The offseason is baseball’s present tense. During the season, the moments are fleeting, as nearly every day produces more information to add to the pile of what we think we know about individual players and teams. Even in the midst of unexpected storylines unfolding, we know that tomorrow is very likely to challenge us with something new. The offseason is different. It’s the only extensive moment during the year when past and future collide to form a discernible present. The information collected, now a past waiting for interpretation, is tacked on to the amoeba-like mass of data from the slightly more distant past and given temporary shape. That is where projections come from.
Based on pre-season projections, the 2014 Rockies experienced fortune and misfortune. For the misfortune, one needs only to recall that Carlos Gonzalez and Troy Tulowitzki played 161 games—combined. But, as David Martin at Rockies Review reminded us in August, “the what if game can be played both ways.” Players such as Charlie Blackmon and Tyler Matzek, for instance, exceeded projections. One of my favorite projection tools is found at Baseball Prospectus. No professional projector maintains that the figures their systems produce are rigid, but Baseball Prospectus lists the range of expectation based on their PECOTA projection system. In other words, each player card has a projected performance ranging from the 90th percentile—unexpected in a good way—all the way down to 10th percentile projections—unexpected in a bad way. Below, I take a brief look at 10 position players who had at least 250 plate appearances in the context of which percentile they ended up fitting in to based on PECOTA’s pre-season projections. I focus on two metrics. The first is True Average (TAv). TAv is Baseball Prospectus’s all-in-one hitting metric based on linear weights, which means that doubles are worth more than singles. It’s weighted to look like regular batting average, so it’s easy to understand what the figures mean–.340 is excellent and .230 is the opposite of excellent. The other metric is Wins Above Replacement Player (WARP), which is BP’s version statistic of total value. For position players, that means defense is included. Both metrics are park adjusted.
The takeaway is to show that in this present moment, “regression” filled prognostication doesn’t always fit—but sometimes it does.
Lowered Floors and Upward Regression
The most disappointing performance among all Rockies’ position players from 2014 was Carlos Gonzalez. CarGo’s TAv was a dismal .234, and he actually found himself on the wrong side of zero in terms of value. He had a -0.7 WARP. His 10th percentile projection had him at .267 TAv and 1.9 WARP. CarGo’s 2014 was truly a worst case scenario. The caveat with using 2014 pre-season projections after the 2014 season is that the next set of projections will be different. His floor will be lowered. Nevertheless, an improved season for Carlos Gonzalez is a sure bet—if healthy. Earlier this offseason, I wrote that given CarGo’s age, contract, and past performance, he offers good value either as the starting right fielder heading into 2015 or as a trade chip. I’m more convinced of that now.
Josh Rutledge continues to unimpress. Troy Tulowitzki’s incumbent injury reserve fits into two centiles. His bat, which is where his value is supposed to come from, was in the 20th percentile of his pre-season projections with a .242 TAv. Due to poor defense according to Baseball Prospectus’s measurements, his WARP was even worse. He was essentially replacement level, which placed him in the 10th percentile of pre-season projections. Rutledge has had back-to-back seasons of severe underperformance. I can see his 2015 projection being dramatically lowered. It’s hard to even see how the pre-season median projection—situated in the 50th percentile—optimistically forecasted a .258 TAv. If you’re looking for regression upward, look elsewhere. Rutledge might fit on a team somewhere in Major League Baseball, but a team like the Rockies where he gets entirely too much playing time isn’t one of them.
Of all the position players on the Rockies last season, I was most critical of Wilin Rosario. Where he ended up regarding pre-season projections explains why. His .238 TAv and 0.4 WARP place him in the 10th percentile. Rosario finished the season with a strong September, at least while standing rather than crouching at the plate, but I’m still not convinced. Rosario’s 28 home run rookie season was propped up by a home run to fly ball ratio resembling that of Giancarlo Stanton. It went down in 2013, but he still got hits. But Rosario’s .292 batting average in 2013 was propped up by a .344 batting average on balls in play. Given his player profile—he’s not fast and didn’t hit a lot of line drives—that, expectedly, also came down. I think 2014 Rosario is closest to the player we should expect. If he’s back next season, it’s reasonable to expect some improvements at the plate, but not much.
If you’re looking for a player most likely to improve (not counting CarGo because it would be surprising if he didn’t improve), it’s the last one: D.J. LeMahieu. D.J. finished the season with a .226 TAv and 0.2 WARP. Both performances fit snugly into the 20th percentile of his projections. As opposed to Rosario and Rutledge, who will likely see their floors lowered, I see D.J. LeMahieu as someone where the “upward regression” designation might apply. That doesn’t mean he’ll become a world-beater. But he doesn’t have to because he’s a good defender. If LeMahieu improves to a .250 TAv he’ll be among the most valuable second baseman in the National League. I base some of this speculation on an article Rob Arthur wrote earlier this season. Arthur, one of the most innovative baseball analysts around, designed a method to predict break-out seasons. Essentially, he uses PITCHf/x data to measure trends regarding zone-distance—the distance a pitched ball is from the center of the strike zone. Essentially, increased distance from the zone reflects pitcher-adjustments due to the perceived skill of a batter. Because there’s little risk in pitching down the middle to a terrible hitter, increased distance from the zone hints a touch of fear with what the batter might do given a pitch to hit. The method has proven affective, and D.J. is among the ten players Arthur identified who saw their zone distance increase in 2014. What we can hope for is that it means he’ll hit more doubles in 2015. From someone who is likely to hit either seventh or eighth in the daily lineup and offer Gold Glove defense, that’s a boon.
Charlie Blackmon finished the season with a .260 TAv and 1.5 WARP, both of which place him in the 70th percentile of his pre-season PECOTA projections. But that result clouds just how precisely PECOTA projected Blackmon to be, albeit in much fewer expected plate appearances. You remember, I trust, that Charlie had an incredible April where his batting average ranged between .400 and .600, and he hit five home runs. This month earned him the additional playing time, although injuries to Carlos Gonzalez and Michael Cuddyer also had something to do with it. The last day Blackmon’s batting average sat above .400 was April 27th. From April 28th until the end of the season, his line was .269/.314/.401. PECOTA’s pre-season median projection was a line of .266/.320/.398. The difference in value resulted from his home runs. He hit 19 whereas even the 90th percentile projection had him hitting just eight. The lesson is this: we know Charlie Blackmon as a baseball player, and it turns out PECOTA already knew what we now know. And that is a player who can respectively play a lot more than he should if necessary.
Troy Tulowitzki finds himself in this category because he missed playing time. In fact, his performance tells two tales, one unfinished. The first is his TAv. Tulo’s .331 TAv is in the 80th percentile of his pre-season projection—the lofty regions of a player established as one of the very best hitters in all of baseball. The other story is his WARP. Playing in just over half of a 162 game season, Tulo accrued 4 WARP, which was just in the 30th percentile of his PECOTA projection. Remember that WARP is a counting stat. Tulo was well on his way to reaching the upper portion of his projections before injury. I expect that his 2015 centile projections will resemble 2014, which maxed out at 7.7 WARP in the 90th percentile, with a disaster season in the 10th percentile being one in which he’s worth about three wins. In other words, a 10th percentile Tulowitzki season roughly equates to a 90th percentile season from LeMahieu.
Raised Ceilings and Downward Regression
Two-thousand-fourteen Justin Morneau was the bizarro 2014 Carlos Gonzalez. While the latter had a dreadful season that nobody can reasonably expect him to replicate in 2015, the former exceeded expectations. Morneau finished the 2014 season with a .298 TAv and 3 WARP. His 90th percentile PECOTA projection placed him with a .301 TAv and 1.9 WARP. The difference in WARP appears to be based almost entirely on defense. PECOTA expected negative value on defense, but he ended up on the plus side. We know more about Justin Morneau than other players because he’s been around so long. If Morneau’s 2015 resembles his 2014, it will mean that he re-produced what might be the absolute most he can do at this point in his career. You might see minor upward adjustments regarding his future projections, but you’re more likely to see his performance go down.
The same applies to Drew Stubbs. Stubbs’s .270 TAv and 1.1 WARP are in the 80th percentile of March forecasts. It looks like Stubbs hit the ceiling of what he can do as a player in 2014. Earlier this year, Matthew Gross suggested that Stubbs might fit on the Rockies more than other teams. Nevertheless, that his 2014 strikeout rate resides north of 30 percent and his BABIP north of .400 are factors that cannot be ignored when looking forward to 2015. Regression fits neatly here.
Regression doesn’t fit for Corey Dickerson. Dickerson finished the season with a WARP of 2.3, which was in the 80th percentile of his pre-season projection, and a TAv of .304, which was in the 90th percentile. If Dickerson replicates his 2015, it shouldn’t be terribly surprising. In fact, I will guess that if he does reproduce the same TAv it will be in the 70th or 80th percentile. That is, his ceiling is going to be raised for future projections. One of the reasons for this is because Dickerson doesn’t quite the trail of numbers behind him that someone like Drew Stubbs does. Additionally, Dickerson’s production is supported by an excellent line drive rate. Of all hitters in baseball with at least 450 plate appearances, Dickerson ranked eighth in line drive rate. That is why Dickerson’s .356 batting average on balls in play is a testament to his skill rather than his fortune. It’s also one reason why Dickerson’s excellence is not the same as Stubbs’s.
Finally, there’s Nolan Arenado. Arenado’s 4.3 WARP was in the 90th percentile of his forecast, while his .273 TAv was in the 70th. Like Dickerson, Arenado is a player whose ceiling has to be adjusted. He exceeded expectations, but not in a way that should cause anyone to appeal to regression I’ve already written a paean to Nolan Arenado’s youth and skill, which bears a link but not repeating here.
In the lower regions of the projections, the 2014 Rockies had one unanticipated disaster (Gonzalez), a couple of unsurprising poor outputs (Rutledge and Rosario), and a player whose season most fans were happy with but with a lot of room to improve (LeMahieu). Added to that, they had a slight overperformance that initially looked like it would be a massive one (Blackmon) and a half-season worthy of a full campaign (Tulowitzki). In the upper regions, the team had two big overperformers (Stubbs and Morneau) and a couple of seasons that demand raised ceilings (Dickerson and Arenado). Things did go wrong for the team, but other things went right. We still have a couple of months before PECOTA and other projection systems input the most recent past to declare what will come. When they do come out, pay attention to them, but also keep in mind that the projections are delimited at extremes, and that mental adjustments can apply to some players, but not others.