In the wake of the Astros' minor league successes and major league failures, Astros fans might be missing out on one of the most interesting baseball-related debates of the last decade. The race for the AL MVP award has lately grabbed national attention because two candidates--Detroit's Miguel Cabrera and Los Angeles' Mike Trout--have separated themselves from the pack but not from each other. Further complicating the matter is that there are strong arguments for both candidates. Cabrera's case is based on traditional counting stats favored by self-proclaimed baseball purisits, and Trout's on advanced metrics championed by sabermetricians. The battle between good/evil, red/white, up/down, vanilla/chocolate has been going on for millennium, and is currently being fought on the baseball statistical battlefield.
As an experiment, clack and I have decided to debate the issue for your enjoyment. In one corner, clack will tout Trout's candidacy because of his superior advanced metrics. In the other, I have chosen to champion Cabrera's because of traditional measures. Let the fun begin.
Sent: Tuesday, October 02, 2012 4:18 PM
Obviously, both Cabrera and Trout are legitimate candidates to win the MVP award and both deserve it for varying reasons. But there can be only one winner, and I think Cabrera's the man to claim the prize. Before diving into my reasons, I'll allay your fears. I will not be arguing that Trout is a 20 year old rookie and will have plenty of other chances to win the MVP award in his career. I've heard that argument on the radio and it's ridiculous. Rather, I think Cabrera's case is slightly better on it's own merit.
At the time of this writing, it's no certain thing that Cabrera will win the Triple Crown. He's darn close...he leads Trout in Batting Average .329 to .325 on Tuesday with two games left to play. But there's a decent chance he will play little or not at all in the interest of saving him for the playoffs. Likewise, Trout has two games left and the Angels have already been eliminated from the playoffs. Trout will probably play because his manager will want to give him a chance to pad his stats for the possible MVP award. Cabrera leads in Home Runs by one over Josh Hamilton and two over Edwin Encarnacion (I can't believe I just typed that about Encarnacion. Can you believe he turned his career around like he did? A topic for another day.). Cabrera leads the RBI category by 10. So there's a better-than-small chance Cabrera won't win the Triple Crown even though he's on pace for it.
Assuming Cabrera does win the Triple Crown, he will be only the sixteenth hitter in MLB history to have done so. The last time a hitter won the Triple Crown was when Carl Yastrzemski famously pulled it off in 1967. Winning the Triple Crown is a historic accomplishment and should be recognized as such with the MVP award. Heck, even coming within two days of winning the Triple Crown can probably only be claimed by a handful of hitters other than those fifteen fellows.
Conversely, Mike Trout currently owns a WAR (Wins Above Replacement) of 10.2, which is tremendous. But out of 14,455 player seasons listed by Fangraphs, hitters have achieved a 10.2 WAR or greater seventy-five times since 1878. While not trying to take away from Trout's accomplishment, the rarity of what Miguel Cabrera has been able to do this season with raw numbers earns him the MVP vote in my book.
What are your thoughts on the subject?
Sent: Tuesday, October 02, 2012 5:44 PM
Chris, obviously everyone can agree that Cabrera and Trout are both enjoying tremendous seasons. The concept of "MVP" can mean different things to different people. Some people will only support a MVP who is leading his team into the playoffs. If that is a criterion, then Trout won’t qualify, given that the Angels fell short of the playoffs. But I prefer a MVP criterion based upon the best player, the guy who helped produce the most wins for his team. And this depends on his own production, and not the production or quality of his teammates.
WAR is a good tool for measuring a player’s overall contribution to his team’s wins. WAR includes defense, offense, and base running. Of those three components, the triple crown includes only a portion of the player’s offense (HRs and batting average), and totally omits defense and base running. In addition, the triple crown includes a category, RBIs, that is dependent on the ability of the player’s teammates to get on base and the placement of the player in the batting order. RBIs are not all that valuable in telling us about value.
When all facets of the player’s value are taken into account, Trout is much more valuable than Cabrera, regardless whether you rely upon Baseball-Reference’s or Fangraph’s WAR measure. Trout’s WAR is 57% higher than Cabrera based on B-Ref’s WAR, and 46% higher based on Fangraphs’ WAR. People can argue which WAR calculation is exactly right. But the margin of Trout’s superiority is large, no matter how you calculate WAR.
I understand that the triple crown is a rare event---maybe coincidence is a better term than "event." Baseball has many rare events, like hitting for the cycle or pitching a perfect game. But hitting for the cycle is rarely mentioned as a hitting attribute in MVP debates. Pitching a perfect game doesn’t earn a pitcher the Cy Young Award. Otherwise, Matt Cain’s perfect game should earn him the CYA. Yet he is not among the favorites to win the the NL Cy Young Award. The triple crown is that rare coincidence when a hitter leads three specific batting categories. But it doesn’t tell us whether the player is the best overall baseball player or even the best offensive player. That’s because the three categories are somewhat narrow indicators of offensive performance. Even setting aside defense, Trout has been a better offensive player than Cabrera, based on metrics like wRC+ and wOBA.
Sent: Tuesday, October 02, 2012 7:55 PM
Clarence, thanks for replying. Trout's case is solid, no doubt. I have some issues about the vagueness associated with the calculation and application of WAR, but you brought up a couple things that hit the center of the argument about the MVP award itself and the value of traditional stats that I'd rather discuss first.
Like you indicated, a central point in the debate over the MVP award is whether the award should go to the best baseball player or the most valuable. While some say there is no difference, I would argue that the difference is a case of quantitative versus qualitative. The Silver Slugger exists to award the best pure hitter at a position. Golden Gloves, flawed as they are, exist to reward defensive goodness. The Hank Aaron award goes to the top hitter in each league. All of those can be argued to be quantitative, as they are based on old-school and/or new-school statistics. But for hitters, the MVP award is the only one where qualitative value is ever debated. That is, how much impact did this player have on his team or on the league standings?
I don't believe that the qualitative argument can be ignored. We live in a real world where our actions matter and our performances have consequences, not in a world of pure mathematics. In September, when both the Tigers and the Angels were vying for a playoff spot, Trout batted a paltry (for him) .257/.380/.455. Meaning, when the going got tough, Trout got...less tough. Meanwhile, Cabrera carried the Tigers on his back all the way into the playoffs by destroying opponents in September to the tune of .308/.378/.654. Doubtless, the MVP needs to be applied to entire body of work for the season, but from a subjective "value" standpoint, Cabrera was an MVP when his team needed him the most, and Trout was...Jed Lowrie.
There, I mentioned the Astros. Our obligation is complete and we can proceed.
I contend that RBI can be used as easily as wOBA or wRC+ to judge a player's past performance. Stats, whether they are simple counting stats or based on an obscure formula rooted in differential equations, quantum mechanics, or copernican heliocentrism, are only useful when applied in a meaningful context. Mike Trout's 10.2 WAR is impressive individually, but his teammates will be watching from the couch in October. Meanwhile, 19% of the Tigers' runs all season are a direct result of Cabrera's RBI's, whereas "only" 11% of the Angels' runs were knocked in by Trout. Cabrera drove in 7% more of his team's runs than Trout did. Not fair, you should say, because of their differences in the lineup order. True, except that Cabrera scored 15% of his team's runs compared to Trout's 17%, for a difference of only 2% in favor of Trout. In my opinion, Trout should have scored a higher percentage of his team's runs because Albert Pujols, Kendrys Morales, and Mark Trumbo hit behind him all season.
Furthermore, sabermetricians have been trying for awhile to define "clutch". I wish them success, but there's a quick and dirty way to look at whether or not players perform well in the highest-impact situations. With Runners in Scoring position this season, Cabrera hit .353/.422/.578. Trout was right behind him with .330/.398/.569. But amping up the pressure a bit, when runners were in scoring positions with two outs, Cabrera hit .420/.491/.720 and Trout hit .286/.435/.347 in the same number of At Bats. Yes, it's a small sample size, so no long-term statistical or predictive meaning can be derived, but what can be said for certainty was that in those high-pressure situations in 2012, Cabrera was definitively the stronger offensive player. And I think that matters.
My point is, it's easy to use traditional stats in a context that shows a player's value on his team. For what it's worth, I don't agree with the classification of a Triple Crown as a coincidence, but that topic isn't nearly as interesting as the one I covered in far too many words already above. But even without winning the Triple Crown, Cabrera's MVP case can be built by showing that his traditional stats, when applied in context of the value he added, provided greater impact than Trout's.
But you are welcome to convince me otherwise!
Sent: Tuesday, October 02, 2012 10:44 PM
Chris, I agree that a ballplayer’s actions matter. But he is only responsible for what is within his control. I understand the narrative that Cabrera carried his team to a playoff spot. But do you realize that Mike Trout’s team has more wins than Cabrera’s Tigers? Surely Trout’s 10.2 WAR had something to do with the Angel’s record. The Angels played in a tougher division than the Tigers, and won more games than the Tigers. That should be counted in Trout’s favor. He has no control over the strength of the division, but he was a better offensive player than Cabrera, while also playing against tougher competition.
I think it’s hard to argue that Trout didn’t score enough runs. He led all major league players in runs scored (129). And he did it in only 137 games, because he began the season in the minors. The remaining top 5 players in runs scored played 10 – 20 games more than Trout---yet Trout scored 20 – 23 runs more than each of them. In fact, Trout’s 10.2 WAR is even more impressive when you consider that he accumulated it in 23 fewer games than Ryan Braun, who posted the next closest WAR total (8.1).
If you want to give Cabrera credit for batting well during the Tiger’s stretch drive, that’s a fair point. But how much extra credit can you give him? Cabrera’s hot performance in September is already incorporated into his season stats. And, like you say, the MVP should be based on the entire body of performance during the season. A win in May or June counts the same as a win in September. Based on wOBA and wRC+, Trout was the best offensive player over the course of the season. And, even if Trout and Cabrera were only roughly equivalent offensive players, Trout’s defense breaks the tie. Stopping the opposing team from scoring runs is important too, and Trout’s defense helped prevent runs from scoring on a day in and day out basis.
I’ll stay away from the philosophical questions about clutch hitting. But I will disagree with your position that Cabrera holds a big advantage in hitting during the highest impact situations. The leverage or impact of a situation is determined by more factors than just the number of runners on base. The score and inning, as well as the base-out state, determine the leverage of the situation. Neither Trout or Cabrera increase their offense (as measured by OPS) in high leverage situations. But both hit well in high leverage situations---Cabrera with a .971 OPS and Trout with a .917 OPS. Given the sample size, I don’t think the difference in their high leverage offense is significant.
Another way to measure the impact of a player’s offense on a game’s outcome is Win Probability Added (WPA), which is based on the extent that the hitter’s offensive actions increased or decreased the team’s probability of winning each game. Mike Trout is No. 1 in WPA. He was the best in the major leagues at improving his team’s chances at winning its games, as measured by WPA. Cabrera is ranked No. 12 in WPA. WPA is descriptive rather than predictive. I wouldn’t use WPA to measure talent or ability, but it is relevant in describing a player’s impact on changing the course of previous games.
As demonstrated in this response, advanced statistics have something to contribute to the qualitative issue you raised. And my opinion remains unchanged that Mike Trout should be MVP.
Sent: Tuesday, October 03, 2012 8:17 AM
You zapped one of my points pretty fairly about the strength of the divisions. But today, Miguel Cabrera is one day closer to the Triple Crown, as he raised his average on Tuesday night by going 2-3. The Triple Crown is an accomplishment that demands historical recognition. Of the fifteen Triple Crown winners, only four were not awarded the MVP, and even that sounds like a travesty when you see who earned the crown in those years (Ted Williams 1947 & 1942, Lou Gehrig 1934, Chuck Klein 1933).
Let me say that I am definitely a stats guy, and I love the sabermetric concepts you bring up, like WAR, WPA, and Fangraphs' Clutch stat. But the beef I have with such stats is the same problem I have with all sabermetrics--there is no consistency of method between experts, and oftentimes the inputs are subjective instead of objective. Take WPA for example. It's calculation is based on Win Expectency (WE). WE in turn is based on computer simulations that use databases of past material to try to predict who is most likely to win a baseball game. The problem with simulations like that is they often don't take into account small variations that affected the past data or can affect the game at hand, such as weather effects, team/player morale, team/player health, manager style, and just plain dumb luck (ever seen Jurassic Park? It's the butterfly effect!). All of these things affect every game, and it's a bold assumption to say that they just "come out in the wash" through multiple simulation iterations. In other words, WE tries to be a predictive engine for a system that has a zillion minute complexities that can't be quantified. And WE is the foundation of WPA. Further, even Tom Tango, one of the leading sabermetric experts, states that WPA is useless in identifying individual player talent:
"WPA is not a way to evaluate the talent of a player." It's entirely context-specific. In other words, it's great for game analysis, but less great for player analysis. It's like the percentages they show on World Series of Poker broadcasts: it just tells you the percentages of victory during that hand, not the talent levels of the poker players involved."
The Clutch stat in turn is based on WPA, which is based on WE, which as I've shown is based on an imperfect model. No slight intended to its creators...it's very good, but a perfect model only exists in an abstract paradigm state of platonic realism. (Believe it or not, I used that same phrase in an engineering report at work. It may have been my finest moment.) So I'm confounded by how Clutch can be used to measure a player's ability to hit in clutch situations when experts state that the variables it is calculated from can not be used to evaluate players. It bothers me how sabermetrics experts put their calculations forward as definitive quantitative measurements when they truly aren't. Their work is very good and very illustrative, but like all other stats they must be put in context to mean anything. And one of the contexts they must be placed in is one of uncertain inputs.
WAR suffers from a similar issue, and more. Fangraphs creates it from the sum of wRAA, UBR, and UZR (all of which are stats whose value are disputed in favor of some other stat by some other website), then applied with a vague positional adjustment that I'm sure is based on back calculations of past data but suffer from the same variations described above. Then the whole mess is compared to a "replacement" player, who is also back-calculated based on past data, but is a hypothetical somebody who doesn't exist, and experts don't even agree on how to define the stats of the imaginary replacement player. Yikes! WAR is exceptionally useful and probably can be used with broad strokes to paint a picture of who has been good and who has been not good. But it suffers from too many flaws and subjective inputs to be used as a definitive measurement of who has been more valuable than who. The "and more" I alluded to is the disagreement among statisticians about how Wins above Replacement is calculated. Fangraphs have their WAR, Baseball Reference has their oWAR and dWAR, Baseball Prospectus has their WARP, WARP1, WARP2, and WARP3. Eventually I think you can time travel if you use WARP10 to slingshot around the sun. This is my biggest beef with sabermetrics. Why haven't the statisticians recognized that their bickering and infighting are the main reasons why mainstream baseball folks have not accepted these concepts? I brought the topic up once at Baseball Prospectus and was greeted with a chorus of, "Because we're right and they're all wrong!" Meanwhile, the flaws I pointed out above go unaddressed. As I stated in my last email, context is everything with stats and I have trouble finding a context where one of the various WAR stats can be used to measure the impact that either Cabrera or Trout had on their team and the league.
On the way in this morning, Mike Tirico was on the radio. After proving that he actually does know what WAR is (he mentioned it as a good measurement of Offense, Defense, and Baserunning), he said, "... but WAR is only good for Third Place. And that's where the Angels are." To sum up, I still think Cabrera has the stronger claim because of the historical achievement of the Triple Crown, the application of traditional stats to show that he had a greater impact on his team's playoff run, and because of the contextual problems with advanced metrics that do not consider all the real-world variables. I've enjoyed the discussion and think this is a good format to present once in a while on the site. And now, you may have the last word!
Sent: Tuesday, October 02, 2012 10:44 PM
First, I’ll comment on your claim that it was a travesty when past triple crown winners were not awarded the MVP. The last time that happened was in 1947, and Joe DiMaggio won the MVP over Ted Williams. DiMaggio, like Trout, played a premium defensive position, and was regarded as a graceful fielder. Writers didn’t have advanced statistics in that day and age. But I have to believe that writers realized that a player’s worth was more than just offense. Williams never really had a good defensive reputation, and he committed the second most errors among LFers that year at a weak defensive position. DiMaggio committed the least errors that year among CFers, a strong defensive position. Are errors a good way to evaluate defense? No. But, other than the "eye test," it was all that was available in those days. Perhaps advanced metrics, like WAR, would have helped Williams win the MVP, if those statistics had been available. Maybe Williams deserved the MVP, but the fact that DiMaggio won the MVP shows that the triple crown does not guarantee the MVP---even without advanced statistics.
Second, don’t get hung up on what you like or don’t like about advanced metrics. Defense and base running have always been regarded as valuable skills---these are virtues held in esteem by traditional baseball people. Why did Roberto Clemente win the NL MVP in 1966, even though he was only the 6th or 7th best offensive player? He was a great defensive outfielder and a terrific base runner. Frankly, Mike Trout has a much better case for a MVP than Joe DiMaggio in 1947. The difference between Trout’s and Cabrera’s offense is not large, even if you use traditional batting stats like batting average. Yet even the eye test would tell us that Trout is a much better defensive player and base runner than Cabrera, who has a bad reputation as both a fielder and base runner. Advanced statistics aren’t required in order to realize that Trout has provided more value in more ways to his team. Mike Trout has been the more complete ballplayer.
Third, I’ll address your general concerns about WAR. Although I don’t agree with all of your criticisms of the advanced statistics, this probably isn’t the right place to get mired in arguing the details of the methodologies. No analytic methodologies are perfect. But if you throw the baby out with the bath water, so to speak, you have no information. I think we agree that WAR has, at least, some usefulness. I just don’t see how making judgements with no information is better than using information which is useful, though imperfect. The alternative is to rely on subjective opinions of defense and base running, for example. This seems much more susceptible to error than using the quantified fielding and base running components of WAR. And, yes, WAR can be calculated in different ways. That’s because it’s a framework for evaluation and not a doctrine. But where is the WAR calculation that favors Cabrera over Trout? I haven’t seen it. The two main sources of differing WAR calculations are Baseball Reference and Fangraphs---yet, in this case, both WAR calculations show decisively that Trout is more valuable than Cabrera. Even if you think the WAR calculations are fuzzy math, it is implausible that any ambiguity in the results is sufficient to overcome Trout’s margin (46% – 57%) over Cabrera.
Fourth, let me clear up some questions you have raised about Win Probability Added. I used this statistic to test your claim that Cabrera hit better than Trout when it matters. If we want to evaluate how the timing of hits contributed to winning games, the family of WPA stats (including WPA clutch) is the best evaluation method. Again, not perfect, but better than the alternative of just speculating. Although WPA doesn’t measure a player’s overall hitting skill, it is useful in telling us what happened, when it happened, and the impact of the timing of the events on the game. Yes, Tango and other saber analysts wouldn’t use it to evaluate a player’s talent---but most saber analysts also wouldn’t view the runner in scoring position batting stats that you presented as the appropriate measure of a player’s talent. That’s because the very concept of "clutch hitting," to a great extent, is not considered a repeatable skill. So, if you want to reject WPA on those grounds, then you should also reject your argument that Cabrera was a better clutch hitter than Trout.
My conclusion: you don’t have to use advanced statistics to find that Mike Trout should be AL MVP, but advanced statistics provides quantifiable information which supports his worthiness of the MVP award.