ZiPS vs. Steamer, 2014: Position Players

A short while ago, I wrote a post comparing and contrasting the Steamer projections and the FAN projections. It can be potentially really interesting to see where different projections diverge, but that particular exercise faced two problems: fan bias, and fan-projection sample size. It wasn’t as satisfying an exercise as it could’ve been, and I’m okay with that, but now we have ZiPS all nice and uploaded, and we can try this kind of thing again. On which players do Steamer and ZiPS most disagree? Is there anything to be learned?

This’ll be broken down into two posts — looking at position players today, and looking at pitchers tomorrow. And it’s worth noting this is a slightly different exercise from the one involving the fans. Fans might be informed by observations, or by gut feelings, or whatever. Steamer and ZiPS will arrive at their conclusions based more or less on the same data. So, you’d expect them to be very similar, and of course they are. It’s interesting, then, where they are not.

Dave wrote this post up almost a year ago. In spirit, this is the same thing, with some tweaks to the methodology. To get to the numbers I wanted, I also had to narrow down the pool of projected baseball players. I’m not real interested in guys in the low minors, and neither, as it happens, is ZiPS. I’m also not real interested in the majority of part-time players. I wound up eliminating everybody who wasn’t projected for at least 300 plate appearances by both projection systems. This left me with 286 guys to compare.

Now, when Dave did this, he ran with projected WAR per 600 plate appearances. I did that too, first, but then I gave it a second thought. What we’re really interested in isn’t where Steamer and ZiPS disagree on baserunning. Likewise, we aren’t all that interested in where Steamer and ZiPS disagree on defense. We want to know where they differ on offense, so why not just look at wOBA and cut out the rest? So, I chose to go with wOBA, which, conveniently, doesn’t require any kind of adjusted denominator. Though it can be meaningful where projection systems disagree on other elements of the game, hitting is the real core of what we care about.

What you’re going to see below, then, are projected wOBAs. For the pool of 286 players, Steamer projects an average .329 wOBA, while ZiPS projects an average .324 wOBA. In order to find the biggest differences, I had to make an adjustment, involving means and standard deviations. It made only a very small difference. That’s probably enough, so why don’t we get to a table? Here are the ten hitters Steamer likes more than ZiPS the most:

Name	Steamer	ZiPS
Wilin Rosario	0.357	0.326
Travis Snider	0.325	0.296
Michael Morse	0.344	0.315
David Freese	0.339	0.311
Norichika Aoki	0.338	0.312
Kole Calhoun	0.345	0.319
Marc Krauss	0.320	0.297
Angel Pagan	0.331	0.308
J.B. Shuck	0.307	0.285
Logan Forsythe	0.314	0.292

Just missing: Will Middlebrooks, then Corey Hart, then Kendrys Morales, then Abraham Almonte. ZiPS doesn’t seem to think so much of the Mariners. In that way, ZiPS is like a lot of people.

The table is something of a mixed bag. There are younger guys, and more proven veterans. There are weaker guys, and more proven sluggers. There’s no bigger positive difference than there is for Wilin Rosario. One way of looking at this is that Steamer likes him more, by 31 wOBA points. It could also be said that ZiPS projects a raw wOBA that’s basically average, whereas Steamer projects a raw wOBA that’s more than a full standard deviation above the mean.

Steamer figures Rosario will cut down the strikeouts a bit. ZiPS goes in the other direction. Steamer figures Rosario will increase his walks more than ZiPS does. Steamer also projects power in between Rosario’s 2012 and 2013, whereas ZiPS puts him much closer to his 2013 figure. ZiPS actually projects a slightly higher BABIP, but put everything together and you’ve got a wRC+ difference of 113 to 92. It’s substantial, as ZiPS doesn’t seem to think so much of the over-aggressive type.

Look at the table again. Rosario was as productive in 2013 as he was in 2012, but his core rates declined and he was lifted by a high BABIP. Snider is coming off a miserable season. Morse is coming off a miserable season. Freese is coming off a down year, relatively speaking. Forsythe is coming off a miserable season. Middlebrooks just missed the cutoff, but he’s coming off a miserable season. It would appear that Steamer is considerably more forgiving when it comes to a rough stretch that follows a more encouraging track record. Steamer, in other words, might believe more in bounce-backs, whereas ZiPS is a cold-hearted cynic.

As for Calhoun, Krauss, and Shuck? Both Calhoun and Krauss were tremendous in Triple-A, with as many walks as strikeouts. Shuck had twice as many Triple-A walks as strikeouts, even though he didn’t have any power. Steamer seems more willing to give credit for high-minors production, and Krauss in particular is interesting, since Steamer sees a 23% strikeout rate while ZiPS comes in at 30%. He struck out 31% of the time in the majors. He struck out 17% of the time in Triple-A.

Time to flip things. Here are the ten hitters ZiPS likes more than Steamer the most:

Name	Steamer	ZiPS
Alfonso Soriano	0.315	0.336
Bryce Harper	0.363	0.382
Mark Trumbo	0.338	0.357
David Ortiz	0.376	0.392
Carlos Quentin	0.344	0.359
Billy Hamilton	0.286	0.302
Marlon Byrd	0.319	0.334
Nick Castellanos	0.313	0.328
Jarrod Saltalamacchia	0.305	0.320
Raul Ibanez	0.313	0.326

Just missing: Ryan Braun, then Jacoby Ellsbury, then John Jaso, then Anthony Rendon. There’s a lot of recent missed time for those guys, which might be some kind of factor.

Interestingly, a year ago, ZiPS liked Harper and Ortiz a lot more than Steamer. Here they are again, along with a host of others, making up another mixed bag of old guys and the considerably lesser-proven.

With Soriano, Ortiz, and Ibanez, my guess is that Steamer is applying a more aggressive aging curve, because all of those guys are up there. They’ve also recently been very productive, and ZiPS is more willing to give them future credit for that. Now, what’s the opposite of an old guy on a presumably downhill trajectory? How about Nick Castellanos or Billy Hamilton?

Steamer doesn’t hate Castellanos — ZiPS just believes in his power more. As for Hamilton, there’s little difference, except in the BABIP category, with Steamer coming in at .300 and ZiPS coming in at .332. Now, his minor-league BABIP is .354, and that’s encouraging, but then again his BABIP in Triple-A a year ago was .310 so it’s not like anything is automatic. BABIP is one part speed and several more parts quality of contact, and with Hamilton, that’s the big question.

Trumbo showing up makes me wonder if there’s something weird going on regarding his transition from Anaheim to Arizona. With Quentin, ZiPS believes a little more in his 2013 BABIP, which was the highest of his career. And ZiPS thinks Harper will keep growing. Steamer puts him right on his career line. ZiPS projects a coming power boost, which I don’t think would take anybody by surprise. A year ago, ZiPS nailed Harper, while Steamer seemed to underrate his ability.

It seems like ZiPS is kinder to older players. It also seems like ZiPS is kinder to prospects and youth, but then you have that Steamer trio of third/fourth outfielders with Triple-A success. Steamer also seems to favor players coming off down seasons, relative to what they’d done previously, so ZiPS might put a little extra weight on the most recent year. It also seems like ZiPS might be a bit more forgiving of recent missed time, but then Steamer’s a lot higher on, say, Corey Hart.

So the differences don’t appear to be of one or two types. But there are differences, and that’s interesting enough, and as always, the safest thing to do might be to average the projections out. Which, conveniently, is exactly what we do for our projected standings and our playoffs-odds page. It might seem like there’s no point in having multiple projection systems, since they’re all trying to predict the future based on the same information. But as long as we can identify player differences, we can know that they’re at least saying slightly different things.