Spring training is a time for pitchers to rebuild their arsenals, hitters to find a rhythm against live competition and teams to begin forming their identity for the season. It’s also a time for statistical forecasts, which — let’s be honest — is the kind of preseason prep we are especially fond of here at FiveThirtyEight. But this year’s forecasts are a fascinating mix of the confident and the uncertain — and with good reason.
The 2020 MLB season, shortened to 60 games and played under pandemic conditions, was one of the weirdest campaigns in memory, even if it ended with the best team rising to the top. And since forecast systems rely on data from previous years to inform their predictions — usually giving the most recent season the most weight by far — the strange nature of 2020 will naturally have a big effect on how we look ahead to 2021.
This subject is one the fantasy guru Ron Shandler deliberates at length in the 2021 installment of his celebrated “Baseball Forecaster” series. “Let me present an idea that nobody wants to consider but could make too much sense to dismiss,” he wrote. “In trying to figure out what is going to happen in 2021, the most prudent path might be to make believe 2020 never happened.” Although the experts Shandler surveyed from throughout the fantasy-baseball industry didn’t go quite that far, their consensus was to weigh 2020 data only about half as much as they would a normal season heading into the following year.
That means players who had unusually good — or bad — 2020 campaigns should probably be viewed with a healthy dose of skepticism.
For instance, after a solid debut season at age 28 in 2019, San Francisco’s Mike Yastrzemski produced the equivalent of 7.2 wins above replacement1 per 162 games in 2020 — an MVP-candidate pace, if it were a normal year. But clearly, the less weight we give 2020 stats (relative to other seasons), the less valuable Yastrzemski projects to be looking ahead to 2021. The same goes for older hitters who revived their careers in 2020 (AL MVP José Abreu), pitchers who performed well beyond their 2018 or 2019 numbers (Corbin Burnes, Yu Darvish) and even phenoms who enjoyed huge breakouts in the short schedule. Yes, San Diego’s Fernando Tatís Jr. will almost certainly remain one of the best players in baseball over a full season, but even his stellar 2020 numbers need to be regressed to the mean more than we would usually do for a 22-year-old wunderkind shortstop, since he still has played only 143 regular season games in his entire MLB career.
We can see this effect at work when looking at whose weighted average WAR from the previous three seasons would change the most depending on whether we apply a standard series of weights — treating 2020 like just any other season (with a weight of 3, giving weights of 2 and 1 to 2019 and 2018 respectively) — or a modified weighting scheme based on Shandler’s consensus (a weight of only 1.8 for 2020, 2.8 for 2019 and 1.4 for 2018):
|WAR/162 in…||Weighted Avg.|
|Fernando Tatís Jr.||SDP||+0.0||+3.9||+7.7||+5.1||+4.1||-1.0|
At the other end of the spectrum, the same effect applies to players who were unusually bad in 2020. Pitchers coming off a big injury, like Washington’s Stephen Strasburg, are always hard to predict. But what to do with hitters like J.D. Martinez, Christian Yelich and Alex Bregman, all of whom had 2020 seasons that fell well short of what they had done over the previous few years? Should they get a pass for those performances — and if not, how much should it be held against them? Beyond that, an even bigger question might surround players who didn’t play at all, choosing (perhaps wisely) to opt out of the chaotic short season entirely. (For what it’s worth, the majority of experts polled by Shandler voted to take those players’ pre-2020 projections and simply age-adjust them forward by a year.)
|WAR/162 in…||Weighted Avg.|
Either way, our views of many players in 2021 will change depending on how much we consider 2020 to have really happened in a meaningful way (or not).
Does this projection logic extend to teams as well? Certainly the champions of 2020 — the Los Angeles Dodgers — are not seeing any signs of regression in their forecasts. According to the weighted mix of projections we use to set preseason Elo ratings,2 L.A. has 100.8 wins of talent on its roster, easily the most in baseball. (The Yankees are No. 2 at 95.5 wins.) That’s due in no small part to the free-agent acquisition of Trevor Bauer last month, to go with the team’s ludicrous base of existing talent — headlined by Mookie Betts, who led MLB in WAR last year. According to research from ESPN’s Stats & Information Group, Bauer is only the second reigning Cy Young Award winner to join the defending champions the following offseason, enrolling in a club that had previously contained only Roger Clemens (who went from Toronto to New York before the 1999 season).
But the Dodgers are the exception, not the rule, this preseason. Compared with the same March forecasts taken from 2019 and 2020, this year’s projection shows many more teams bunched in the middle of the pack, with 80 to 85 wins:
This surely speaks to factors beyond uncertainty around how projections should treat 2020 stats; the postseason structure MLB used last season (and wants to use again) rewards spending less for midpack results, for instance. But just the same, it’s also true that we simply know less about which players (and teams) will be good or bad going into 2021 than in a normal season. That doesn’t necessarily mean we should toss out everything we saw with our eyes last year. But it does mean that we should look back on 2020’s numbers with much more skepticism than usual as we try to balance them against previous seasons — and stats from 2021, when those finally start piling up.