FanPost

Red Sox first half surprises and busts (OPS contest results)


For the second spring training in a row, I challenged Over The Monster readers to try their hand at predicting the first-half OPS's of the Red Sox players this year. The first half (first 81 games) ended just over a week ago, and I was ready with all the numbers on Wednesday night - but alas, as a working dad, sometimes writing posts just takes a while. Here we are now. I hope there is some interest.

I must wonder, though ... this time, instead of the healthy number of entries received last year, there was only ... one. Well, two, if you count mine, but I can't claim any credit for drawing my own interest. Perhaps folks realized that there aren't any actual prizes for the winners? Well regardless, thank you, Gtpx, for entering!

And it turns out, you are the winner! Your overall score of 366.4 came in under my score of 398.6 (lowest score wins). These scores were arrived at by multiplying, for each player, the contestant's error in OPS by the player's number of plate appearances through the first 81 games, then adding these.

But a contest with two contestants isn't terribly interesting. So to make this more interesting, back in spring training I recorded the season predictions made for each Red Sox player by the projection systems ATC, ZiPS, Steamer, and Fans. I did this in last year's contest, too, and every projection systems smoked every human, with the humans having been too optimistic.

This time, that didn't happen:

Contestant Score
Gtpx 366.4
ZiPS 373.0
ATC 388.4
Steamer 396.0
Fans 396.7
Tomisphere 398.6

Gtpx beats all! Humans can beat formulas!

Or, um, in my case, lose to all the formulas. Hey, at least you all know I'm not rigging the results.

I had planned to have a look at just why Gtpx did so well, and why I didn't, but instead I think it will be more interesting for everyone if I talk about what the results have to say about the individual players.

Everybody's best guess was for the same player - Sandy Leon. His actual OPS was just above .657. The best guess was off by less than .001. (That was mine, btw, .658. At least I can do one thing well.) The guesses ranged from .655 to .675, the narrowest range by far, with a standard deviation among them of .0084, the smallest of those by far. Had we looked at this narrow spread in the guesses back in spring training, we might have predicted that Leon would come the closest of anyone to these predictions, and that's exactly what happened. He's also the only player who can be said to have ended up where he was projected to, neither above nor below - every other player on the team either exceeded all six predictions by a significant amount, or fell below all six predictions by a significant amount. We might conclude from this that we can expect Sandy Leon's bat to stay right about where it is at.

We might then wonder, based on this, whether the player that had the second-narrowest spread of guesses came in pretty close to the predictions for him, and if we were to speculate that was the case, we'd be exactly wrong. The second-lowest spread of guesses belonged to Mookie Betts, ranging from .855 to .887, with a standard deviation of .0124. Yet everybody's worst guess (when weighted by plate appearances) was for Mookie Betts, whose OPS came in at 1.111, which is 0.237 above the average guess of .875.

What does this say about Mookie? One, he made some real changes to his approach at the plate this spring, and they've made the difference. Like Mike Trout, he's not only gifted athletically, but he's also humble and constantly looking to improve his game, reasons why Betts and Trout now stand above all other players in baseball in terms of results and value added. Two, he's been freaking awesome this year, because despite having had the second highest OPS predictions on the whole team (second only to J.D. Martinez), he's still exceeded predictions by more than anyone else. He's been a whopping 19 standard deviations over the average value of the predictions for him, more than double anyone else on the team. It'll be exciting to see if he can finish the year strong like he has in years past.

So if Betts was the farthest above expectations, who was farthest below? I'm not counting Sam Travis nor Dustin Pedroia, both of whom had only 13 plate appearances, nor Tzu-Wei Lin, who had only 41. Blake Swihart comes in as farthest below, his .404 OPS having been .265 below the average prediction of .669 for him. Certainly a disappointment, but with only 82 plate appearances, one can't be too disappointed.

What if instead of asking who was farthest above or below expectations, we ask, whose results were most surprising? I see this as a different question, because there were differing amounts of certainty in the original predictions. When there was a lot of variability in the predictions, I take that to mean we weren't very sure of our predictions. Standard deviation of the predictions makes a good measure of this variability. When we measure that way, we see the following.

Most surprisingly poor:

Player Avg prediction Actual OPS Std Dev above
Rafael Devers 0.820 0.708 -8.4
Jackie Bradley Jr. 0.777 0.612 -7.1
Blake Swihart 0.669 0.404 -6.2

In this sense, Devers' results are most surprising, although all told all three are probably a pretty similar level of surprising.

Devers now seems to have righted what was wrong with his approach or his swing, so here's hoping he'll continue bringing that OPS up. All three have brought theirs up in the week since we hit the halfway point. That could assure playing time for Bradley and Devers, but Swihart may get a trade.

Now for those who were most surprisingly good:

Player Avg prediction Actual Std Dev above
Mookie Betts 0.875 1.111 19.1

He deserves to be set apart.

Others who exceeded everyone's expectation:

Player Avg prediction Actual OPS Std Dev above
Brock Holt 0.681 0.799 8.4
J.D. Martinez 0.936 1.050 5
Mitch Moreland 0.766 0.908 4.8
Andrew Benintendi 0.826 0.892 4
Xander Bogaerts 0.807 0.844 1.7

Others who fell below everyone's expectations:

Player Avg prediction Actual OPS Std Dev above
Christian Vazquez 0.702 0.568 -4.5
Eduardo Nunez 0.758 0.649 -4.3
Hanley Ramirez 0.814 0.708 -4

Where we expected:

Player Avg prediction Actual Std Dev above
Sandy Leon 0.664 0.657 -0.8

Xander Bogaerts did come close to joining Sandy Leon on the "where we expected him to be" list, but he did come out above expectation for everybody. And I will just end there, because this post is a week late already.