So I’ve gotten some flak for not having Marlins fan favorite Chris Coghlan on my NL Rookie of the Year ballot for the BBA, favoring the Pittsburgh Pirates’ Andrew McCutchen. In the ballot, I used WAR, calculated on my own using the guidelines as broken down by FanGraphs and by Tom Tango, who came up with the general WAR formula. In this space, I’ll try to break down how I achieved the conclusion.
First off, I’ll begin by defining how I planned on choosing the Rookie of the Year (as I will with all the other awards) and what I’ll do to evaluate this choice. Perhaps I was not clear enough in my ballot with regards to what the award meant to me. I am judging the Rookie of the Year as the most productive rookie over the course of the rookie’s playing time in the 2009 season. That’s the definition of the award in my opinion. I’m not awarding the award to “the rookie who is likely to be the best player over the course of his career,” (if you’d like to know my opinion on that, I think it’s Tommy Hanson of the Atlanta Braves) neither am I measuring “which rookie had the most historic stretch of offense.” I’m not even judging “the best rookie hitter” either, though that does get more to the point. I’m measuring production, context neutral so I can identify work done by the individual player rather than what he’s done in the context of his team (in other words, no quoting runs scored or RBI’s here).
So now that I’ve defined upon what I deem the award should be based, I need to define how I am going to evaluate it. For that, I chose WAR, Wins Above Replacement, as the statistic that would evaluate total production over the course of the season. WAR tries to put a player’s contributions in pitching, hitting, and defense in common terms so that they can be evaluated against each other. Furthermore, it places the player’s contributions in terms of runs and wins, which are ultimately the things that teams are interested in getting. After all, players don’t get paid to hit, pitch, or defend, but rather to save or add runs and wins to a team. WAR is, in my opinion, the best approximation we have to determining this player contribution to a team.
What follows is an careful breakdown of each WAR component between the two players, along with some explanations about what they mean. If you’d like a quick breakdown of how WAR works, I direct you to my other site, Intro to Sabermetrics 101 for an explanation within the glossary section. I also welcome fans of the Maniac who are interested in sabermetrics to check out the site for weekly content about the saber world and explanations on crucial saber-topics.
All right, let’s get knee deep into this explanation, shall we?
For offense, I’ll use wOBA from FanGraphs as my run estimator. It’s a linear weights formula, assigns run values to each event, and correctly values OBP as compared to slugging, unlike OPS. The reason why I use is because it’s extremely easy to convert wOBA into Runs Above Average, which is how we’ll compare offensive players.
Coghlan had a monstrous second half offensively, and unlike another well-respected writer who inexplicably casually dismissed Coghlan’s hitting with a weird explanation, I won’t do the same. I’ll say it right now: Chris Coghlan was the best rookie on offense this year in the National League. I’m not disputing any claims regarding his offense, and as a Marlins fan, it was a joy to watch Coghlan hit. He was great in the second half, and if you’d like to use the word “historic,” that may indeed apply. In all accounts, he was excellent, one of the best hitters in the second half in all of baseball.
Should the second half erase his first half, when he was below average? Absolutely not. Remember, the definition is production over the course of the season. Coghlan had wOBAs of .306 (May, 82 PA), .346 (June, 116 PA), and .293 (July, 83 PA). That tallies to a wOBA of .319, about .010 points below average. That’s why we’ll use his total numbers. Coghlan ended up with a wOBA of .372 according to FanGraphs; my calculations had him at .371. After park adjustment, both our numbers yielded some 20.7 runs above average.
Where I think Marlins fans are mistaken is in the evaluation of McCutchen’s offense. Truthfully, McCutchen was not very far behind. FanGraphs has his wOBA at .368, I have him at .366. This difference between the two agrees mostly with their OPS differences; Coghlan’s OPS of .850 isn’t that much higher than McCutchen’s .836. After park adjustment, FanGraphs has a value of 16.7 runs above average; I have a value of 14.7 runs.
What does this mean? Well, according to the calculations laid out here, Coghlan appears to be something along the lines of four to six runs better than McCutchen. My park adjustments are bad (my process isn’t down right) so I’d give more weight to FanGraphs’ numbers, but it does generally have Coghlan in the range of five runs better.
A lot of people have issues with defensive run totals, and I don’t blame them. In sample sizes we’re talking about here (almost a full season’s worth of chances for either player, surprisingly), UZR has error bars of +/- five (5) runs, which is very high uncertainty. With more data points, I’d be more certain, but at the moment I don’t have any other statistics to go by other than UZR. Still, UZR tells a not-so-pretty tale for Chris Coghlan, painting him as a -10 left fielder this season. McCutchen is at around -1 runs in center field this year.
People are quick to jump on these calculations and say that they’re bogus. Should we throw these out? No, but we should consider the error possibilities. There are some ways we can do that. Primarily, I would pick out more stats, but the only ones I have are the Fans Scouting Report data, and for the moment the data isn’t in the proper scale to convert to runs. However, if we buy that the fans are generally perceptive, we can use their voting tallies and estimate a qualitative difference.
McCutchen: 4.39 out of 5 as a center fielder (57 votes)
Coghlan: 3.18 out of 5 as a left fielder (16 votes)
What’s the difference between these numbers? Your guess is as good as mine, but taking it out the scale, you would say this is the difference between a good and great outfielder. Is five runs too much? When the season ends and TotalZone data from Retrosheet is available, along with a Fans Scouting Report that is in the 0-100 scale and can be converted approximately to runs, I can properly tackle this question once again.
The other option is to regress the data. For the purposes of this piece, I regressed both Coghlan’s and McCutchen’s UZR numbers with 100 games of average (0 runs) defense and reapplied it to their defensive game totals. With the regression, Coghlan rates at -6 runs, while McCutchen rates at -0.5 runs. With that, you’d still have a difference of 5.5 runs, which would make up the offensive total difference. Again, once we have more data, I’ll tackle the issue once again, but for now, this is all we have to work with.
Defense at each position is not worth the same amount. A +5 run shortstop is not the same as a +5 run first baseman, because the skill required to play shortstop far outweighs the skill required to play first base. This is why a positional adjustment is needed, to even out the defensive value in between positions.
Again, you can check out the Intro to Sabermetrics 101 site glossary for a more detailed look at this subject, along with links to the actual research involved. The research uses players who have played multiple positions and measures the differences between the metric’s numbers in their performance over multiple seasons. Here, I’ll simply pose a question: given two players were average defensively at their respective positions, how much more valuable is the average center fielder than the average left fielder? For all the defensive metrics, the difference measured through the research is around 10 runs over a full season. This difference is similar to the difference between a center fielder and a catcher, as an average catcher has been found to be worth around eight to 10 runs more than an average center fielder.
Qualitatively, this makes sense. Teams usually accept a lower value of offense from catchers because of the position they play, thus they are inherently adding a positional adjustment to the player’s worth due to their position. First baseman usually are your best offensive players, because first base requires more offense due to the ease of the position. Catchers are worth something like 25 runs more than first baseman in terms of defense due to position.
Which brings us to our question between Coghlan and McCutchen. Coghlan played left field, while McCutchen played center. Ignoring the quality of their play, Coghlan was docked 5.9 runs according to FanGraphs for his playing time in left field, while McCutchen was awarded 1.7 runs for playing center field as much as he did. The importance here is the difference; due to their playing time, the difference turned out to be around 7.5 runs, a bonus McCutchen receives for playing a more difficult position.
You don’t have to understand too much about replacement level to get why this adjustment is needed. The replacement level adjustment awards players for playing time, because otherwise the team would have to field a “replacement level” player worth approximately the league minimum in salary. Of course, theoretically a bench player replaces your starter, and not all bench players are at replacement level, but this does provide a convenient baseline for awarding playing time. Quite simply, the more you play, the more a replacement level player doesn’t have to play.
The replacement level adjustment is worth something like 20 runs per 600 or 650 PA. I use 600 PA, and so does FanGraphs. According to them, Coghlan accumulated 18.8 runs for being on the field, while McCutchen received 16.7 runs. Here the difference is about two runs to Coghlan for his 70 more PA.
As a whole, what do we get? If we say that the difference between Coghlan’s and McCutchen’s defense is the regressed 5.5 runs instead of the measured nine runs, we get the result that McCutchen outperformed Coghlan by around six runs. That difference would have been enough to leave Coghlan at a virtual dead heat with Hanson in my ballot for third place. If you considered the two players were even defensively for their positions, you’d have a case that Coghlan and McCutchen were essentially tied in terms of runs above replacement.
In other words, even if you don’t take into account Coghlan’s defense, which most of us will say was a bit below average this year, you’d still have the two players in a virtual tie for the Rookie of the Year. The difference between Coghlan’s playing time and slightly extra offensive production is entirely offset by the difference in difficulty between the two positions. And that would probably make sense as well; the two were very close in offensive production, with Coghlan in the lead, but you’d take McCutchen’s level of production from center field just as quickly as you’d take Coghlan’s from left, provided they both played their positions equally. Both the numbers and the fans report disagree with this notion, which definitely tips the scales toward McCutchen.
Are there flaws in the way I did this? Perhaps. In comparing these two players, playing time could make some difference, making the rate of replacement level adjustment important. However, such a difference would not make up a six run gap. More defensive statistics would help, and I promise that I’ll revisit this topic when these extra stats become available. I have enough confidence in the defensive adjustments that I don’t think they should be overwhelmingly off. And in terms of offense, I am very confident in the run differences, especially when tallied and park corrected by FanGraphs.
There’s my full explanation. Please couch any arguments in terms of the definitions initially provided; introducing other definitions of this award is not useful in discussing how I voted on the award. I welcome any critiques and discussion my Maniac readers may have on this apparently incendiary topic.