clock menu more-arrow no yes

Filed under:

Splitting the run differential: blowouts and close games

New, 12 comments

Yes, I’m bringing up run differential again. Sorry, not sorry.

MLB: Houston Astros at Seattle Mariners Joe Nicholson-USA TODAY Sports

Yesterday’s loss was a bit of an aberration for the 2018 Mariners. They’ve allowed more than nine runs in fifteen games this season and they’ve lost all of them. In those losses, the deficit has been three runs or fewer just twice (including yesterday’s result). The offensive struggles of this team have made staging big comebacks extremely difficult, that’s no surprise. But rarely does the offense pick up any of the slack if the pitching staff starts allowing bunches of runs.

I’ll apologize up front about bringing this topic back up. It seems like the Mariners run differential has become this rigid pillar around which arguments have circled all season long. This latest loss to the Astros has pushed it to a season low negative 43. They’re on pace to set the major league record for the largest discrepancy between a team’s actual win percentage and their Pythagorean estimated win percentage.

Earlier this year, I looked into the Mariners Cluster Luck in the month of April. You may recall there were a number of conversations around the Mariners run differential happening around that time too. Cluster Luck attempts to provide some additional context for a team’s performance that a simple run differential may miss.

I’d like to introduce another way of looking at run differential that provides a little more context and nuance. The other day, I saw an interesting thread about run differential on Twitter. Former Mariners employee Tom Tango had noted the 12.5 game difference between the Red Sox and the Astros win percentage despite very similar run differentials. This led to a discussion about how runs were scored and large blowout wins (or losses) can skew a run differential. As Adam Levine argues, the Astros have won many more games by larger margins, giving them an artificially inflated run differential.

Well, the same principle is true for the Mariners, just the other way around. In the histogram below, I’ve plotted all of the Mariners games this season by their final run differential.

Look at all those one run wins! Look at all those blowout losses! The Mariners have lost as many blowouts (5 runs or more) as they have close games (1-2 runs). That has a huge effect on their overall run differential. And those blowout losses are not the best barometer for assessing the team’s talent level moving forward either.

In that Cluster Luck article from April, I used the Base Runs formula to calculate the Mariners expected run differential, providing some additional context to the actual results on the field. I’ve kept up with calculating the Mariners expected runs and Cluster Luck throughout the season (they’re at -6.5 luck runs right now). Let’s see what that histogram looks like when we use expected run differential.

The expected run differential results are far more distributed than the actual results. Yes, the Mariners have suffered 16 blowouts per Base Runs. But a lot of those losses by three or more runs have shifted to the right by expected run differential. The huge center column is also interesting. Those games are basically coin flips, decided by less than one expected run. All those closely contested games give the Mariners very few opportunities to pad their run differential, and yet they’ve turned them into actual wins for the most part thanks to Edwin Díaz. Base Runs also thinks the Mariners should have had a few more blowout wins of their own too, for good measure.

Lots of blowout wins and losses skew run totals. That’s easy enough to understand. The Mariners, in particular, have suffered a number of blowout losses. Scott Servais has shown a tendency to cut his losses during those games. He’ll often leave his long relievers on the mound, even if they’re allowing the opponent to add on to the already lopsided score and he’s quick to make substitutions to give his key players some rest. That creates situations where the final score is more crooked than expected.

Given what we’ve seen in these two histograms, we should take their raw run differential with many large grains of salt (this is true for any team, not just the Mariners). Are the Mariners as bad as their -43 run differential tells us they are? Almost certainly not. Is it still concerning? Absolutely.