How Shortened Seasons Affect Future Projections
Unusual situations create interesting problems for analysts to solve. While it’s still not clear just what the 2020 season will look like — if indeed there is a 2020 season — the one thing that’s guaranteed is that baseball this year will probably look very different than any season in living memory. (Incidentally, I won’t say all-time: 19th century baseball saw a team disband during warmups because no one was in attendance and some of the home squad left with the visiting team.)
Interesting challenges are usually fun to tackle, though admittedly I much prefer those that don’t involve a lack of baseball. The general belief around the game is that we will get some baseball this year, making the challenge of the moment figuring out what a shortened season will mean for the projections; regardless of what form this season takes, come 2021, I’ll have to churn out ZiPS projections, which should prove to be a trickier-than-typical offseason task.
The natural hypothesis is that in the short-term, the projections will be worse than usual, due to greater uncertainty and simply fewer games played by the players in question. If the season opens in late June or early July, it’s likely that teams will play somewhere in the neighborhood of 100 games. The good news is that while we haven’t had a lost season, we have two strike-shortened seasons, in 1981 and 1994, that can help guide us. They’re not 100% comparable — 1981’s strike was in the middle of the season and 1995 was slightly shorter as well — but they’re the main historical comps we have to look at.
ZiPS projections didn’t exist during these seasons, as I was in high school in 1994 and was just figuring out how to use a toilet in 1981. So to get an idea, I ran some very simplified projections, using only the basic data and simple aging, a stripped-down projection that’s similar to Tom Tango’s Marcel the Monkey forecasting system. Read the rest of this entry »

Dan Szymborski