Are Writers Picking Better MVPs and Cy Youngs?
When Josh Hamilton and Joey Votto brought home MVP trophies this offseason, it marked the second consecutive year that the American and National League MVPs were also the league leaders in wins above replacement. For the Cy Young, three of the last four leaders in league WAR also took home the hardware.
It begs the question: Are the writers finally getting it? Has the statistics revolution finally gotten to the point where mainstream writers not only understand wOBA, xFIP, and UZR, but use them on their awards ballots?
To examine this question, I ran correlations between awards votes and WAR for both the MVP and the Cy Young for the past 10 seasons and looked for the positive trend that I was expecting to see.
So, are the voters using these numbers? In a word: nope. In fact, there has been a slightly negative trend for both awards over the past 10 years.
This result is somewhat counter to the narrative that surrounded Felix Hernandez‘s Cy Young victory this year, which was widely hailed as a sign of the revolution. This decade has seen major shifts in attitude toward baseball statistics on every level. Sites and blogs have popped up at every corner, there have been countless books printed, numerous statheads have garnered employment in front offices, and ESPN has incorporated things like WAR into their baseball coverage. Yet, the reality seems to be that, while some things are changing, award voters are not using advanced statistics any more today than they did 10 years ago.
It comes as no surprise that a subjective award voted on by mainstream baseball journalists does not perfectly correlate with an advanced statistic that many baseball fans have never heard of. Also, I am not suggesting that every ballot should list the WAR leaders in order, but I found it awfully surprising that these measures of overall player value do not align more today than in 2000.
But what about the last two seasons?
Well, the last two years were decent in terms of the correlations to WAR. However, there were still plenty of head scratching votes that dragged the correlations down. Tim Hudson finishing ahead of Josh Johnson for this year’s NL Cy Young, and the 7.2 WAR Ryan Zimmerman finishing 16th for this year’s NL MVP to name a couple examples.
Again, I am not suggesting that voters will ever solely use WAR to decide their awards votes, nor should they. However, given the talk that surrounded Felix’s victory despite his lack of victories, it is worth noting that the writers don’t appear to be moving towards WAR at all. They simply moved from one mainstream stat to another. The move from wins to ERA is progress, but it isn’t the bloodless revolution that has been described.
Curious to know what the correlations would be if B-R WAR was used. Wouldn’t expect much difference in the MVP vote, but the Cy Young might be markedly different.
Eventually, old people who refuse to change will die and then it will get better.
I almost thumbs upped death.
Yeah, we thought that was going to happen when the people who lived in the 60s were getting voted into office. Nothing’s changed. But that’s a different discussion.
>implying radical change is a good thing
In other words–
Meet the new boss…same as the old boss.
Not quite the same, but still something of an indication of advanced statistics going semi-main stream:
http://www.rasmussenreports.com/public_content/lifestyle/sports/july_2010/jeter_mauer_are_american_league_all_star_favorites
http://www.rasmussenreports.com/public_content/lifestyle/sports/july_2010/pujols_is_national_league_favorite_for_all_star_game
This is certainly not a sports specialist website, but the stats are still used. It’s also interesting to note that often the fans perceived most valuable players line up nicely with Fangraph’s All-Stars. I feel like we often feel as if there is a big difference between WAR and traditionally valued stats, but that’s not always the case. The main difference is that fans tend to value historical production as well as current.
Sorry, still can’t use pitching WAR seriously as a metric of determining Cy Young when all it looks at is FIP.
Its ok. Dont worry about it. IF we are lucky the world can move past this.
I’m not sure how useful this graph is. The correlation between voting and WAR for a given year is probably very closely correlated with the correlation between old-school stats and WAR for that year.
That is, if the best (by WAR) players happen to have great W-L records and RBI marks, then the voters will look really “good” on this graph whether they are using old stats or new stats.
I bet that the variation in that correlation (between WAR and old-school stats) from year to year is high enough that it overwhelms any slow move towards new-school stats by voters.
This.
It could be people are voting on something like the top 3 using defensible metrics, but use their other votes to vote for personal fave’s, gritty players, etc.
Also, my guess is you used regular correlation, when in fact I’m pretty sure you’d need to use non-parametric – spearman or kendall-tau – given the data.
So, if you run again looking:
1) at just the top 5 vote getters
and /or
2) using non-parametric regression
you might find something better.
Exactly… just going to say this. It’s all about the media vibe – advanced stats have made inroads into the mainstream press, the voters pick up on who is getting hype, they vote for said hyped player.
Then with the rest of the ballot they refer to RBIs/wins.
I’d love to see a top 5 votes-only MVP correlation.
“Yet, the reality seems to be that, while some things are changing, award voters are not using advanced statistics any more today than they did 10 years ago.”
It’s almost entirely the same people as 10 years ago, isn’t it? We’re only going to see different philosophies when the writers get replaced.
People can’t change philosophies/learn?
Generally, no.
People who are open to advanced methodologies will change methodologies, but guys who talk about “playing the game the right way” aren’t going to suddenly pick up advanced stats.
Coincidence. They are probably looking at numbers besides Wins and RBIs these days (thanks to the sabermetric movement,) but the WAR leaders have also been OPS, ERA, K, and Quality Start leaders… far more accessible statistics at this point. I doubt very many of these writers know what wOBA, xFIP, and UZR are. The evidence is in their content.
I don’t know how much better they’ve gotten in the last 10 years, but they have improved greatly since the 70s and 80s. I doubt we’d see such hilariously terrible votings like Boog Powell over Carl Yastzemski in 1970, Willie Hernandez’s MVP in 84, or, god forbid, the 1987 MVP voting in either league. (off the top of my head, NL players more deserving Andre Dawson for MVP in no particular order: Jack Clark,Tim Raines, Eric Davis, Dale Murphy, , Darryl Strawberry, Mike Schmidt, Pedro Guerrero ,Tony Gwynn, Will Clark…and that’s just off the top of my head).
I think you’re forgetting the piece d’ resistance, Rollie Fingers winning the Cy Young AND the MVP in 1981 as a reliever who threw only 78 innings. Yes, I get that 1981 was a weird year with a labor dispute, but there had to be 4 better MVP candidates on his own team.
There’s probably too much noise here to be useful. It doesn’t seem like votes are weighted by position, which is where it seems the biggest sea change has taken place (ie, third or second place votes becoming first place ones).
I was going to write what dfan said… this correlation isn’t very meaningful one way or another.
There’s a great little book called “Statistics as principled argument” by Abelson that I wish more scientists, let alone people who play with baseball stats, were aware of — basically, the point is you’re making an argument when doing stats and your argument will have strengths/weakness depending on the analysis one chooses.
I’d be interested to know the correlation between WAR and voting at the very top of the ballot (say, top 3 for each of MVP/Cy Young). I’d bet that correlates more closely. Downballot, you have all sorts of reasons why players get votes.
Also, defense is a huge (I would say too big) part of WAR, and it’s unreliable. So perhaps the voters rightly discount defense. I would think that if you looked at wOBA or wRC+ or some measure of purely offensive ability, it would correlate more closely.
Or if you included the variable of team record. Sure, Ryan Zimmerman is great, but should a guy on the Nats finish third in NL MVP voting? If you regress that variable, I would guess a closer correlation.
In short, I think WAR and the traditional stats that are frequently called meaningless (and the traditional writers frequently called ignorant) are actually pretty close to each other in voting once other factors have been considered. I mean, sure, we all know W-L and RBI aren’t the best metrics to evaluate players, but they still do tell us something.
Re: Your last point – Yes and no. Wins and RBI often do tell us something about PLAYING TIME (as do many contunting stats), which is itself a large component of WAR. However, I wouldn’t say they tell us very much about context-neutral performance.
It seems to me like a lot of those peaks can be accounted for by having a slam-dunk candidate. You don’t need to know anything about WAR to understand that Pedro and Unit were the best pitchers in their respective leagues in 2000 or that Bonds was the best hitter in the NL from 2001-4.
Also, as downballot MVP votes are typically cast for reasons for other than on-field performance, I would be interested to see the study re-done with a cutoff after, say, #3 or #5.
Jesse, I think this qualm raised by Adam is particularly worrisome. Letting these correlations be partially driven by down ballot votes glazes over the purpose of this analysis. We care about whether the top-tier votes are going to the correct people rather than whether some random writer through a 3rd place vote at his local guy because he knew it wouldn’t count. Although I haven’t seen the data I would be willing to be that each year there are at least a half-dozen players collecting pity votes.
The problem with Adam’s proposed solution is that we would only have 3-5 data points each year, which is really not enough to determine a correlation and have confidence in its meaning.
Although correlations are cool, intuitive, and pretty to display I don’t think it’s the appropriate metric here. I think the better way to analyze this is to compare the top 3-5 vote receivers each year and see how they ranked up with the top 3-5 WAR players. This method doesn’t have the flash that a correlation plot does and it’s harder to engineer some statistical test for evaluating the change (I guarantee you could find a non-parametric test that would do the trick though), it’s much more attuned to what you’re trying to evaluate.
Zimmerman didn’t deserve to finish 3rd in MVP voting, but neither did he deserve to finish 16th.
Part of that is the fact that he played for a last place team and put up very good – but not eye popping – numbers.
The other part may have been that even among supposedly enlightened saber-friendly voters, some of them still use Baseball Reference’s WAR. Which is inferior to FanGraphs WAR because of their respective defensive components.
That 2005 Bartolo Colon vote still makes me laugh. And this graph illustrates it perfectly.
Kinda like Pedroia winning the mvp in 2008
By the way, if you’re one of the new writers in the fold, welcome to the family – hope mom moved some of those boxes of shit to the side of the garage so you’ll have plenty of room to write lengthy screeds from her basement. =)
[1] Which WAR? fWAR or bWAR?
I thought a recent conversation indicated the writers were much more cloesly aligned with bWAR, which makes sense due it being based on runs allowed and not FIP.
Other research ashowed that run allowed in one seasons is a better predictor of ERA the following season than FIP was, but FIP was still a pretty good predictive measure.
[2] IMO, writers are using ERA and strikeouts … and generally have since the 80s. There are always exceptions, but if you’re strong in ERA and K’s you’ll likely take the award. Then again, if you’re strong in those 2 stats, you probably have strong stats all around.
I don’t think WAR is a generally accepted stats, and given the uncertainty around the relaibility of defensive measures and their weighting in WAR, as well as, using FIP or runs allowed, maybe WAR shouldn’t be the “determining stat”. I know that is heresy here, but there are discussions taking place regarding both comments/situations.
If writers look at ERA and K for picthers, and look at a variety of measures for batters, then I’m cool. We’ve seen the general trend since the 5-man rotations that pitcher wins carry less importance (unless they rack up a big disparity between the next guy). That’s what saberists have basically been after.
In any year, there are generally 1-3 guys that could “deserve” the award.
However, even saber-friendly writers, like Joe Posnanski, had CC and Price ahead of Lee and Bucholz (who I’ll point out just boasts wins and era, and not K’s and IP).
I meant to add “increased bullpen use” to the “5-man rotation” line.
Previously, when there were 4-man rotations and FAR more complete games, pitchers had more influence on pitcher wins than they do now. So, the league leader in wins was very often deserving of the award.
But the voting has changed along with those changes.
I think part of the issue with Ryan Zimmerman’s WAR is how much defense is incorporated. When such a large part of his WAR comes from defense, and writers ignoring a player’s defense when considering who should receive votes (excluding the DH, who voters will penalize for not playing defense at all), if that great defensive player has good but not great stats, he’ll finish very low in the MVP voting despite ranking much higher in WAR.
About 20% of Zimmerman’s 2010 WAR came from defense. That’s a good amount, but not exactly Brett Gardner territory.
If there’s a writer that ignores defense, chances are that they’re ignoring WAR too. I’m not talking about that group.
My concern is about the saber-friendly writers who do value defense and have a health respect for and understand of WAR.
If they look at WAR from FanGaphs, Zimmerman at least rates being part of the MVP discussion because of very good offense combined with excellent defense.
If they look at WAR from Baseball Reference, they’ll see the same very good offense, combined with… average defense. Which drops him and his WAR right out of the MVP discussion.
Yeah, the defensive component of Baseball Reference’s WAR says that he’s average defensively.
To beg the question does NOT mean the same thing as to raise the question. To raise the question means that a certain situation intuitively leads to the asking of a certain question, which is what you mean in the first line of your second paragraph. To beg the question is a specific and technical logical fallacy. How do you know which to use? Just never use “beg the question” and you’ll be OK.
If you fell in a ditch, I would laugh.
Is that a no, then?
It does beg the question.
At least somebody gets it.
Well stated. We ask the writers to learn more about proper use of statistics, so why not ask the statisticians to learn more about proper writing?
This recent abuse of “begging the question” has become so common that many simply believe it is correct, refuse to accept otherwise, and mock anyone who dares correct them. Ironically, these are many of the same people who are subjected to similar treatment when criticising use of Win totals and RBI’s.
In a further irony, the other most-abused concept is irony. No, wait — that last one is just a coincidence, not irony.