Tulo’s WPA And The MVP Award
One of the objections I heard today to my suggestion that Troy Tulowitzki deserves MVP consideration is that there’s a massive gap between him and Joey Votto in WPA. And it’s true, there is. Votto has the highest WPA in baseball, coming in at +6.47, while Tulowitzki’s +1.38 ranks him just 46th among position players, never mind all the pitchers who also rank ahead of him.
One of the main reasons why Tulowitzki’s WPA is so low is that a lot of his offense has come in games that had already been decided. In low leverage at-bats, he is hitting .361/.427/.649, but just .261/.357/.472 in high leverage situations. This gap is pretty large, and the opposite of Votto’s, who has hit better in high leverage situations this year. That makes a difference to a lot of people, and I understand why.
However, let me use an example from Saturday to show why you should be careful using context-dependent stats like WPA to determine how valuable a specific player was.
9/18, Colorado vs LA
Top 1st, 2 outs, runner on 1st, no score.
Tulowitzki hits a home run, giving the Rockies a 2-0 lead. Since their chances of winning jumped from 48 percent to 68 percent, Tulowitzki gets 0.20 added to his WPA total for the season.
Top 3rd, 2 outs, no one on, 2-0 Rockies.
Tulowitzki doubles to center. Because the Rockies were already likely to win the game, this double was worth a little less than a normal double, and he is credited with just 0.02 WPA. He would eventually score on Melvin Mora’s single, and the Rockies would take a 3-0 lead.
Top 5th, 1 out, runner on 1st, 3-0 Rockies.
Tulowitzki hits a home run, giving the Rockies a 5-0 lead. Because the Rockies already had an advantage, these runs were worth less than the earlier runs, so Tulowitzki only got 0.08 WPA for pushing their win expectancy from 86 percent to 94 percent.
I’d imagine you can see what happened here. Tulowitzki received minimal credit for his 5th inning home run because the Rockies already stood a really good chance of winning the game, and yet they had those good odds of winning because of what Tulowitzki did in his prior two at-bats. He essentially created his own low-leverage situation by single-handedly beating the Dodgers on Saturday.
Do you really want to say that Tulowitzki’s two-run homer in the 5th was less than half as valuable as his two run homer in the first? In terms of the story of the game, those runs didn’t have the same impact, but were they really less valuable? The fact is that the WPA of that HR is decided by how Tulowitzki performed in his first two at-bats of the game – should we hold a player’s good performances in less standing if they come after his previous good performances in the same game? Or, alternately, should we punish him for performing well on the same day that Jhoulis Chacin was throwing a shutout? If his first inning home run had tied the game, rather than breaking a tie, and his second home run had broken open a closer affair, the WPA story would be a lot different, even though the only variable we’re changing is the performance of the Rockies starting pitcher.
For these reasons, I lean towards context neutral performance. For me, WPA is like overall team performance – use it as a tie-breaker in a situation where it is otherwise hard to distinguish between two candidates, but don’t lean too heavily upon it. I’ll give Votto a bit of a boost for just how awesome he’s been in high pressure situations this year, but I’m not going to ignore how great Tulowitzki has been in situations where the game wasn’t close just because he helped make it not close in the first place.
WPA is great at telling the story of how a game unfolded. It is less great at evaluating how much a player helped his team win through his own actions. This doesn’t make it a bad stat, but it was designed for a specific purpose, and it is best used in that purpose. Trying to take it and turn it into a value stat for player performance is asking it to do something it was never intended for in the first place.
Dave is the Managing Editor of FanGraphs.
I take it you aren’t a fan of something like http://www.draysbay.com/2010/8/2/1600933/clutch-war-measuring-value-for then?
And that’s why we have WPA/LI…of which Votto doubles Tulo’s score in…
Yep. Votto’s been the better hitter, there’s no doubt about that. The question is simply by how much, and whether the defense and positional scarcity make up for that gap. So whether you use a +2 win offensive advantage for Votto (context neutral), a +3 win advantage (WPA/LI), or a +5 win advantage (WPA) matters.
Right, but I’m still trying to find a single statistic that favors Tulo. I’m totally confused – in your two articles, I see bunch of “what ifs” but every single number so far has favored Votto – what defense can you possibly give for Tulo? It seems the only defense you’ve given is “the numbers can be misleading” – but that’s the argument from ignorance fallacy discussed in the Gardenhire article today.
“till trying to find a single statistic that favors Tulo.”
How about all of his rates? For a SS all of Tulo’s offensive rates are better than Votto at 1B. If Tulo had the same number of PAs as Votto this article would not exist and they would be carving Tulo’s name into the MVP trophy.
Of course he has to get penalized for having less cumulative numbers, but you don’t to act like Tulo’s rates don’t considerably beat Votto’s from a pure rate perspective.
.420 wOBA from SS
.437 wOBA from 1B
Defense aside, you take the .420 from SS all day, every day. There, that is a stat for you.
I don’t actually mean “all of his rates” clearly Votto’s OBP beats Tulo’s even with the positional consideration, but the point remains, especially with wOBA, which I think we can agree is the most important rate stat we have.
@ Andy — UZR favors Tulo, as does the positional scarcity stat. Exactly how hard did you look to find one that favors Tulo?
I’m sure he means overall value chuck.
So value per game or value for the season?
Value per the season – that should be the only thing that matters in the MVP race.
Wrong, HRs and RBIs are the only things that matters. Defense doesn’t matter to MVP voters, please just give me the list of 1B in the national league so I can check off the little boxes.
k thks.
Oh, and the players in question need to play for a team that makes the post season, this is obvious.
So Tulo is far enough ahead where WPA shouldn’t be a tiebreaker? Didn’t realize Tulo was that far ahead of the field.
Just curious your argument on the value of an intentional walk vs a normal walk the other day when talking about Cabrera, that have any context factored in?
Frankly, I don’t see what’s wrong with counting Tulo’s 2nd homer less than his first in a single game, it doesn’t help his team as much as the counting stats would suggest, and a game like this is among the minority rather than the norm
The WAR difference with the bat is roughly 2 wins. The difference using WPA/LI is 3 wins, using RE24 is 3.5 wins, and WPA is 4.5 wins. The difference between the two has been about more than just Tulo tacking on runs in meaningless innings. He just hasn’t hit as well in situations where his at-bats had significant value in terms of scoring runs. In assessing value, I think that WPA may exaggerate the difference here, but some adjustment for the value of their hits is probably due. WPA may exaggerate the difference, but I would argue that those other differences are worth more than just being a tiebreaker.
Wait – when did the MVP award become the Most Clutch award? I’m not saying Tulo should be the MVP, I just wasn’t aware WPA quite this important in finding the best player this year. Interesting.
While it’s not called the “Most Clutch Player” it IS the “Most Valuable Player.” To say that a player like Adam Dunn (who always seems to hit solo shots when his team is already winning or losing by 4+ runs) is more *valuable* than a guy like Votto, who comes up big in clutch situations, is just plain wrong. Obviously Tulo>Dunn, but for my money “Valuable” should be the guy that helps his team win ball games.
If the game was tied in the bottom of the 9th, a Double gets less WPA than a single that knocks him in, but clearly the double did more to win the game.
If the double was just a single, then a follow up single wouldn’t have scored him.
Dave Cameron for president. That is all.
You can’t use WPA for comparing players for MVP. That’s just retarded. Unless you want to compare how many back pages they got in a season, it’s meaningless.
And by that I mean, “I agree with Dave.” Don’t tell me Votto deserves the MVP because of WPA. Plenty of other reasons to consider Votto for the award, however.
As for Tulowitski, I know he singlehandedly destroyed my fantasy team when I needed a win to make the playoffs. That means more to me than his WPA!
What? I think this post is the only retarded thing here.
How about “Dave Cameron for president.”?
I follow your examples on the follies of WPA. But it seams to me that WPA is very much supposed to penalize the value of the second home run, despite the fact that Tulo effectively hurt his own numbers with exeptional play. I – again this is my understanding of WPA – think that if Chacin had given up a few runs so that Tulo’s second home run put the Rockies in the lead that WPA should give him a whole lot more credit. So I agree that WPA should likely only be used as a tie breaker, but I believe this is already how many commentors were using it in your last thread.
The way I see it the entire purpose of this article is to remind people why WPA is a poor tool to use when evaluating MVP. I think you were successful in showing some of the potential pitfals, but I don’t think you ever made a good case for why you think it should carry so little weight in MVP evaluations.
Please let me know if you think that I have missed the point.
I reviewed the comments of the other post and I rescind my assertion that the commentators we using WPA as a tie breaker.
In that case it seems to me that your conclusion seems reasonable: That WPA is mostly telling a story and is not a great evaluator of single season performance.
Full Disclosure: I love that you put out a lot of content and start a lot of debates, but I will be honest when I say that your writing tends make me very defensive.
I think the case has been clearly made all over the place. WPA is a “story” stat. It’s an amazing way to look at a game and see what happened and who was involved.
Say a player hits two two run home runs with the team up by 4 runs. So now they’re up by 8. And then the opposing team comes back and scores 7 runs and loses. Were those 4 runs any less valuable than the first four to that win? Of course not.
This is not a criticism of WPA. It’s just a criticism of those who misuse it.
WPA/LI is a completely different story. That’s a great stat for MVP discussions.
Exactly. You never stop playing. The game isn’t over until the 27th out has been recorded. You can’t penalize a guy for beating the other team more.
WPA, and all “clutch” stats, are really dumb.
I don’t know how someone can make that statement. I’ll agree that WPA and clutch stats don’t necessarily mean one player is better than another. I also don’t think that being “clutch” is something that is necessarily repeatable.
However, WPA and clutch stats do a great job of telling a story. The story that WPA told this year is that Joey Votto is the hands down NL MVP for the 2010 season and Tulo doesn’t belong in the discussion. I cannot think of a better stat for choosing an MVP. It describes who was the most “clutch” for the year. We don’t need to use it to predict who will be most “clutch” in the future.
I’ve lost a lot of respect for Dave with these last couple of Tulo articles.
I am sure he is devastated to hear that
Ranking Strasburg as having the second highest trade value in BASEBALL didnt do it for you???
DC is a moron.
*groan*
WPA and “clutch” stats don’t show who is more valuable than who. They are completely reliant on context and don’t actually relate to which player had the best season. The reason I think whoever had the best season should always win the MVP is because no matter where he plays or what the rest of his team does, he will have the most value. Since “clutch” isn’t a repeatable skill, he is also the most reliable in any situation. It doesn’t make sense to say that the guy who did the best in high leverage situations is the most valuable since it is reliant on the player’s surroundings and it isn’t repeatable.
But yeah, Votto should win MVP. Zimmerman would be my number 2, with Pujols third.
I agree with you that WPA is misleading and probably shouldn’t be used, but I think this post doesn’t tell us anything. You can’t use this anecdotal example of a time Tulo’s later performance was made irrelevant by his own previous performance in the same game to show that this is what he’s been doing all year.
What about 5/23, when Tulo hits a solo HR to make it 10-4 in the 6th. This is a pretty worthless HR (.025 WPA) after the game is in hand and Tulo hadn’t done much to help contribute to the Rockies having the game in hand. He had a single but also struck out with two men on base.
On 8/4, Tulo hit a solo HR in the 6th to make it 6-1. This was worth .030 WPA, and Tulo didn’t contribute to his team scoring the first 5 runs that have put the game in hand.
Point being, better examples are present to show how misleading WPA can be. What if Chase Utley hits a HR against Josh Johnson in the top of the first inning, and Halladay goes on to outduel him 1-0? That HR isn’t worth much in WPA but it was obviously crucial, clutch, and whatever else you want to call it. In fact, Tulo is a good player to use to show the worthlessness of WPA. There are a few situations where he knocks in runs early in the game that end up being the difference in close games the Rockies won.
I agree with your point, but not how you demonstrated it.
WPA is a zero-sum stat. Utley’s HR didn’t increase the team’s chances of winning nearly as much as Halladay’s pitching did.
While I don’t think WPA should be used as a primary factor in determining the winner of the MVP, I don’t see what’s wrong with factoring in. I know all about the studies that have been done about clutch hitting and how it supposedly doesn’t exist, so I understand that WPA is not a good tool for evaluating a player’s performance. However, the way I interpret it, the MVP award isn’t necessarily meant for the player who had the best season. It’s unfortunate for Tulowitzki that so many of his home runs have come in low leverage situations, but that doesn’t change the fact that those low leverage home runs didn’t help the Rockies as much as Votto’s high leverage home runs helped the Reds.
I realize that WPA is not a great evaluative stat, but it does help show how much a player helped his team win games during the course of the season. Obviously it shouldn’t be the only stat used and having a low WPA shouldn’t kill a player’s chances. However, it does describe how a player impacted his team’s chances of winning, so it isn’t useless in the MVP conversation.
I don’t understand how WPA is misleading in this case. The whole point of WPA is to reward a players timing. In this case, Tulo’s homer was essentially worthless because it didn’t improve the Rockies odds of winning much. If he were to strike out, it wouldn’t make a big difference either way.
If you literally think that win probability added is a good value stat, than there is no problem with WPA.
VEP,
I agree with you completely that Dave’s example doesn’t make his point.
WPA can be screwy. On May 25th, Tulo hit a solo HR in the 3rd Inning to put the Rockies up 3-0. That was worth 0.77 WPA and 1.00 RE24. They would go on to win 3-2.
Now, sure that doesn’t make Tulo “clutch.” But it does make his performance extremely “important” or “valuable,” and this is what WPA is missing. What WPA is not missing is where Tulo hits another HR in a game in hand to which he has already contributed. I don’t see why he would get more credit for a meaningless HR because he was one of the players that made the HR meaningless.
The word value means ‘important’ not skill. If John McDonald came up as a pinch hitter in the bottom of the ninth and hits a grand slam to win the game for the jays by a score of 12-11, that is the most important hit of the game, not the 3 solo shots by Jose Bautista in the first half of the game. Regardless of whether Johnny Mac was in the right place at the right time, his homerun is the difference. If Jose is up in that same situation and strikes out, the Jays lose even with his 3 homeruns in the bag.
The combined WPA for the individuals on a team add up to the EXACT wins above average for that team. It is what happened EXACTLY, and regardless of luck or ‘right place, right time’ that is what won you those games.
Guys/Dave:
With all due respect, this article & most of the commentary illustrates how misguided sabermetrics on FG has become.
Most Valuable Player in essence is WPA.
If you are measuring a player’s ACTUAL contribution to his team’s ability to win, then what’s the point of the data/metric.
Me thinks you guys have been drinking from the same Kool-Aid way too long and have forgotten what our patron saint of SaberMetrics (Bill James) said long ago:
““Good analysis never begins with statistics. Good analysis always begins with a specific question, and a question which is of interest to baseball people whether they are actuaries or artists or aging scouts.
Proposing a system which instantly evaluates everything that every player does is analogous to fixing insurance rates for drivers by attaching a box of sensors to the hood of every automobile and keeping track of how often every driver does something dangerous, and calculating exactly how dangerous that was.
It’s not the real world. It’s not practical, and it’s not useful.
Maybe, in 50 years, it will be practical or it will be useful, but it’s not now. Our general knowledge is limited by our specific knowledge. Our ability to have an impact on the discussion cannot be larger than our ability to find a question which has an actual answer.”
It appears Dave is guilty of what many of us unwittingly do more frequently: we have our favorite players we somehow identify with and seek metrics which prove our point despite the prevailing viewpoint to the contrary.
That’s a very slippery slope to dance on.
One that the anti-sabermetrics folks have traditionally complained about as their primary objection to embracing our analysis.
We need to re-examine our motives here Dave, to guard against not taking the same plunge down that slippery slope to join our saber forefathers who shrilly proclaimed their own biased views, using pet metrics, on their way to irrelevant oblivion.
I agree that sabermetrics should start with a question first – that is the scientific method after all. However, I see nothing wrong with designing a system to answer that question for the masses. WAR values and projection systems are some of the biggest advances ins sabermetrics over the years. As long as the emphasis is explaining the rational for the stats you choose to use (or create) I don’t see the problem.
Okay, so how about a rate stat that incorporates scarcity?
Something like PWOBA – position-weight on base average?
A simple quick and dirty way to do it is to z-score the WOBA relative to all other SS or 1B in the game. (Alternatively, you could incorporate all factors into a PWOBA+ type stat that takes into park effects, league, etc.).
This simple one gives you
Votto: 2.1
Tulo: 3.1
Hanley’s back at 1.8 for perspective; Miggy is at 1.7, perhaps surprisingly. If you’re more comfortable with BA-scaled numbers between .200 and .400 you can do .272 + .0255 X
Votto: .326
Tulo: .351
Et voila! There’s your one value-stat showing Tulo well ahead.
see this makes a lot of sense since what we are trying to determine is “value” so since Hanley is a butcher of a SS I hope that you can then incorporate that into the new “Value” measurement
The problem is that this is essentially what WAR does, except it does it in a much more nuanced and carefully researched way.
This is crucially a RATE stat, so Tulo doesn’t suffer from lost playing time, whereas WAR is a COUNTING stat.
Well then you could just use WAR/150 or whatever. That keeps the superior construction of WAR, namely the fact that it compares players to their *defensive peers*, which is how positions are assigned, and the fact that it includes a measure of defense.
So do you like Fangraphs’ version of WAR or B-R version of WAR? Both are different; namely, Fangraphs uses the UZR statistic for a single season (which is pretty unreliable only given a single season — Teix has one of the lowest UZR of any 1B in the game) and yet Dave Cameron’s disciples cite WAR as if it’s the end-all, be-all stat seemingly forgetting that it’s an EQUATION… one that varies from site to site. For example, Josh Hamilton has the higher WAR on Fangraphs, but Miguel Cabrera has the higher WAR on B-R… how do you justify using Fangraphs’ calculation over B-R? (I’m only asking because you seem so certain about WAR’s value as a metric of player value… so presumably you have a reason for using one over the other. What is flawed about B-R’s?)
I call bullcrap last year Tulo had a UZR/150 of -9.8 at home and +14.8 on the road how can a stat like that be a consistent measure of performance?
was his second HR less valuable? better question- did his second HR have any value? if the only thing that matters in the regular season is getting the most wins, then his second HR was worthless.
What if the other team comes back and they win by only one run? How is a home run ever worthless? The WPA does not change retroactively, and yet they needed those runs to win.
You don’t know how much a run means until the game is over. WPA is loaded with context. It measures exactly what we think a hit, out, etc. means in the exact moment it happens. It’s a statistical guide to the cheers and groans of the crowd that’s watching the game. But when the game is over, those facts very often change in hind sight.
Further, it’s not WA. It’s WPA. Someone could have a huge hit, netting huge WPA for himself, and the team can still lose. Plus, that hit could be huge simply because three other guys got on base before you. The guy on third and the guy on second get no share of the WPA when you hit a single to score them.
WPA is the saber version of Wins for pitchers. It’s a meaningless stat when it comes to player value. Its value depends heavily on what your teammates have done. I’m sure many of you touting WPA also disparage wins just as passionately.
Rather than have Dave give us his methodology first, then the results second, he has instead seemed to pick Tulo first, then backward engineer what ever logic he needs to support that selection.
We should really dismiss any and all pieces written in this manner. It’s the “Jim Rice is feared” HOF argument with fancier numbers.
1. Summarize your methodology.
2. Let your winner present itself accordingly.
If you are able to see this, you need to stop reading internet baseball blogs right now. It will only lead to frustration.
You are correct sir. Your problem is that you think that these people care about truth. These sites have nothing to do with truth.
*Cackles and jams cheetos into his maw*
Take THAT Cameron! Mwahahahaha. Sheeple.
*Licks cheeto dust off of pudgy digits*
I just want to comment that this comment section has been excellent, thanks Dave (seriously).
And to add fuel to the fire, what does Joey Votto? He goes 0-3 with 2 strikeouts before crushing a 2 run homer in the 8th inning to break a 2-2 tie. You want to tell me someone besides Votto is MVP because this is exactly what an MVP does, provide the big hit when his team needs it most. Sure it is fun to argue about who the MVP will be, but really does anyone actually believe the writers will choose someone besides Joey because I see no way its happening.
No, the MVP provides the most value to a team. If that means he hits his homers in a tie game in the 8th inning, fine. Had Votto not gone 0-3 with 2 Ks he may not have put himself in that situation to be “clutch.”
Over the last ten years, the league leader in WPA has won the MVP 6 times (Pujols ’09, A-Rod ’07, Bonds ’01-’04). In the other four years one of the two MVP’s finished in the top 3 in WPA. And only 3 of the 20 MVP’s were not in the top 20 (Ichiro finished 27th in ’01, Rollins 25th in ’07, and Pedroia 22nd in ’08). So apparently there is a correlation between WPA and the MVP.
To compare, I looked up MVP wOBA standings over the last ten years too. The leader has won 7 times (Pujols ’08 & ’09, A-Rod ’07, Bonds ’01-’04). In the other 3 years, a top 3 guy won at least one of the MVP’s. And we had 5 MVP’s not in the top 20 (Ichiro 49th, Tejada 44th in ’02, Morneau 24th in ’06, Rollins 32nd, and Pedroia 24th). Again, the trend says the MVP is usually one of the leaders in wOBA.
The fact is that the leaders in WBA are probably leading because they are having outstanding seasons and are worthy of MVP consideration. Clutchness may not be the best thing to evaluate the MVP, but it is in fact something that most MVP’s excel at. And in the context of Votto or Tulo for MVP, WBA is one of many numbers that say Votto is the choice for MVP.
No amount of defense at any position can overshadow what Joey Bag-a-donuts has done this year.
Offense >>>>> Defense
It’s funny to me that the argument in this thread is “how valuable is WPA as a stat?” and not more frequently “Tulowitzki is probably #3 in MVP voters’ minds right now and WPA further distances Votto from Tulowitzki regardless of how pin-point accurate it is.” Votto has been more valuable all season by value metrics and has been far more “clutch.” Even if clutch is circumstantial, it decides MVP winners. If Tulowitzki plays 110 games and has comparable rate stats at the SS position, he loses the award because he missed 25% of the season and because he was bad in high leverage situations, if the comparison is Votto to Tulowitzki only. The only rate stat that seems to favor Tulowitzki are widely varying fielding metrics and wOBA, which in all likelihood would be vastly lower given 150 games, I know that doesn’t count, but that’s why the fact that he’s missed time does.
I don’t think we’re having this argument if someone doesn’t write about it. It’s pretty insipid.
Dave, were you not the same guy who advised the Mariner’s front office to utilize flawed defensive data/analysis in their 2010 roster construction?
I ask that as it appears with your equally suspicious application of WAR to elevate Tulo’s relative value against Votto, you’re once again playing the same thing:
Selective application of flawed data to prove your biased picked instead of being more objective.
Correct me if I’m wrong, but given the known unreliability of single season defensive data (due to how its collected), I’m not satisfied with WAR for that reason and for a larger one in that it completely ignores the context of winning.
After all, isn’t that what the voters (meaning sports writers) go by in casting their ballots: what they see in game winning impact?
We need to re-think our application of analysis like what Dave has done as being not in the best interests of saber-metrics becoming more widely adopted.
Or, do we even care as in the movie a “Few Good Men”: we can’t handle the truth!
I think saying WAR completely ignores winning is a bit of a stretch. The idea of WAR, and Win Shares, and all of these such comprehensive measures, is to first determine what events in baseball lead to either scoring runs or preventing runs (the two things you need to do to win games), then determine a value for each of those events, and then add them up to see who provided the most value.
Someone can provide a huge amount of WAR to a losing team, but it’s not that guy’s fault, that’s for sure. It’s still a team sport where very little can be accomplished without the help of other quality players around you.
It is extremely helpful to have a context neutral stat to determine a player’s individual contribution to his team’s efforts at winning, regardless of the actual success of that team. It’s not the end of the analysis of course, but as an example, someone who’s not top 20 in WAR should probably not even be considered for the MVP.
Once someone’s in that list, you can maybe drop someone off because their team stinks if you feel contending is important, or maybe you drop pitchers because you think it should be an every day player, etc. But I firmly believe WAR should be the first slice someone takes when whittling down the list of candidates.
I delight in, result in I found exactly what I used to be having a look for. You have ended my four day long hunt! God Bless you man. Have a nice day. Bye