Hello! I’ve spent a little while in one of my spreadsheets, because I’m working up another thing. But that doesn’t mean I can’t provide a smaller thing in the meantime. As I’ve mentioned on a few occasions in the past, I have projected team records going back to 2005. Of course, the methods aren’t all consistent, because certain projections haven’t existed for that long, but all projections follow the same general rules — use a best-guess depth chart and then project player performance based on what those players have done in the recent past. I’m not saying the 2005 projections were as good as the 2016 projections, but they weren’t crazy. So let’s look at a little data! You don’t have anything better to do.
First, and most simply, here are projected wins and actual wins, for all 360 team-seasons.
There’s enough signal there to know the projections are onto something, and there’s enough noise there to keep baseball entertainingly unpredictable. The greatest over-achiever since 2005: those 2012 Baltimore Orioles, who won 93 games after having been projected to win a measly 70. The greatest under-achiever since 2005: the 2012 Boston Red Sox, who won 69 games after having been projected to win an impressive 91. In other words, the Orioles won like the Red Sox were supposed to, and the Red Sox lost like the Orioles were supposed to. I guess you could say the numbers were right, but they were misplaced.
How have the individual team breakdowns looked? I’m not including this because I think it’s in any way predictive. It’s just here to sate some curiosity. I calculated error in two ways. Here’s one, where I took the absolute value of each miss, and then added them up over the 12 years.
The projections have had the greatest error with the Indians, missing by an average of almost 10 wins per season. At the other end, welp, check out the Yankees. The average error there is about three wins per season. The Braves aren’t even particularly close to that. For whatever reason, the Yankees have been reasonably predictable over the past decade and change.
Here’s the other way of calculating error, just subtracting projected wins from actual wins over the time window. Absolute values have no place here.
How to read this: The Rangers have ranked 15th in projected wins, but they’ve been seventh in actual wins. Hence their error of +48. The Mariners are tied for 20th in projected wins, but they’ve been 26th in actual wins, hence their error of -45. The Rangers are out in front here by nine wins; the Mariners trail the next-worst team by 10 wins. I don’t think this means anything about the teams moving forward, but this provides some partial background, when you consider how various fans respond to the 2017 projections over the offseason. Projections have looked wrong before, and they’ll look wrong again. If they didn’t, we’d hate them.
