaw's coming off back problems, and even though he'll say he's fine, no pitcher is going to turn down a chance to get into a game like this unless his pitching arm has been amputated (and some might insist on "going? Pat Venditte" anyway). Second, there's actually some evidence that inexperienced pitchers get more of a home-field advantage than more experienced pitchers, a small factor, but one that has to be noted: You're not doing a young pitcher any favors by starting him in an elimination game on the road instead of at home.

No matter what happens, the Dodgers made the wrong decision here. They maximized their chance of playing a Game 5 rather than their chances of winning Game 4 and Game 5, the prerequisite to attending the NLCS.