I’ve wondered for awhile if, for expected record purposes, blowouts should have their runs capped. If any team has given us reason to consider this in more depth, it’s the Pirates.

Right now, the Brewers and Pirates both are 13-16. The Brewers have scored and allowed the same amount of runs, while the Pirates have a deficit of 77 runs, leading to an expected record of 8-21 (so says Baseball-Reference). Of course, the issue with both run differentials is that they are the direct result of their clashes from April 20 to 22, in which the Brewers outscored the Pirates 36-1 (8-1, 8-0, 20-0). Take those games out of the equation and the teams look even, which probably is more appropriate to their true abilities.

In my understanding of the theory, there should be an in-game run differential after which runs for Pythagorean record purposes should be capped. Maybe it’s just me, but it’s a bit superficial to say that the Pirates’ ability to lose by 20 runs has much impact on their ability to win or lose other games. Even if the Pirates give up 1000 runs this year, April 22′s ignominy would be two percent of their runs allowed on the season, which is quite a lot for a game that only contributes one loss.

Since I’m interested in the topic but rather simplistic in my understanding of it, I leave it to the faithful readers. Is there a point at which the effect of blowouts ought to be minimized in calculating Pythagorean records? Is Pittsburgh’s situation so rare that we ought not tinker with the formula? Debate it how you want, though I suggest a cheese dip for the afterparty.

David Pinto said...

Great teams tend to win big, and horrible teams tend to lose big. Those Pirates blow outs tell us much more about the Pirates (given the small sample of games) than their close wins.

What’s more interesting is the Brewers, who appear only to be able to win big, unless Gallardo is on the mound.

Mike Savino said...

I thought that was the whole reasoning behind the pythag projections? That winning by 1 run wasn’t sustainable but blowing teams out was. So, I feel like the whole idea is to measure the blow outs.

Brandon Isleib said...

But is winning by 20 runs sustainable or a measure of anything in particular? If it isn’t, should its impact be limited? That’s what I’m after. I agree with the general principles; I’m just curious if there’s an upper limit on one game’s utility to the run differential.

Tree said...

Losing by 20 runs does tell you more than losing by 10 runs, and there already is sort of a diminishing returns built into the way pythag works (the difference between a huge negative differential and a slightly huger differential is small.)

The main problem with trying to refine the pythag formulas is that, at some point, you are better off using team/player projections and something like Bill James’ log 5 method to talk about what should have happened.

Dave Studeman said...

I’ve played with all sorts of ways to change the pythagorean formula, including capping runs scored and/or allowed in a single game at various levels. I’ve never significantly improved the accuracy of the basic formula.

I have a feeling I’ve written about this lots of times on this here site, but I’m too lazy to search for the articles.