The Los Angeles Dodgers deserve a ton of credit for how their 2017 season turned out. Manager Dave Roberts steered his club to 104 regular-season wins and sailed through the playoffs before facing a Houston Astros team playing at just as high of a level. But for a manager who isn’t afraid to employ analytics to drive his decisions, he either ignored the numbers or was simply looking at the wrong ones before Game 7 of the World Series.

Roberts decided to go with Yu Darvish on the mound for the winner-take-all game Wednesday night, a curious decision given how poorly Darvish fared against the Astros in 2017.  Overall, Darvish had allowed a weighted on-base percentage — a version of on-base percentage that accounts for how a player reached base, instead of simply considering whether a player reached base — of .340 to the Astros in 2017 before taking the mound in Game 7 — and that includes his starts against Houston as a member of the Texas Rangers this season. To put that wOBA against in perspective, Houston led the majors with a .349 wOBA during the regular season, the highest mark by a team over the past five years.

So, to answer the question of “how did Darvish fare against the Astros in 2017?” The answer is, “not great.”

Darvish’s disastrous outing in Game 3 of the World Series provided further evidence. He allowed four earned runs and six hits in 1.2 innings, resulting in a game score — a metric devised by Bill James to determine the strength of a pitcher in any particular baseball game — of 26, a career low among his playoff performances and the fourth-worst by a starter during the 2017 MLB playoffs.

Cheers, tears, and hugs among fans as the Astros win their first World Series. Rough Cut (no reporter narration).

Despite all those metric-minded red flags, Roberts stuck with him, only to see Darvish throw another clunker: three hits, four earned runs in 1.2 innings, a performance so bad it reduced Los Angeles’s chances of winning the World Series by 35 percentage points before the second inning was over. That was even worse than Game 3, when he lowered the Dodgers’ title hopes by just 11 percent.

LAD Pitcher in Game 7 Win expectancy at start of appearance Win expectancy at end of appearance Difference
Yu Darvish 45% 10% -35%
Brandon Morrow 10% 11% +1%
Clayton Kershaw 8% 5% -4%
Kenley Jansen 5% 6% +1%
Alex Wood 3% 2% -2%

This all could have been avoided, perhaps, if Roberts had instead let Clayton Kershaw start the game on two days rest.

Begin with the obvious issue: Kershaw was going to be throwing ahead of schedule. But, 1) it’s not like there’s a Game 8 for which you need to save him. And 2) Kershaw has been pretty darn good on short rest.

Kershaw took over for Darvish in the third, striking out four while walking two, allowing another two hits during his stint in relief.

Performance in Game 7 of 2017 World Series IP H ER BB SO ERA
Clayton Kershaw 4.0 2 0 2 4 0.00
Yu Darvish 1.2 3 4 1 0 21.60

In addition, Kershaw allowed a wOBA against of .249 to the Astros before taking the mound in Game 7, which, when compared to Darvish, is roughly the difference in performance between the best and worst hitting clubs in the majors this season.

If short rest was the key deterrent in Roberts’s decision, it shouldn’t have been. Statistics show starters asked to pitch on short rest (three or fewer days) in the playoffs aren’t as bad as you might expect. According to FanGraphs, run expectancy, RE24 in short hand, tells you the number of expected runs given the current number of outs and placement of base runners, with higher numbers indicating more runs saved by a pitcher. Since 2012, the first year under the current playoff format, starters on short rest allowed just a fraction of a run more than expected in these playoff situations when you take into account men on base and outs left in the inning (minus-0.21 RE24). Compare that to pitchers on normal five days’ rest giving up only half a run less than expected (0.42 RE24) and it’s hardly an impact at all, considering what is at stake.

Not to mention Kershaw has been really good when asked to pitch on short rest during the playoffs, only once out of five postseason appearances allowing more runs to score than expected.

Season Result Days rest IP H ER BB SO Game Score RE24
2013 NLDS, Game 4 W 4-3 3 6 3 0 1 6 67 0.552
2014 NLDS, Game 4 L 2-3 3 6 4 3 2 9 59 -0.349
2015 NLDS, Game 4 W 3-1 3 7 3 1 1 8 74 2.147
2016 NLDS, Game 4 W 6-5 3 6.2 7 5 2 11 49 0.611
2016 NLCS, Game 2 W 1-0 2 7 2 0 1 6 78 3.500

Factor in that Kershaw is a much better pitcher than almost anyone in the majors and the choice should have been clear who should have been in to start the game.

Read more on the World Series: