clock menu more-arrow no yes mobile

Filed under:

Rockies predictions for 2016 according to Marcel

Taking a look at 2016 forecasts based on the simplest, but still useful, projection system.

Denis Poroy/Getty Images

All of us (okay, most of us [fine, many of us] {some of us?] ((me, at least)) have been eagerly awaiting ZiPS and PECOTA projections for the 2016 Rockies. The anticipation isn't because the projections are gospel, but because they offer a forecast for players after taking into account their most recent performances—that, and because it is loads of fun to mull over and discuss a slice of possibility. All the while, however, other projections were just sitting there under our nose.

Over at Beyond the Box Score, Henry Druschel reviewed how well some of the major projection systems did last year. After looking at ZiPS, PECOTA, Steamer, and Marcel, he found that the Tom Tango’s Marcel the Monkey projection system outperformed many of the others. That’s interesting because Marcel is by far the simplest of them all.

Marcel projects performance based solely on what a player has done in the last three seasons, with the most recent seasons weighted more heavily, along with a slight age adjustment. For players with no major league experience, it simply pegs them at league average. Looking at Marcel’s output is essentially applying Occam’s razor to projection systems: The best answer is most often the simplest. Let’s take a look at what Marcel tells us about the 2016 Rockies.

★ ★ ★

First, the infielders:

Nick Hundley 418 0.262 0.306 0.415 0.721 12
Tom Murphy 220 0.263 0.329 0.439 0.768 8
Ben Paulsen 384 0.274 0.325 0.459 0.783 13
Mark Reynolds 459 0.222 0.303 0.399 0.703 18
DJ LeMahieu 564 0.283 0.335 0.381 0.715 7
Nolan Arenado 579 0.285 0.326 0.521 0.847 27
Jose Reyes 525 0.275 0.319 0.390 0.710 9

The first takeaway is from the top of the table. A season like that from Tom Murphy would be an enormous success. If we imagine the rate stats remaining the same, mentally add in about 80 plate appearances and a few home runs, the Rockies suddenly have a very good young catcher. And that doesn’t even include the bits of evidence that suggest Murphy is a quality receiver. In fact, the surface projections indicate that Murphy will be much better than incumbent starter Nick Hundley.

The other surprising highlight is from Ben Paulsen. The counter to my skepticism that Paulsen is a good major league baseball player is his decent major league baseball numbers over the past couple of years. The Marcel figures are more optimistic.

Other bright spots in the infield are found in the two 2015 All-Stars, Nolan Arenado and DJ LeMahieu. The sense evoked from looking at LeMahieu’s Marcel projection is "same ol', same ol'." That’s not a bad thing. In fact, it highlights how consistent LeMahieu has been over the past two seasons. Arenado’s projection looks entirely reasonable. As I’ve noted before, Arenado might not hit more than 40 home runs again in a season, but that doesn’t mean he can’t still improve at the plate. If the Marcel forecast is the baseline Arenado season, then he’s the star we think he is.

The infielders with the lowest projected OPS are also the lowest OPS among all position players. José Reyes is sure to get suspended, so we can adjust his PAs downward. If the Rockies bring him back after his suspension and the rate stats hold, then the team has the same flawed player they brought on last season. He’s still projected to have a better OPS than Reynolds, who is forecasted to be the Rockies’ worst position player this side of Brandon Barnes. Reynolds might hit some dingers, but his OBP looks like it will kill any value his slugging provides.

Let's turn to the outfield:

Corey Dickerson 365 0.293 0.342 0.516 0.858 15
Charlie Blackmon 606 0.285 0.337 0.446 0.783 17
Carlos González 532 0.267 0.325 0.509 0.834 29
Gerardo Parra 552 0.273 0.321 0.413 0.734 11

The outfielders are interesting, not least of all because one or more might not begin 2016 with the Rockies. According to these Marcel projections, the freshly signed Gerardo Parra will have the worst season of the four, and he’s the only one sure to not be going anywhere. The takeaway is that the Rockies should net good value for trading one of the other outfielders. We have a very good idea of the player Charlie Blackmon is, and the Marcel projections reflect that notion.

There’s indication that CarGo isn’t going to be traded during the offseason. He’s the most expensive and, notably, he’s not even projected to be the best outfielder. If Corey Dickerson posts the rate stats above with 150-200 more plate appearances, he’d end up the best of the bunch. In this simple look at simple projections, Marcel thinks Dickerson is the best of the Rockies’ four outfielders right now. I suspect a lot of people would agree with the monkey.

The projections for the starting pitching also has some notable nuggets:

Player IP ERA WHIP HR/9 BB/9 K/9
Jorge De La Rosa 153 4.12 1.340 1.0 3.5 7.3
Jon Gray 80 4.39 1.363 1.0 2.9 8.3
Jordan Lyles 96 4.50 1.365 0.8 3.1 6.8
Chad Bettis 105 4.54 1.410 1.0 3.2 7.5
Chris Rusin 122 4.80 1.475 1.2 3.0 6.5

This is the best guess of an Opening Day rotation, even though throughout the season five to seven other pitchers are likely to start a bunch of games for the Rockies. There’s nothing terribly surprising here. Jorge De La Rosa looks to be the rotation’s stalwart in what might be his final season with the Rockies. Jordan Lyles and Chad Bettis look like reasonable options given where the team is at the moment, while Chris Rusin’s projection fits the mold of a fifth starter on a non-contender.

Most interestingly, Marcel thinks that Jon Gray will be the Rockies best starter. Remember, this projection system only looks at Gray’s MLB performance from last season and adjusts for age. There need not be major math and adjustments to determine that Gray’s 5.53 ERA in 2015 was a fluke—even a monkey can do it! He’s better than that, and he might be the team’s best starter in 2016.

★ ★ ★

In the Beyond the Box Score article that spawned what you’re reading here, Druschel notes, "Baseball is extremely hard to predict, and using Marcel is going to be almost as good as using the more complicated systems in the vast majority of situations." That is a truly compelling conclusion. We can pull out a couple more conclusions.

First, Marcel is an argument against defaulting to "regression to the mean" as a form of player analysis. The monkey just points to the numbers without deeper investigation and draws fairly accurate conclusions That doesn’t mean that we shouldn’t look for caveats, though; it means that the caveats can’t completely stand in for everything else.

At the same time, Marcel rejects the impulse to look too far into a player’s past to determine how he will perform in the future. Most projection systems have long memories—Marcel’s is short. For instance, when perusing CarGo’s FanGraphs or Baseball Reference page, the eyes are drawn to 2010. As far as Marcel is concerned, 2010 never happened. It was too long ago to provide any insight into what he will do in 2016.

While ZiPS and PECOTA will still provide grist for our conversational mill, we can get a head start on them by looking at the simplest, and still useful, projections from Marcel the Monkey.