On Tuesday, the Hall of Fame did something unusual: it actually did us proud. It was the first time since 1955 that as many as four players had been elected to the Hall, and Randy Johnson, Pedro Martínez, John Smoltz, and Craig Biggio were all eminently worthy—that's something we should appreciate. Would I have liked to see even more make the grade? Of course. But we knew entering the 2pm announcement that this was likely the best we could hope for.
We knew that thanks to pre–Hall of Fame election "polling" done by Darren Viola and Ryan Thibs, who do a great public service by collecting and recording ballots made public before the announcement. Public ballots don't tell the whole story, though, since they're biased toward types of candidates, so for the past several years I've "unskewed" them to predict the final outcome. With some adjustments to my model this year—and, surely, more than a few strokes of luck—I'm humbled to say they did better than ever in 2015:
The exit polls, as taken from Viola's HOF Ballot Collecting Gizmo, ended 2015 with a larger-than-usual average error of 4.7 percentage points (the median error was four points). My projections carried a mean deviation of 2.6 points and a median of 2.5 points. According to tweep Ben Dilday, the root mean square of my projections was the second-lowest among Hall of Fame projection systems this year, behind only Dilday's himself (I'm self-taught at this stuff so I'll admit to a lot of his methodology going over my head, but his uncanny projections are here). One of the systems I beat out was father-of-sabermetrics Tom Tango's, which is kind of like beating Meryl Streep at the Oscars—a huge honor that I'll enjoy for the brief time it will last.
Some of my more notable calls included not being fooled by the polls' over- and underrating, respectively, of Tim Raines and Lee Smith. However, in my mind, these calls aren't that impressive; Raines and Smith have become very very predictable over the years in how much they underperform and overperform their polls. Others I called within a percentage point included John Smoltz (that one I am proud of, considering he had no over- or underperformance history to go off), Roger Clemens, Barry Bonds (those two's middling showings have become as foreseeable as the sunrise), Jeff Kent, and Larry Walker.
While I am quite gratified, there are always things I could have called better. This year, my biggest misses were overestimating Curt Schilling by 7.8 points, overestimating Jeff Bagwell by 5.4 points, and underestimating Gary Sheffield by 5.2 points. Schilling is perhaps forgivable, as he suffered an almost unprecedented 19.2-point gap between his public- and private-ballot performance. Meanwhile, I'm still baffled by Sheffield—a surly steroid user who is actually more popular with the sport's crusty traditionalists?!
However, Sheffield shared his good fortune with most of the other players who made up the bottom of the ballot, including Nomar Garciaparra and Carlos Delgado, which I did not expect. However, I'm now wised up to it for next year. Nomar does seem to be a perfect intellectual heir to the newly departed (from the ballot) Don Mattingly, who always gained ground relative to his pre-election polling numbers. Like Mattingly, Garciaparra was an elite player with an awesome peak, but injuries cut short out both their careers. Narrative-driven voters may dwell more on their mythology-buoyed memories of Mattingly's and Garciaparra's primes—and it certainly doesn't hurt that they had those primes on the biggest stages, in New York and in Boston.
Finally, one small but important prediction that I slightly muffed was voter turnout, which I use to predict the proportion of public-to-private ballots. Instead of my predicted 570, turnout dropped slightly from past years to 549 voters. If I had nailed the turnout, could my projections have been even better? As it turns out, they would have been worse—a mean error of 2.8 points and a median error of 3.1. This is what I mean when I say I got lucky!