The ZiPS (Almost) Midseason Update – American League
When looking at the differences between midseason and original projections, it’s always fun to see where reality shredded expectations the most. The American League in 2019, on the other hand, is fairly boring. We have one big surprise, bordering on the edge of truly affecting the playoff hunt, and a relatively mild switcheroo in the AL Central leader. Sure, the White Sox are a bit better than projected and the Angels a bit worse, but it’s generally a league in which most teams are at least in the same time zone as their preseason win prognostications.
So how do the ZiPS in-season projections work? For the Big Official ones, I use the full-on ZiPS model rather than the comparatively simple in-season one, to try to get the best estimates possible. Each player gets a percentile projection, with ZiPS randomly selecting from each player’s distribution to get a range of the expected roster strength for each individual team. Then each team is projected against every other team in their schedule a million times for the rest of the year. All this has the benefit of getting more accurate tails as opposed to the binomial distribution when you’re working with an assumed roster strength; one of the most important things in ZiPS is that on all layers, it’s designed to be skeptical about its own accuracy.
So let’s dive right into the American League. Read the rest of this entry »