The PECOTA projections from Baseball Prospectus came out yesterday, and once again they hate the Royals. That's OK. However, with the release of PECOTA, there are four major projection systems that have released their data for the upcoming season. Let's take a look at how each system compares on the main position players, including Jarrod Dyson, Paulo Orlando, and Christian Colon.
I grabbed Steamer and ZiPS from FanGraphs, Marcel from Baseball Reference, and PECOTA from Baseball Prospectus. One thing to note about these projections - Steamer, ZiPS, and PECOTA are actively managed by people whose goal is to improve the projection. The Marcel projections, developed by Tom Tango, are meant to serve as a baseline which any projection system should aim to beat. Since its inception, Marcel has not been updated (as far as I know). Marcel's beauty is in its simplicity.
Wins Above Replacement
You'll find that in calculations of overall value, PECOTA never comes in "first". For no player does PECOTA project the highest WAR of the three systems that project some WAR value (Marcel does not have a WAR value on Baseball Reference).
The largest differences lie with Salvador Perez and Mike Moustakas. For Perez, my guess is that framing data unique to Baseball Prospectus factors in. BP's catcher framing metrics hate Perez and think he's a terrible pitch framer. In the case of Moustakas, PECOTA just really does not buy into his 2015 improvement. Other than those two, PECOTA is not terribly far from the other two systems but is consistently lower.
On Base Percentage
First, notice that I set the minimum in the x axis to .250 to show the differences a little better. Regarding on base percentage, there are some similar trends. PECOTA is often the lowest, but it does project Salvador Perez's OBP as the highest of the four systems. The spread in his projected OBP is very small though.
Two things jump out at me (aside from the PECOTA hate). First, Marcel really likes Christian Colon. ZiPS, Steamer, and PECOTA project a .318, .316, and .308 OBP respectively, but Marcel has a .337 OBP. In other words, Marcel projects a well above-average OBP for Colon, while the other three project a below-average OBP (after removing pitchers from the league average). Colon has all of 168 plate appearances in the bigs, but he has a .361 OBP in that time. It looks like a small sample size is revealing differences in weighting methodology between the systems. I just don't know what that difference is.
Second, Steamer likes Mike Moustakas a little more than the other systems. It definitely buys into Oppo Moose more compared to the other three. In an oldish-looking Powerpoint presentation, Dash Davidson noted that what makes Steamer unique is that they "use a different system for each component (K%, BB%, HR%...)." Without a deeper knowledge of the methods, I can't say much about why Steamer likes Moustakas more.
First, notice that I set the x axis to a minimum of .300 to show the differences a bit better. Regarding slugging, there are a couple things (or thing) to notice. Marcel likes Paulo Orlando, Christian Colon, and Jarrod Dyson to slug the ball around the diamond a bit.
After not showing much power in the minors, Orlando hit seven dingers in 251 plate appearances at the big show. Adding in his three homers from AAA, Orlando hit 10 homers in 2015. Orlando hadn't hit double-digit homers since a 2010 stint in AA. I would imagine that Marcel is weighing his MLB data more than the other systems. I wouldn't blame you for disagreeing with Marcel here. Colon is yet to hit a homer in the bigs, so yea. Dyson's "power" comes from legging out extra base hits. His .380 slugging in 2015 likely weighs more heavily on Marcel than the other systems since Marcel does not account for speed vs. actual power.
The systems all see heavy regression for Lorenzo Cain. Cain hit the ball with authority, limiting soft contact and increasing hard contact at the same time. His HR/FB rate climbed, his ground-ball rate fell, and his popup rate fell. There are a lot of indicators to say that his 2015 performance was for real, so maybe the projections are underrating him a bit. They see numbers in agreement with 2014 rather than 2015.
Despite Perez's steady trend downward in offense, consisting of increased aggressiveness and overuse, the systems all see an offensive rebound in about the same neighborhood. They all see a below-average OBP but still better than both 2014 and 2015, while they also see a SLG similar to last year.
The systems all buy into Kendrys Morales' rebound, though they see 2013 (119 wRC+) much more than 2015 (131 wRC+). Like Cain, Morales increased his hard contact rate, elevated the ball more, and popped up less.
I think those are all the interesting things to note. Here is a Google Doc with the numbers. Pitchers to come next time.