The Projections and You, Revisited

As always happens around this time of year, we’re arguing the utility and accuracy of WAR as a player-evaluation statistic and model. Every year, the argument gets a little smarter on both sides, but every year, there’s the same struggle over acceptance. I’m sure we’re not done talking about this, not with Dave Cameron included as a BBWAA voter for the Most Valuable Player award, but you might recall that, a few months ago, the argument then was about the utility of projections. More specifically, it was about projections and identifying breakouts and collapses. Again, it wasn’t a new argument, but sometimes it’s the repeating arguments that manage to push us all forward.

Around that time, in the middle of June, I published a post with 20 players and 20 polls. There were five offensive over-achievers, five offensive under-achievers, five pitching over-achievers, and five pitching under-achievers. In each poll, I asked the audience to select how they felt about the player’s projection going forward. Which possible collapses or breakouts were people buying? Which were they dismissing? I know the season’s not over yet, but I thought this could be a fine time to look back on the post and on what has happened since.

The most important point in this whole post: this isn’t a study, with far-reaching conclusions. From this, we’re not definitively learning anything about projections or human intuition. We’re essentially reviewing 20 small samples of data, and then seeing what that might mean, and I beg of you not to make more of this than I am. For me, this is just sating my own curiosity. Anyhow, let us proceed to the comprehensive table:

Player	Projection	Poll	Performance	Difference
Brian Dozier	0.317	64% looks low	0.321	1%
Michael Brantley	0.334	59% looks right	0.362	8%
Lonnie Chisenhall	0.342	79% looks right	0.286	-16%
Nelson Cruz	0.355	80% looks right	0.325	-8%
Victor Martinez	0.356	59% looks right	0.406	14%
Jedd Gyorko	0.304	76% looks right	0.336	11%
Brad Miller	0.308	67% looks right	0.350	14%
Mike Moustakas	0.312	59% looks high	0.293	-6%
Domonic Brown	0.331	67% looks high	0.306	-8%
David Wright	0.347	62% looks right	0.310	-11%
Garrett Richards	3.75	58% looks right	2.59	-31%
Dallas Keuchel	3.83	67% looks high	3.75	-2%
Jake Arrieta	4.12	59% looks right	2.46	-40%
Jake Odorizzi	4.28	67% looks right	3.65	-15%
Collin McHugh	4.36	59% looks right	3.20	-27%
Justin Verlander	3.62	58% looks right	3.79	5%
Matt Cain	3.83	74% looks right	4.10	7%
Shelby Miller	3.95	61% looks right	4.94	25%
Clay Buchholz	4.12	58% looks right	3.62	-12%
Felix Doubront	4.28	52% looks right	4.60	7%

Here’s a fun fact for you: the five offensive over-achievers were projected for an average .341 wOBA. Since the post was published, they’ve put up an average .340 wOBA. Meanwhile, the five offensive under-achievers were projected for an average .320 wOBA. Since the post was published, they’ve put up an average .319 wOBA. That’s remarkable in its simplicity. Again, small samples, and again, don’t draw too much from this, but the projections have had those particular players nailed.

There’s a different story on the pitching side. The five pitching over-achievers were projected for an average 4.07 FIP. Since publishing, they’ve put up an average 3.13 FIP. The five pitching under-achievers were projected for an average 3.96 FIP. Since publishing, they’ve put up an average 4.21 FIP. Four of the five over-achieving pitchers have meaningfully beaten their June 16 projections. Four of the five under-achieving pitchers have done worse. It’s also hard to know what to make of Justin Verlander — he was projected for a 3.62 FIP, and he’s come in at 3.79, which would be a very mild decline. But he’s also yielded a 5.00 ERA, and that’s dreadful, and this is just what happens sometimes when you’re trying to analyze pitcher performance. You don’t know which category is necessarily the most meaningful.

So, about the polls. This won’t surprise you, but it’s a mixed bag. The player for whom people believed the projection the least was Domonic Brown. Indeed, he’s undershot his projection by 8%. People also figured the projections were too high on Mike Moustakas, and he’s undershot his projection by 6%. Yet, two-thirds of people were willing to buy Dallas Keuchel, and he’s basically equaled his projection. Brian Dozier has equaled his projection, despite the same kind of vote distribution. Voters were a lot more willing to believe in Keuchel than they were willing to believe in Collin McHugh, and it’s McHugh who’s continued to pitch like a front-of-the-rotation starter. Not that Keuchel’s been bad or anything, and he did have a wrist issue for some time, but before the post he had a 16.2% K-BB% and since publishing he’s come in at 7.7%.

At the other end, people were relatively unwilling to buy Nelson Cruz, and he’s sure enough slowed down. Same for Lonnie Chisenhall, despite his dinger just a little while ago. Three-quarters of people were also unwilling to think Jedd Gyorko was a pile of crap, and he’s rebounded to be better than average. The crowd kind of missed the boat on Matt Cain, who’s now out with surgery. They also missed on Jake Odorizzi.

There is some other stuff you can try to take from this. You can do that at your leisure! And it’s interesting that, in even the most extreme case shown, 20% of people still didn’t believe the projections. You can say that 80% of people had the right idea about Nelson Cruz, but even there one out of five figured Cruz had made an improvement. One out of five people figured that Chisenhall had made an improvement, and he’s actually gotten a good deal worse. So, each case provides a little mental fodder, for your consideration, but maybe the best thing to be done with all this is to try the same thing again in 2015, and in 2016, and so on and so forth. Maybe, in time, we could build a real sample of polling data and results. If not, at least we’d get to think about baseball.