What’s a Most Valuable Player?
There’s more to Anthony Rizzo’s NL MVP case than meets the eye. (via Ben Grey)
Let’s say you play in a baseball simulation league, something like APBA or Strat-O-Matic, and you have a choice between two batter cards. The first batter, let’s call him Joe Slugger, will put up amazing slugging stats—the best in the league. His On-Base Percentage (OBP) and Slugging Percentage (SLG) will be at the top of the charts, perhaps historically great.
The second batter, Eddie Situation, will post very good numbers but not as good as Slugger’s. The difference is that he will excel in key situations. According to his card, his performance will increase with batters on base while Slugger’s performance will get worse with runners on base.
This is a very sophisticated game you play, and Situation’s card also shows that he will perform best in close games, while Slugger’s stats are more likely to be accrued in runaway contests. You know that a home run in a one-run contest is worth more than a home run in a five-run contest, so you’re even more intrigued by Situation’s card. When you run all the numbers, you find that Situation’s contribution to your team’s wins will be higher than Slugger’s.
Which batter will you choose?
Me, I’d choose Situation instead of Slugger. I don’t want to lead my league in counting stats—I want to lead my league in wins. Choosing the guy who hits better with runners in scoring position and in close games will result in more wins for my team. I don’t want the best hitter on my team, I want the most valuable hitter.
See where I’m going with this?
Currently, the consensus seems to be that Bryce Harper is the clear choice for the National League Most Valuable Player Award. After all, he’s batting .333/.463/.648 with 36 home runs, 84 RBI and 104 runs scored (all numbers as of Sept. 14). However, I don’t think we should totally hang our hat on those counting stats. Allow me to offer another way to look at things.
Here are the current (as of Sept. 14, and that’s the last time I’ll say it) NL leaders in wRC (weighted Runs Created, which is a similar scale to Runs Scored and RBI), along with some other key stats:
| Name | Team | PA | AVG | OBP | SLG | wRC | wRAA |
| Bryce Harper | Nationals | 574 | .333 | .463 | .648 | 133 | 68.3 |
| Joey Votto | Reds | 616 | .315 | .459 | .555 | 128 | 58.2 |
| Paul Goldschmidt | Diamondbacks | 620 | .316 | .431 | .553 | 118 | 48.0 |
| Anthony Rizzo | Cubs | 616 | .278 | .388 | .523 | 107 | 37.2 |
| Andrew McCutchen | Pirates | 601 | .299 | .401 | .509 | 103 | 35.4 |
As you can see, Harper leads the league in wRC, five ahead of Joey Votto and 15 more than Paul Goldschmidt. He also leads in wRAA, which is the same thing as wRC, but it’s expressed in runs above average instead of total runs. You can see Harper moves ahead in wRAA by posting the highest wRC total but in fewer plate appearances. Votto and Anthony Rizzo, who have the exact same number of plate appearances, also have the exact same difference in both wRC and wRAA. I like to look at both stats as a starting point.
Not all hits have the same impact, however. A single with a runner on third and two outs is worth more than a single with no one on base and two outs, for just one example. The best research I’ve seen that captures this phenomenon was posted by Tom Ruane at Retrosheet many years ago. Tom calls it Value Added Batting Runs, but it’s called RE24 at FanGraphs and Baseball-Reference. Shorter names fit better on teeny columns.
You calculate each batter’s Value Added Runs—or RE24—by calculating the difference in expected runs from before his plate appearance to after his plate appearance, adding in the number of runs that scored. You can calculate the difference in expected runs by…well, just read the article.
RE24 is a real gem of a stat and vastly underused. It tells you very different things about batters. Let’s look at the RE24 of our top five MVP contenders and compare it to their wRAA.
| Name | Team | wRAA | RE24 | Diff |
| Bryce Harper | Nationals | 68.3 | 68.06 | -0.24 |
| Joey Votto | Reds | 58.2 | 64.35 | 6.15 |
| Paul Goldschmidt | Diamondbacks | 48.0 | 51.41 | 3.41 |
| Anthony Rizzo | Cubs | 37.2 | 55.47 | 18.27 |
| Andrew McCutchen | Pirates | 35.4 | 47.69 | 12.29 |
Harper’s RE24 is about the same as his wRAA (too many acronyms!), but every other batter increases his impact once we add in his performance in each specific situation. How does this happen? Well, here are a couple of factoids (out of many possible factoids) to consider:
- With runners in scoring position, Bryce Harper has batted .294/.476/.559. With no one on, he’s batted .336/.443/.695.
- With runners in scoring position, Anthony Rizzo has batted .305/.425/.602. With no one on, he’s batted .249/.344/.488.
With runners in scoring position, Harper has actually been worse than with no one on, but Rizzo has been much, much better than his no-one-on stats. The nice thing about RE24 is that it takes all the differences in base/out situations (there are 24 of them) and sums them up in a single number.
This is called situational hitting. People tend to ignore it because it’s not very likely Rizzo will repeat this breakout again. But MVP awards aren’t given for repeatable performances. They’re given based on what actually happened. Rizzo’s RISP performance actually happened.
You probably know all of this; maybe you’ve looked up RE24 on baseball websites many times before. But there’s another wrinkle to consider, something I first researched in 2007. That is, the margin of victory in a game.
This is a simple yet powerful idea, too: runs in close contests are more important than runs in blowouts. The object of the game is to win the thing, not to run up a big run total. Thanks to WPA, we can quantify exactly how much events matter to winning in close games vs. those in blowouts. For instance, a batting event in a one-run game was worth 1.38 more than average, while a batting event in a three-run game was worth 0.97 of average. I’m going to call this a Margin Factor (and that’s all the detail I’m giving you here; read the article for more).
So I multiplied each batter’s RE24 in each game by the Margin Factor of the game. Below you can see how our Big Five rank in wRAA, RE24 and this last stat, which I’ll call Game-Adjusted RE24 (GameRE24 for short).
| Name | Team | wRAA | RE24 | GameRE24 |
| Bryce Harper | Nationals | 68.3 | 68.06 | 56.4 |
| Joey Votto | Reds | 58.2 | 64.35 | 60.6 |
| Paul Goldschmidt | Diamondbacks | 48.0 | 51.41 | 55.5 |
| Anthony Rizzo | Cubs | 37.2 | 55.47 | 54.0 |
| Andrew McCutchen | Pirates | 35.4 | 47.69 | 43.4 |
Most of our batters lost ground in the transition from RE24 to GameRE24 because one-run games tend to be low-scoring affairs. However, Paul Goldschmidt actually increased his total impact by factoring in the game situation. (Fun fact: Goldschmidt leads the majors in RE24 in one-run games at 25.4. Kris Bryant is second at 19.1).
In our overall GameRE24 totals, Votto is now in the lead, with Harper, Goldschmidt and Rizzo virtually tied for second. It’s not a runaway race anymore; it’s a dead heat.
Of course, you probably would want to plug these numbers into WAR, where Harper’s (and McCutchen’s) positions would factor in. So, too, would Goldschmidt’s fielding numbers. You probably also should include the fact that Harper has accrued fewer plate appearances. There is much still to play with, but the basic concept is fully formed.
In 1979, Don Baylor won the American League MVP Award, largely thanks to an impressive RBI total of 139. Back then, we didn’t have any advanced stats or breakouts to make sense of those numbers, so we grabbed onto some evidence of situational hitting wherever we found it.
Here’s what we now know. Baylor came to bat with runners in scoring position 258 times in 1979, 22 more times than the major league runner-up (Darrell Porter). His OPS in those situations was .981, which is obviously very good but was still just 15th in the majors. His RBI count was driven more by opportunity than performance.
If we had been able to see his RE24 at the time, we would have seen a total of 37.9, eighth in the AL behind Fred Lynn’s 60.6. In fact, Lynn led the majors in OPS with runners in scoring position at 1.188. Remarkably, his RE24 was much higher than Baylor’s despite coming to bat only 166 times with runners in scoring position.
This debate of “What does valuable mean?” has gone on long enough. Many people have given up on the topic, ceding the floor to simple counting stats because, you know, who really knows? But we do know. Value = Winning. Winning is a function of hitting, hitting in the right situations and in the closest games. If you want to honor the best hitter, vote for the Silver Slugger. But if you really want to honor the Most Valuable Player, take a good look at his Game-Adjusted RE24. That is where value lies.
References & Resources (and Caveats)
- As always, thanks to Retrosheet, FanGraphs and Baseball-Reference for their wonderful work and tools. Particular thanks to Tom Ruane for his research. Tangotiger has been my guide in much of my WPA research over the years.
- I used the 2007 Margin Factors from my original articles. Ideally, I’d update those for the current run environment. Also, the RE24 stats at FanGraphs and Baseball Reference are based on evolving run expectancy tables and won’t be finalized until the season is over. I’m taking it on faith that these RE24 numbers are pretty close to the final deal.
- Tom Ruane, Retrosheet, “The Value Added Approach to Evaluating Performance”
- Dave Studeman, The Hardball Times, “Long Live Baseball Analysis”
Sorry, I couldn’t resist to lol when I read this article. According to your logic, why not simply use WPA to evaluate MVP?
WPA values events at the end of close games more highly than those at the beginning of close games. This is the most legitimate complaint people have about using WPA for MVP discussions. The approach outlined here doesn’t do that.
Hey, but:
“MVP awards aren’t given for repeatable performances. They’re given based on what actually happened.”
And:
“I don’t want to lead my league in counting stats—I want to lead my league in wins.”
Rizzo’s WPA(6.84 to date, which leads MLB) actually happened, right? And his WPA leads to actual wins, right?
And late inning performance in close games/clutch situation is surely a very valuable skill, if it’s a repeatable skill. Just as hitting with runners on base.
This is just following your logic in the article.
I think it’s more to be interpreted as: “He made a bunch of hard contact with guys on base. He should be rewarded for making that hard contact, regardless of whether or not he would be expected to do so. He should not be rewarded for how much that hard contact happened to matter because of how many runs the other team scored or did not score.”
“And his WPA leads to actual wins, right?”
No. And that’s where there’s the disconnect.
WPA doesn’t measure where wins come from – it measures how dramatic the wins are. A game-winning homer in the 1st is just as valuable as a game-winning homer in the 9th, but WPA will weigh the latter much more heavily.
I want the guy who will win me 10 boring games over the guy who will win me 5 dramatic ones. This is why WPA is a bad measure for this exercise. WPA/LI might be better though.
WPA/LI does seem like it’s a fair thing to look at–it avoids penalizing players because their teammates don’t give them opportunities to come to bat in high-leverage situations, and measures how well they did in the situations they faced.
As it happens, Harper is blowing everyone else away in WPA/LI:
Harper 8.30
Votto 6.96
Rizzo 5.68
Goldschmidt 5.17
Upton (!) 5.11
McCutchen 3.98
I’m a fan of WPA/LI, too, but it is hell to describe. In particular, it’s very hard to describe what context is stripped out of it, and what is left in. I know Tango has taken a good shot at it, but I can’t find his description off the top of my head.
Here’s a link to an article I wrote about WPA/LI. Check out the comments, in particular. The important thing is that WPA/LI treats every plate appearance the same, while RE24, WPA and the approach I’m laying out here give more weight to certain plate appearances and games.
http://www.hardballtimes.com/talking-situational-wins/
Having re-read that article and thought about it, I’m not sure that WPA/LI is what we want to use for a Value award. To me, the point of establishing value is acknowledging that some PA’s are indeed more important than others. WPA/LI, while an awesome stat, doesn’t do that.
So WPA weighs events at the end of close games greater than at the beginning. It “weighs” each event by its importance at the time it occurred, naturally, by measuring how it affected the likelihood one team wins.
The absurd contrivance of “GameRE24” weighs a leadoff single more if the game ended up close than if it didn’t, while its forebear, RE24, quite reasonably credits a hitter that leads off an inning the same weather he’s ultimately driven in or not. Do we give hitters credit for events that came after or do we not?
I feel like WAR + Clutch is a fairly easy way to express this combination of production and context. Using FanGraphs’ versions of those stats, Bryant is just ahead of Cutch for the league lead.
That’s an interesting idea, but Clutch is a different animal. It doesn’t capture as much of the situational hitting that RE24 captures, and it also puts a premium on when an event happens within a game. I’m trying to avoid that here. (see above response to the WPA comment)
Yeah, I’m not sure why I thought it would be different. I’ve used that 1st vs. 9th inning homer example before to explain value, but apparently forgot to think about that here. Thanks for the response.
BTW, if you like WPA, then you should just insert a batter’s WPA into the hitting part of WAR. Makes it easy, and also makes Rizzo the clear choice for MVP. However, adding Clutch to Batting Runs is a strange hybrid that misses a lot of value provided by total WPA.
My approach here is designed to capture the best of WPA (that is, bases and outs, game margin) but to also set aside the most objectionable part of WPA (inning). Innings make for drama and story, but they don’t provide value.
To clarify again, and including Matt’s comment from above:
“And late inning performance in close games/clutch situation is surely a very valuable skill, if it’s a repeatable skill. Just as hitting with runners on base.”
This is covered in the article I originally wrote on this subject in 2007. The point is that a home run in the first inning of a one-run game has just as much value as a home run in the ninth inning. Hitting that homer late in the game isn’t more “valuable”, it’s more dramatic.
Hey, no one is a bigger fan of WPA than me. I’ve been writing about it for ten years. But it’s important to recognize what it is and what it isn’t.
This is laughable. How do you know it will be a one-run game when you hit a homer in 1ST INNING? It will have a small chance (if the score is tied, then around 30%) to be a one-run game, and the other times it won’t be a one-run game. That’s it.
Of course the solo HR in anywhere in a game is of the same value IF the game’s margin of victory is fixed. But you can’t assume the margin of victory is fixed.
And 30% is by summing +1 and -1 margin of victory. +1 and -1 is about 15% respectively.
The point is baseball is a sport that you can only know what happens before, not what happens after. (Same as any sport.)
I’m glad you have such a terrific sense of humor.
“Hitting that homer late in the game isn’t more “valuable”, it’s more dramatic.”
I have a decent sense of humor when I found this laughable claim.
BTW, if you think home runs late in games are more valuable than those early in games, then you must also think that wins at the end of the season are more valuable than wins in the beginning of the season. If so, you would like the Drama Index and the Championship Value Added stat that Sky Andrecheck and I created and published in the THT Annual from 2009 to 2011. Check it out if you get a chance.
“Wins at the end of the season are more valuable than wins in the beginning of the season. ”
It’s correct. But of course, only to teams in playoff hunt in late season. To those teams in 1st overall pick race or already booked a division win, their wins at the end of the season is less(or much less) valuable.
The point is that you are not so sure that if you are in playoff hunt or not in the beginning of the season. It’s only a relative small probability to most teams.
First of all, please give Dave the respect he deserves — he literally created WPA, so I think he has a pretty good idea of what it measures, as well as a good grasp on the topic of win probability as a whole. Disagree with his points all you want, but to call them laughable is just closed-minded and disrespectful.
To the point at hand, the entire argument Dave is making is that it doesn’t matter that we didn’t know the final score at the time of the first inning HR. We’re just trying to evaluate the plays in hindsight, and when we do so, we have the privilege of already knowing the outcome of the game. This allows us to evaluate each play in the context of the entire game, rather than in the context of just what had happened so far.
To put it another way, WPA tells us how excited the player and the fans should be about the outcome of the play. If Rizzo hits a 1st inning HR in the tie game, certainly he should be less excited than if he hit the same HR in the 9th. But when we look back on the game, and we see that it ended with a one run margin, it doesn’t matter when or how those runs were scored. They each had the same impact on the end result. The player who hit the 1st inning HR had just as much responsibility for the win as the player who hit the walk-off. Just as getting an eagle on the first hole in golf has the same impact on your score as getting an eagle on the last hole, even if the latter was more “clutch”.
Hitting that homer late in the game isn’t more “valuable”, it’s more dramatic.
Of course it is. The value of a solo home run with no outs in the ninth inning, for example, is greater than a solo home run with no outs in the first inning. There’s less future innings to play when you’re in the ninth inning, which has implications for the probability of winning the game; thus increasing the importance or value of each play on the margins. It’s *no* different in principle from what you’re doing with your stat, which is to contextualize each offensive play and derive its value based on context (e.g., a single with no outs and no runners on base vs. a single with no outs and a runner on first).
But when we look back on the game, and we see that it ended with a one run margin, it doesn’t matter when or how those runs were scored. They each had the same impact on the end result.
This is entirely false for the reasons I stated above.
Ask a team if they’d rather make up a 1-0 deficit in the first inning or make up a 1-0 deficit in the ninth inning. This isn’t just about the excitememt of late-inning plays; these plays have actual added impact on your team’s chances to win the game. Now, as I stated below, there’s a significant drawback with computing value based on contextual analysis, a downside that plagues both WPA and RE24. Unfortunately for Mr. Studeman, he fails to recognize this drawback with regards to RE24.
It’s true that WPA is a natural extension to RE24, which I find to be one of its biggest appeals. But there is a difference, too.
To me, the point of this exercise is to isolate and identify players who most affected wins and losses during the season. The unit here is the final score of each specific game. It doesn’t matter to this process when runs are scored in the game (which is WPA) or when wins are achieved during the season (Championship Value Added). But winning games requires runs.
RE24 is the best system I know for attributing specific runs scored (and outs made) in a game to players because it does take context into account while also giving fair value to each event in the context. But runs are all the context you need in this approach (and not the timing of those runs), because runs are the final arbiters of games.
If Mr. Hunter requests it, I’ll type an appropriate counter to this rebuttal. As you’ve shown in your posts below, you’re not interested in debate.
I don’t understand why “Margin Factor” is relevant to the discussion. If Bryce Harper hits a solo homer to break a tie game with two outs in the ninth and the Nationals then tack on 4 more runs so that the club wins by 5, how is this less valuable than if Paul Goldschmidt does the same thing and the D-Backs don’t add the insurance?
Harper has hit .312/.429/.623 in tie games, and .328/.452/.627 in one-run games. Levying a significant “clutch” penalty on him seems to me to be harsh. He’s easily been the best player in the National League; it’s not an easy thing to do when the team around him has been falling apart.
Well, if you think *when* an event occurs during a game, then you should use WPA. But the logic of saying that an event has more impact on a win in a close game seems irrefutable to me. And, Harper doesn’t lose his standing because of the margin factor–he loses it because of the difference in his hitting with runners on base.
Dave, I’m not quibbling about the wRAA to RE24. When you make the adjustments (which take into account hitting with runners on base), Harper is ahead of Votto and way ahead of the others. It only becomes a 4 way race if you take into account margin of victory.
Are there hitters (and pitchers) whose performance with runners on base is different from with no one on? Of course. That matters. For instance, Joey Votto has hit significant better both this year and over his career with runners on base. I see no problem with crediting him for that. He’s still a lesser hitter than Harper, and then if you add in the defence, it’s not close. The only way that it is a close race is if you take into account “margin of victory”, which is an ex post facto way of looking at clutch performance. Bad enough that you take into account something which is essentially flukey but worse to do it ex post facto.
Thanks, Mike. I understand what you’re saying. It does sound like you might be a fan of WPA, which looks at game margin in real time, as well as the base situation and number of outs. If you don’t like WPA, I’d be interested in understanding why.
I use ex post facto game margin because it captures some of what WPA captures–the intensity of the game–on an overall basis instead of an at-bat basis. By approaching the game intensity that way, we’re able to avoid some of the wide swings you see between early and late at bats in a game.
It’s analogous to a season. An win early in the season is worth as much as a win late in the season–after all, they use total wins and losses to determine the division leader–but players on teams that contend are often given more credit than those that don’t. This is an ex post facto fact too, but it’s a way of capturing the intensity of play.
I don’t think you really answered his question. Your value stat rewards the guy who hits a HR in the top of the 9th that wins a game by 1 run, more than the guy who hits that same home run and is then followed round the bases by four of his teammates. Is this a fair analysis?
Do you not consider this result to be a flaw in the system you’re advocating?
Quoting tz from a later comment:
It’s the byproduct of two items: (1) recognizing that the relative value of a run in a close game is greater than it is in a blowout and (2) attempting to remove the difference in win value across innings (like you’d have with WPA).
I don’t believe there’s a perfect way to accomplish both of those and not get the type of anomaly you describe. So you go with the stat that incorporates the elements you want for a particular use, and live with the limitations that arise from it.
As also mentioned in the comments, Tango suggested that I create two different sets of factors: one for the winning team and one for the losing team. It’s a good idea I believe that would help mitigate the specific situation you describe.
By the way, I should also say that it’s the fairest system I can conceive of, given my goals. Tango’s tweak is a good thought, and I’m also open to other ideas to make it fairer still.
I think you need a third adjustment for importance of the game. Goldschmidt’s production comes at more important times within games, but his team’s games are less important.
Maybe this is just differing opinions of ‘valuable’. If you included a game importance/leverage factor, the ‘valuable’ bogey would be contributing to the team making the playoffs, vs. contributing to winning any game.
Yes, I’ve written about game importance before, such as here: http://www.hardballtimes.com/the-drama-index/
You’ll notice that I called this a Drama Index instead of a Value Index. In the end, a win in April means just as much as a win in September. They’re both of equal value, just as a home run in the first inning of a close game is as valuable as a home run at the end of a close game.
If you want to debit Goldschmidt’s stats in general because his team isn’t going to make it to the postseason, I can understand that. To me, that’s a subjective decision.
It’s interesting to me that Harper’s performance declines with RISP while Rizzo’s improves, but their resulting RISP numbers are pretty similar (Harper = .294/.476/.559/1.035; Rizzo = .305/.425/.602/1.027).
So other than their nearly equal numbers in RISP situations, Harper’s numbers (.336/.443/.695/1.138) blow Rizzo’s (.249/.344/.488/.832) away.
This is a good example of why you can’t just look at one split, but need to accrue all the split information into one number, which is what RE24 does.
A factoid is something purported to to true that actually is not. Factoid is one of the most commonly misused words.
Well, at least I’m not alone.
I think that “factoid” has been used that way enough that its meaning has now changed.
If you were talking about two fairly similarly valued players, I could see looking at some factors like this, but giving the MVP to Rizzo over Harper is just ludicrous. The Difference in value between Harper and Rizzo (using FG WAR) is greater than the difference between Rizzo and Michael Taylor.
There should be an award given to the best player each year. I don’t understand getting hung up on how to define “most valuable” (which by the way if you take literally would require looking at salaries). Just give the MVP to the best player. It’s not rocket science.
That’s why I created the introduction in the way I did. I think there is a place for a “Most Valuable” award and I think value can be clearly defined as contributing the most to winning. I think I’ve shown pretty well that four guys all contributed about the same amount to their team’s wins.
I agree that many people have given up and just want to give the award to the *best* player. I guess this article was written for those who still believe in the idea of value.
I like that your arguments for Rizzo (using RE24 and adjusting it specifically to hurt Harper) STILL have Harper better than Rizzo, even though it’s closer.
Blaming Harper for his teammates being bad is not really fair.
While Harper may not be as “valuable” to you because Rizzo or the others perform better with men on base, Harper is creating value by “starting” a rally – getting a key hit with nobody on base. That’s still pretty valuable from where I’m standing.
You can frame the argument however you want. It’s “clutch” to start the rally or “clutch” to get a hit with guys on base. Bottom line is Harper has been the best player in the NL, and has created the most value, even if his team has sucked.
Well, that’s interesting. If a guy gets a hit and then no one rallies behind him, has he created less value than the guy who gets a hit and has people rally behind him? Aren’t you confusing value provided by the following batters vs. the first batter?
And I still haven’t expressed my own idea…
My opinion:
WPA is the only thing directly corresponds to (batting) wins.
In MVP process though, you could eliminate something from WPA though. My opinion is that you should eliminate all unrepeatable and luck-based skills, retain repeatable skills only. (You surely cannot randomly and unlogically eliminate one thing and leave others there, like what this article does. Use WPA as batting value is acceptable imo, but I won’t suggest it.)
Season woba is (mostly) repeatable, so retain the repeatable part of it.
Hitting with runners on, hitting in late inning, hitting in close games, are mostly unrepeatable and luck-inflated though, so most of all these should be cleared out from MVP discussion.
I have to agree with Mark earlier above, when he said, “Harper is creating value by “starting” a rally – getting a key hit with nobody on base. That’s still pretty valuable from where I’m standing.” Or as the Detroit Tigers broadcasters like to say at every Tiger game of the year – “Do you like dinner time with the family? If so, there are people that need to set the table. There are also people that need to eat.” Those that set the table are typically the top of the order guys who have high OBP and can get on base, and those that eat are the power RBI guys, the 3-5 sometimes 6th hitter in the order.
But I would take it a step further. Every batter that steps up to the plate in MLB should think BOTH ways and become what the situation requires, because let’s face it after the first batter in the very first inning, that is the ONLY guaranteed time you are an OBP guy where the goal is to get on base, after that, its where you leave the 3rd out in the batter order that determines who are the “leadoff/OBP guys” and your “power hitting/RBI guys.”
This is why I created a Bases Moved stat, along with a contact consistency stat, and a plate appearance outcome stat. Those three together totally encapsulate value in my opinion. And I’ll tell you something else, the rate stats of batting average/OBP/SLG/OPS totally do not capture all the positive things a hitter does. Nowhere do those stats factor in getting on base via error, or the value of contact outs. However, my bases moved stat does. Sure, contact outs and errors, ARE included in the DENOMINATOR of those rate stats as a NEGATIVE, that LOWERS all these rates, but it does NOT include contact outs or errors as a POSITIVE in the NUMERATOR.
Of course, RE24 includes a “rally starting” component. The only reason to add anything more is if you believe that Harper’s rally-starting hits are somehow more “inspirational” than others’.
Is the reason we have to spend so much time looking for reasons that Harper ISN’T the MVP solely because he’s an insufferable prick? Is it because the division races are largely decided and boring? Maybe because all the best prospects were called up way before September, so not much to analyze there?
I know WAR has fallen out of fashion, but Rizzo isn’t even first on his own team in it. Harper leads Votto by more than a win and a half… and if you want to switch from wRC to wRC+. well, Harper is having the 9th best full season in the 140 year history of the National League, and 5 of those seasons came from fully juiced Bonds and ’98 McGwire. Let’s not overthink this.
This isn’t personal at all. If the facts supported Harper, I’d be all for the guy. I know some will think I’m overthinking things, but that’s what people say about sabermetrics in general (including the people who invented WAR). I’m also not advocating against WAR, just suggesting a different approach to the offensive component of WAR.
Left unsaid in your text, but clear in the tables is… Harper is still ahead of Rizzo in the other categories. And where he trails others, the suite of stats together again make Harper the obvious choice.
These hoops should only be jumped through when you are trying to decide between two players who are otherwise close, as they are not here.
Not sure what other stats you categories you are referring to, but to me, these aren’t “hoops.” They’re the foundation of a basic understanding of value.
Again, I didn’t write this article because I’m against Harper receiving the MVP. It’s surprising to me that people are interpreting the article this way. I wrote it because I strongly believe we can still benefit from identifying Value.
Then you need to write it differently, rather than frame it as, and headline-grab us with, Rizzo’s MVP case.
I didn’t frame it as Rizzo’s MVP case. You’re confusing a picture caption with the actual headline and article.
By the way, Tango has a post about this article, which you can read here:
http://tangotiger.com/index.php/site/article/context-based-value
His idea is to add a factor based on whether the player’s team won the game or not.
Wow, my only guess was that this article was written as some kind of very about whether you could make Rizzo seem like they logical choice for MVP over Harper. Talk about some mental gymnastics!?!
First, in games that are close you overlook one important detail : With men on base Harper’s OBP jumps over 30 points. Without looking it up, I’d venture a guess that the number of fastballs he sees also decreases. Why could this be? Perhaps, and this is a big stretch, teams don’t pitch to Harper with the game on the line. I know I wouldn’t.
Second, even if you’re going to give Rizzo the advantage for being CLUTCH! shouldn’t we then deduct points for being so much worse than Harper when they game’s not in the line? After all, let the course if 162 games winning big helps a team dance their bullpen (or in the Nationals case provides a buffer for its likely implosion).
Finally, outs, the object of hitting is to not make outs, Harper was way better at that than anyone this year.
Stupid autocorrect
Why do people keep saying that I want to give the award to Rizzo? I don’t say that at all. In fact, my analysis suggests that four batters are in a virtual tie.
The analysis doesn’t overlook anything. It takes into account all events, and it debits Rizzo’s stats with no one on base. It also includes the impact of walks with men on base for Harper. IF pitchers are taking the bat out of his hands, well, that makes him less valuable.
I suggest that people read the linked article to RE24 in order to better understand how it works.
Probably because of the photo and, to a lesser extent, the caption. Very interesting article – who are the AL leaders?
Donaldson, by a good margin.
Well, because Harper is clearly the MVP, Votto and Goldschmidt are clearly 2-3 although the order is debatable and Rizzo is distant 4th. Rizzo’s mere inclusion seems forced.
Still, while most agree that lineup protection doesn’t factor in over the course of a season, I do believe that teams approach the Nats lineup with a “don’t let Harper beat us” attitude.
As for understanding RE24, patronizing much? Being walked in key situations will naturally decrease your opportunity to excel. Here’s a little experiment, replace RE24 with RBI’s and see if it changes your mind because basically that’s the argument you’re making. It also fails to account for the increase in pitch counts or the batters behind Harper getting to see more pitches from someone in the stretch.
Finally, we haven’t even accounted for the fact that UZR over a single season is garbage so comparing fWAR is a fool’s errand.
I don’t know what you mean about Rizzo’s participation being forced. I started with the Top 5 in wRC and winnowed the field from there. Rizzo was fourth. Seems pretty unforced to me.
Sorry you found my remark about RE24 partronizing. My particular point there was that, if Harper is walked in critical situations, that impacts his value. The defense takes the bat out of his hands. So the system gives him appropriate level of credit for that. I don’t see the need to give credit beyond the RE24 impact of a walk. An intentional walk hardly counts as any pitches thrown in my book. As for pitching out of the stretch, are you saying that Harper is often intentionally walked with the bases empty?
Finally, I don’t understand your comment about fWAR and UZR. I only alluded to those at the end of the article.
It doesn’t take into account wRC+, and the adjustments it puts into the formula… which makes Harper CLEARLY better than the other three.
RE24 is adjusted for ballpark and general run environment, just as wOBA is. What adjustments do you feel are missing from my approach?
Sorry…I meant wRC+.
Getting back to 1979, you could simply use WAR to see that Baylor was nowhere near the most valuable player in the American League:
Baylor: 3.7
Lynn: 8.8
(Baseball Reference WAR)
I can see using sequencing when the players are very close by other measurements, but even though I agree with the premise that we would like to base the award on what actually happened, using sequencing seems to me to move back towards rewarding or punishing a player for what his teammates did rather than what he did.
Regarding Baylor, Lynn and WAR: Many people understood that Lynn had better overall stats (especially people in Boston!) but the writers were looking for an indication of Value. This is my point, that many voters still want to find that elusive definition of Value and that’s what I’m presenting here. WAR, as it’s currently configured at BRef and Fangraphs, is the ultimate counting stat. It doesn’t capture the Value that many MVP voters are looking for.
Regarding sequencing, we have talked about creating an RE Leverage Index, which you could use to adjust RE24. I’m sure it would make some difference in the numbers, but I doubt that it would make a big difference. Just a gut feel, however.
totally agree you don’t make the playoffs, you’re not the most valuable player.
Disclaimer: I am a total layman when it comes to sabermetrics. But here is my opinion, such as it is.
In the “Major League Baseball Most Valuable Player Award” article in Wikipedia, it states: “The BBWAA does not offer a clear-cut definition of what ‘most valuable’ means, instead leaving the judgment to the individual voter.”
So it seems there is at least some leeway for judgment here. As far as I am concerned, value in baseball translates to runs, wins, and titles, or some combination of the three. I suppose some may interpret value heavily in favor of individual stats and achievements. But baseball is a team sport, so I am going with valuable as meaning valuable to the team’s goals.
To that end, I see a player such as Yoenis Cedspedis as being an MVP candidate. If one assumes Eddie Situation as a preferable player in a particular game, as is argued in this article, then Cespedes appears to be the Eddie Situation of a particular part of the season. He arrived on the Mets and as far as I can see has sparked them into contention where before he arrived they were floundering.
I know that correlation is not causation and that the Nats were not exactly doing a lot to thwart the Mets’ rise. But to me Cespedes is the essence of what I consider a valuable player to be – a relentless catalyst for runs, which help get wins, which help a team get to the postseason. If a player hit a solo home run in every game, that would be a mammoth individual achievement, but if his team finished in third place without a playoff berth, I would not see that player as an MVP candidate.
Given that I am a layman, I also think that the non-trade of Wilmer Flores and the return of David Wright also played a part in jump-starting the Mets. I don’t think those events are as easy to quantify as what Cespedes has done (and Wright has been a perennial statistical disappointment to me so I see his influence as leadership/grit), but I would be interested to see if there are numbers backing up that (hypothetical) effect. It could all be just a statistical coincidence based on the makeup of the Mets after putting those player-pieces back in place, not some ineffable effect of inspiration or emotion.
Having said that, I’d be in favor of the MVP’s going to Clayton Kershaw, but I guess pitchers have the Cy Young Award so he probably wouldn’t get it. Of the players listed in the article, I would have to lean towards Rizzo or McCutcheon, depending on which one has contributed more to the team’s position in the playoff hunt.
RE24 handles all of the base-out situations, so that the “starting” the rally is included. Trying to simply it for ease of discussion: you get +.25 runs for each base the batter takes, or that the runners move up. E.g., a solo HR (4 bases) gets you +1 run. A bases-loaded triple gets you +3 bases for the batter, and a total of +6 bases for three runners, for a total of 9 bases, or +2.25 runs.
That’s basically how that works. So, what Dave is trying to show is that while you do get positive credit for great performances with bases empty, the “multiplier effect” with runners on simply is more powerful.
So, that’s all RE24 is: count the bases and count the outs. RE24 simply gives it more precision for each base and each out. Moving a runner from 2b to 3b isn’t as valuable as moving him from 3b to home. So, while each is 1 base, the value of that last base is greater.
As someone, I think it was Dave, noted: RE24 is the bridge stat between RBIs and Runs Created.
I’d like to leave an overall comment. A lot of people are referring to the method in this article as “hoops”, but there are not more hoops involved here than in all the basic WAR calculations. wRC is a complicated stat, after all. In some ways, RE24 with an adjustment for Game Margin is an easier stat to grasp and calculate.
I’m suggesting that there is a better way to calculate Value, that we shouldn’t give up on it. If you replace what ever run-based stat you prefer with GameRE24, you’d be much closer to a Value metric. And it wouldn’t require any more hoops than the current systems.
Isn’t RE24 influenced by not just a player’s performance but also his teammates’ performance, though? Taking extreme hypotheticals, isn’t a player with better hitting stats who always hits with 2 outs and nobody on going to lag a player with worse hitting stats who usually comes up with runners in scoring position?
Relying heavily on RE24 therefore strikes me as penalizing a player just because the rest of his team – and maybe in particular the players hitting in front of him – aren’t very good.
Yes, Hank G. made the same comment above. Here’s my response:
“Regarding sequencing, we have talked about creating an RE Leverage Index, which you could use to adjust RE24. I’m sure it would make some difference in the numbers, but I doubt that it would make a big difference. Just a gut feel, however.”
Ok, thank you.
Maybe I misunderstand what you mean by RE Leverage Index, because I’m interpreting that as putting even more importance on having the opportunity to hit with runners on base (i.e., good teammates).
Leverage Index is a way of calculating the potential impact of each base/out situation. You can take the RE24 of each outcome and divide it by its LI to “normalize” RE24.
As a proxy, I looked at the WPA-based LI for each batter. Harper is at 0.97, Votto and Rizzo are at 1.01 and Goldschmidt is at 1.0. If you take Harper’s RE24 of 68.06 and divide by 0.97, you get a normalized total of 70.16. So there’s an impact of about two runs up for Harper, and probably one run down for Votto and Rizzo.
Dave T., I was wrong about the potential variety of situations seen and you were right to bring it up. Normalizing has a bigger impact than I thought it would–see my comment at the bottom of these comments. And thanks for bringing it up.
Fascinating article Dave. I’m sold on RE24, but not so sure about your inclusion of hitting in a close game.
Here’s a hypothetical:
Harper hits a 1st inning solo home run, and the Nats go on to win the game 8-0.
Rizzo hits a 1st inning solo home run, and the Cubs go on to win the game 1-0.
In these two games, I agree that Rizzo’s home run turned out to be “more valuable” than Harper’s because it was a close game. But the individual player’s contribution is the same. It seems to me that the extra value of Rizzo’s home run was determined by what the rest of his team (and the opposing team) did or didn’t do. Seems strange to give him extra individual credit for it.
Thanks, C. I see your point. You might be interested in WPA, which gives both of those home runs the same value.
Perhaps there is some way to capture the difference between those two events without going full WPA, but I haven’t figured out what it is. In the meantime, I’m okay with the difference. Perhaps the difference is in teammates, but perhaps the difference is the quality of the opposing pitcher. Figuring that out would take a whole lot of hoops.
lol at your use of “hoops”
It’s the byproduct of two items: (1) recognizing that the relative value of a run in a close game is greater than it is in a blowout and (2) attempting to remove the difference in win value across innings (like you’d have with WPA).
I don’t believe there’s a perfect way to accomplish both of those and not get the type of anomaly you describe. So you go with the stat that incorporates the elements you want for a particular use, and live with the limitations that arise from it.
How about this example: you’re trailing by one in the top of the 9th, two outs, bases loaded. You hit a single and everyone advances one base. Then the next batter hits a grand slam.
The grand slam devalues your single, even though without your single it could not have happened. This is an extreme case, but of course every event will affect the ones that come after. If you hit a leadoff single the next batter will have an elevated BABIP on ground balls. If you give your team the lead late better relievers enter the game on your side, and worse relievers on the other. That’s the only the most obvious way strategy changes based on the mid-game score; the relative importance of each subsequent run determines strategy for pitching (pitch to contact or pitch around a hitter?), offense (pinch-running, steals, sacrifices, hit-and-runs), and defense (infield in with a one-run lead, outfield in the “no doubles” alignment with a multi-run lead). The context that matters is the context in the moment, not a homogenized context spread over the whole game.
If we impose the game’s ultimate margin on the weight of intermediate events, we might as well include which team won, right? We know that leverage varies on this basis: with a lead, batters’ leverage drops and pitchers’ leverage rises. For batters this goes against the intuition that the runs scored in a loss are futile and the runs scored in wins are vital (and we are literally in the business of weighing runs), but if you run the numbers you’ll probably find higher average LI for batters in losses than wins. This contradiction is at the heart of the problem with GameRE24. If I’m wrong about the numbers and average LI for batters is greater in wins than losses, the contradiction shifts merely shifts to pitchers. There’s no escape from it without contradicting the empirical results for one group or the other.
The only way the contradiction is resolved is if leverage changes during the game as the score changes. If you hate inning-by-inning variation so much, maybe calculate WPA with the first-inning table applied to every inning. If game margin is applied, especially if direction of the margin is considered, I don’t see how that can be any margin but the one in actual context.
Hey, here’s a comment I left in an old article, and I still like it:
1. If you think each event should be treated equally, independent of the situation, use linear weights.
2. If you think each event should be weighted by runners on and outs of each situation, use RE24.
3. If you think each event should be weighted by all the game factors of each situation, use WPA
4. If you think each situation should be treated equally, use Situational Wins
In this article, I’m proposing that something between #2 and #3 is most appropriate for the Valuable award, because it most easily lines up with actual game wins.
Nice article. You won me over.
I just want to say that the author is not:
1) proposing this stat as the “One Ring to Rule Them All” stat. He is quite clear that he is advocating its use for a very specific purpose.
2) the MVP for the purposes of this article is not the best player in baseball.
Interesting, thought provoking article. However, which makes it important, I believe, is the frequency of this occurence. That is, if such a large disparity often exists between what one may consider “talent” (linear weights) and value (WPA/RE24) then this piece is extremely relevant to how we may approach the MVP award. However, if such a large gap is extremely rare, then, I don’t think theres a problem with quantifying offense value via linear weights based metrics. This may be why it seems to some of the commenters that you’re making mountains out of mole hills. My guess is such a huge gap is quite rare, but I very easily could be wrong.
Give us some context!
Tony
I’ll work on it, Antonio. You might well be right that a situation like this is quite rare, but why not just use a metric like this in the first place? At least, that’s my opinion.
The problem that I have with using statistics like these when trying evaluate value first, is that we don’t have a clear operational definition of what “value” really is. Some people use WPA, some use RE24, and now I’d imagine that someone will adopt your analysis in this article. As evidenced in the comment section, sabermetricans and non-saber people alike are going to differ on what value is. This, to me, call into question construct validity of value, ie: does it really exist? I have a background in science, and when there is a lack of converging operations ( slightly different operational definitions of the same construct leading to the same outcome) the construct is soon discarded. Now, I’m not saying that value doesn’t exist, but it’s an elusive concept. What I normally do is first look at linear weights based metrics when evaluating performance. In my opinion, it’s a more parsimonious way of evaluating performance. If two players are roughly equal, then I look into the metrics that have been mention in this piece. You seem to be a fan of going about this in the opposite way (correct me if I’m wrong) where you’d look at the metrics mentioned in this article first, and then perhaps looks at linear weights based metrics. I’m not sure I agree with this approach given what I’ve said above.
I get it, Antonio. There’s no universal definition of value, so you fall back on the basic counting stats and make value a secondary consideration. I wrote my article in order to push back on this mindset, but I understand if I haven’t convinced you. Gave it my best shot.
If we can go back to the “doesn’t matter if a solo HR in a 1-0 game is hit in the 1st or 9th” thing…This sabr-newb is struggling with the concept. It’s a game theory axiom that, all things equal, it’s better to reveal information as late as possible. ‘Revealing’ a run late in a 0-0 game gives the losing team fewer chances to make optimal offense-for-defense subs than if the run had scored earlier. In the long term, you wouldn’t have let your starter go 9 innings for a complete game shutout if you knew he’d lose on a walkoff solo HR anyway, so by scoring later, I’ve caused you to put more miles on you’re starter’s arm, a potential advantage later in the season/series. What am I overlooking here?
That’s a point of view I haven’t quite heard before. How would you go about quantifying the impact/advantage of a late event in a game.
My interpretation is that (hopefully) managers manage according to the probabilities of the time of the game. They (hopefully) wouldn’t leave a starter who’s getting tired in a tie game and they (hopefully) take the long view on arm fatigue regardless of the situation. So I have a hard time seeing how your point could be incorporated into a metric.
Even assuming optimal managerial decision-making (usually what one ought to assume), we can still only make decisions based on what’s occurred. If my guy hits a solo HR to go up 1-0 in the sixth, you pinch hit for your pitcher in the top of the seventh. If I wait to hit that homer until the bottom of the seventh, you would have left him in. Both decisions are optimal based on available info, but the late homer gave you one less quality AB. No idea how this could be accounted for, or if things like WPA reflect its relevant aspects. Just wanted to see if the case for “Later=better” holds up.
If your taste for “value” falls at this point on the contextuality scale, that’s reasonable, but you need to lay out more of a case for why this specific point if you want people to buy it. We have s lot of options and you’re selling a new one!
If you’re willing to use final game score margin, why not go farther and count only when as player’s run contribution tipped the game from a loss to a win?After all, if a player creates some runs in a game his team loses by one, it didn’t matter in the wins column. Where’s the value in that?
Where does GARE24 fall relative to WPA/LI in contextuality, actually? More or less or both?
GARE24 surely isn’t value until you include positional value!
Your intro doesn’t seem quite fair. Sure, if you knew “will excel” prospectively you’d pick McClutchy, but “will happen” is a much stronger claim than “did happen”. If I can predict the future like that, give me the guy who hits/pitches to the score, just enough to win.
Hi Jetsy. Regarding your questions…
I guess you could include only certain hits in a metric and not others, but I’d rather contextualize all hits to see where they come out. I want to value a batter based on all of his performance rather than just some of it.
GARE24 is an entirely different beast from WPA/LI. I don’t think you can directly compare the context of the two. As I said above, you use WPA/LI when you want each situation to be considered equal. What I’m doing (and what WPA does) emphasizes certain situations and outcomes more than others.
You’re right that I didn’t include positional value (or fielding, for that matter). I said so in the article.
I fashioned my intro to make a point. If something happened, and you can establish a clear tie between what happened and a team winning through a reasonable simulation process, then I believe the thing that happened ought to be included in MVP consideration.
Well, I totally forgot that Baseball Reference publishes a “deleveraged” version of RE24. That is, it normalizes RE24 so that batters aren’t penalized if they bat in relatively fewer potent situations (for instance, coming to bat with the bases empty and two out more often than other batters). They call it RE24/boLI (for “base out Leverage Index”)
And according to Baseball Reference, the major league leader in RE24/boLI is one Mr Bryce Harper…and by a good margin. His RE24/boLI is 80. Votto is second at 65. I didn’t know that boLI could have such a big impact on RE24, but I trust that these tables are solid.
(Footnote: here’s another good article on the subject by Colin Wyers: http://www.baseballprospectus.com/article.php?articleid=21727)
So now we’re all happy. The system I’m advocating selects the guy that makes readers least angry. Between this and changing the Game Margin factor to include wins and losses, I’ve got two good next steps to improve this basic idea. Thanks for the feedback!
When you do WPA/LI or you do RE24/boLI, what you are saying is: every PA is equally valuable. That’s the intent there.
What those metrics are doing are “recentering” each event, so their impact is measured relative to their exact situation. So, a walk with bases loaded bottom of 9th tie game, that has the EQUAL impact to a HR in that situation. WPA gives us that.
And WPA/LI will make sure that that PA has no more weight than a PA in a blowout.
That’s as succinct as I can put it.
Hmm. That’s a good point. I’m not sure that I want each plate appearance to be weighted equally–some are indeed more important than others–but I do want to adjust if there is an inequity in opportunities between players.
Given that’s my goal, would a better approach be to take total RE24 and divide by the average boLI across all PA’s?
I don’t know that that’s a better approach. Two players hit identically, but one always with bases empty (LI = 0.7) and the other always with men on base (LI = 1.4).
The first guy has an RE24 of +21 runs and the second guy has an RE24 of +42 runs. Do we necessarily want to adjust them both to +30 runs? That’s what RE24/boLI would give you.
In your example:
PlayerA is +21 runs with bases empty and 0 runs with men on base, and faces both situation 50% of the time. LI = 1, and RE24 = +21.
PlayerB is +42 runs with men on base and 0 runs with bases empty, and faces both situation 50% of the time. LI = 1, and RE24 = +42.
So, which do you want? I don’t know, but it helps if you lay out various illustrations.
Thanks, Tango. Laying out the math with examples is one of your real strengths.
I’m pretty clear in my mind that I want the second version, the 21 and the 42. The two players had a similar set of opportunities (as measured by average LI) yet the second guy delivered with men on base. This leads directly to runs, which leads directly to wins, which is value in my book.
Let’s take your example, but say that Player B comes to bat 55% of the time with bases empty and 45% with men on base. At the same rates that you describe (not exactly but close to), his total RE24 declines to around 38 and his boLI also declines. Dividing one by the other gets them back to 42 vs. 21.
FYI for people interested, each batter’s average boLI is listed at Baseball Reference (something else I didn’t know until now!).
Ok, so let’s use the player who averages +.03 runs per PA, when the LI = 0.7 (bases empty) and +0 runs per PA when the LI = 1.4 (runners on base). We have these guys:
Player PA_empty PA_runners LI Runs Runs/LI
A 700 000 0.70 21 30.0
B 600 100 0.80 18 22.5
C 500 200 0.90 15 16.7
D 400 300 1.00 12 12.0
E 300 400 1.10 9 8.2
F 200 500 1.20 6 5.0
G 100 600 1.30 3 2.3
H 000 700 1.40 0 0.0
So, what happens is that the better you do with bases empty AND the more you come up with the bases empty, the better your runs/LI metric will come out.
Compare the C and E players:
C 500 200 0.90 15 16.7
E 300 400 1.10 9 8.2
Both were +.03 runs per PA with bases empty and 0 runs per PA with runners on base. But, when you are trying to standardize their opportunities, player C goes up and player E goes down. Even though if you really tried to standardize, they’d be equals.
What’s really happening is that you are “pro-rating” the player’s overall performance in the “missing” PA. If a guy didn’t get much chance to perform with men on base, it gets pro-rated using a lot of his bases empty performance to fill in his men on base performance.
Right. Without equal evidence of the player’s performance with men on base (which is only available at LI of 1), you’re pro-rating evidence from the bigger sample onto the other PA’s. Your extreme example notwithstanding, that sounds like a reasonable thing to me. Do you have any other approaches that would address the twin goals of treating PA’s according to their importance while also acknowledging differences in opportunities between players?
thanks for example, by the way. Very helpful.
Studes: I don’t know that I have a good solution really. Your point is valid that once we have the totals as presented (runs at +15 and LI at 0.9), we lose how we got to there. And so we have to assume that the poor guy would have done better had he had better opportunities (even though we know, in this case, he pro-rates to worse).
Really, this discussion should generate two questions for every answer someone offers. There are many layers to consider.
The leverage argument is poor logic that I see in many sports groups on the topic of value. Yes, getting hits with runners in scoring position is more valuble *on the margins*, but the margins exist *because of the hits that were provided by those runners in scoring position*! I don’t give hitters extra value or credit for piggybacking off his teammate’s success.
If you feel like giving Rizzo credit for leverage situations that did not even *cause*, then have at it. But, if Harper and Rizzo play on the same team as the 3 and 4 hitters in the lineup, Harper gets his double with no runners on, Rizzo gets his double with the guy *who is the cause of Rizzo’s run-scoring opportunity* on second, and you want to give Rizzo an award for that; then you’re cherry-picking what you want for your value.
RE24 does give credit to the batter who got on base before Rizzo. What’s more, it deletes the expected value of the base/out situation from Rizzo’s total. RE24 only credits added value beyond that expected.
If people are interested in player contributions to theoretical wins instead of actual wins, then what I’m suggesting won’t work. If they’re interested in measuring contributions to actual wins, then they might be interested in this approach.
And how does RE24 credit added value beyond that expected? Can you provide an example?
Here the link provided in the article. It explains all about RE24 (called Value Added Runs in the article):
http://www.retrosheet.org/Research/RuaneT/valueadd_art.htm
Thanks for the link.
The proposed methodology doesn’t solve the issue, however. If my math is correct; in the example I provided, Rizzo is still credited with more value than Harper, even after the expected value is deleted. This is logically incoherent.
Part of that extra credit comes from giving hitters with runners in scoring position credit for scoring a run. And that goes back to the same fundamental issue I brought up: why should he get that credit when he didn’t put that runner on base?
Analyzing the game on the margins is useful…for analyzing the game on the margins. But the margins themselves presuppose the events that cause those margins in the first place.
It’s not logically incoherent. It’s perfectly logical. You just don’t like the logic.
Then how about *explaining* it then, instead of making a vacuous post?
At least Mr. Studeman attempted to explain the methodlogy and provided a link (even though I pointed out issues with the methodology).
Ah. You’re Mr. Studeman, using a different alias.
The reply that you gave wasn’t valid.
Oh, it was perfectly valid. The inherent logic is laid out in Tom’s article. You don’t need me to further describe it to you. You just don’t like the implications of the logic.
I pointed out particular problems with the logic in the article as it pertains to atttributing overall value to a hitter, and you didn’t even attempt to counter them. Instead, you parrot the same, intellectually lazy response.
Yes, yes, I’m so intellectually lazy. Woe is me!
Sarcastic wit isn’t your strong suit, either. I’ll field the question to a baseball guru/baseball stats board interested in honest debate.
It’s true that sarcasm isn’t my strong suit. What do you do when “honestly debating” with a blank wall?
Didn’t see this reply, but you weren’t debating. I started off by talking about issues with the stat, then you provided a clarification and a link to th logic, then after I read the link and talked about issues with the logic in the link you provided, you reply in a childish manner by failing to demonstrate how the provided infromation is “perfectly logical” (which is the same logic that’s under dispute here) and levying ad hominem attacks. I expect better from a writer of your stature.
No offense, but your condescending attitude in these posts received the reply they deserved. You use insulting words such as “logically incoherent” and “intellectually lazy” and “poor logic.” None of these adjectives were appropriate for this particular discussion. You started off with a premise–that a batter shouldn’t receive extra value for an event with higher run-scoring potential–and judged everything in that light. If you start with a certain value premise and don’t move from it, there is no room for an honest debate. That is what a troll does.
No offense taken; even if you incorrectly labeled me a troll. However, if you’re taking offense to posters attacking the logic of your arguments — which is all I did here; I even called your responses intellectually lazy, not you — then, respectfully speaking Mr. Studeman, this isn’t the business foryou.
Now someone tells me!
Studes — great to see another article by you — I’ve printed out maybe a dozen sabermetric articles over the years — ones that that I like to go back to — and you’ve written maybe half of them.
Still, I think RE24 is better than Game RE24. If Harper comes up with the bases loaded and two out, and hits a grand slam, I’m not sure why he should get less credit if the Nationals lost 10-4 than if they lost 5-4. He did all he could do in either situation. He shouldn’t be penalized if, say, Drew Storen got bombed.
I can see why WPA (or WPA/LI) captures something additional. But GameRE24 just seems like an arbitrary way to bridge the gap – kind of like Molly Ivins’ dead armadillo.
Hey Craig, thanks for checking back and thanks for the kind words.
I don’t believe that the Game Margin adjustment is arbitrary. Close games have higher average LI’s, to use WPA terms, and I basically apply a game LI to all the plate appearances instead of just some. To me, it’s obvious that events in a close game have a bigger impact on who won the game, compared to the same events in a blowout. The average LI from those games bears that out.
I agree that this results in some distortions when you look at specific plate appearances, and I’m open to other ideas. But I think the “Game LI” factor is an important situational factor to capture in the equation.
I agree with Craig’s line of reasoning here. It doesn’t make sense to me that more credit should be given to the 5-4 game versus the 10-4 game.
If the 10 runs vs. 5 runs is beyond the players control, then it’s unfair. As another commenter said, just by being on a playoff team, each of those games should be more valuable too. Where do you stop with this?
I think that situational hitting is important, and it should be factored in. But it’s more of a bonus, and is less reliable.
Another way of looking at this. If you put Harper in Rizzo’s place for all of Rizzo’s at bats, who would create more wins?
I think we should lean towards Harper.
Also, have we considered the sample size of the different situations?
Dave – So in last night’s Nationals game where Harper went 3 for 3 with 2 home runs and his team wins 4-0 his RE24 would get heavily devalued even though he provide the entire margin of victory? I think you should give up on GameRE24 and just stick with RE24. At the very least you should use run differential – player’s game RE24, but you would still have the problem that a player could still be penalized because his pitching and defense gave up to few runs.
I don’t see any reason to give up on anything. Why can’t we just let the discussion move toward there?
Peter has a good example. Say you have one player that provides all the runs, say 4 solo HR in a game won 4-0.
Now, say you have another player also with 4 solo HR, but that game was won 4-3, with all 3 runs allowed in the 9th.
You have a third player, 4 solo HR, also won the game 4-3, but the defense gave up the three runs in the 1st.
How do we handle these three situations? What is the end result that we want, and can we get there in a consistent fashion for these three players?
Heck, let me add a fourth one, the same guy gets 4 solo HR, but his team loses 5-4.
thanks for the scenarios, Tango (and for sticking up for the idea. You too, tz). My goal here is to essentially develop a Game-Level Leverage Index, so that individual game performances can be adjusted based on the final outcome of the game.
So, to me, I don’t care about *when* the events occur (early or late in the game). Some of the commenters have an issue with the “ex post facto” nature of this, but I’m okay with it. You can’t argue that “a first-inning home run in a one-run game is just as important as a ninth-inning home run in a one-run game” without being willing to live with the consequences.
I like your idea of separating the Game Factor into wins and losses, as I think that will marginally address some of the concerns that have been raised. And I think there’s something to Peter’s idea of taking the batter’s RE24 into account re: the game margin but that’s a complicated one. Got to think it through.
Nothing we do here will be perfect and everyone will have their own tolerance level for the imperfections. The question to me is whether it gets us nearer than anything else to the ideal expressed at the beginning of the article.
Since there is no such thing as “clutch hitting ability” separate from “hitting ability”, why are you giving a hitter extra credit because he happens to get a disproportionate number of his hits with runners on one year? Why don’t you give the extra credit to his teammates who happened to get on base before he got a hit? Both are random.
No offense, but I think I covered this question pretty well in the article. We are talking about the MVP here, not the person most likely to repeat his performance. Wins are the result of not just hitting, but hitting well in particular situations. I am proposing that Value (as in MVP) is that thing that most drives winning games, and situational hitting is part of that.
RE24 DOES give credit to hitters who got on base before the batter.
RE24 gives credit to runners who reach base, but they get the same credit regardless of what subsequent hitters do. For example, if a batter hits a one-out double, he gets a certain run value whether he scores later or is stranded. But if there happens to be a runner on base when he hits the double, he gets additional credit, although it is completely random that there was a runner on.
Yes, wins are the result of hitting well in certain situations, but the credit doesn’t all belong to the batter who is last in the scoring sequence. The scoring is the result of the interaction of what the batter did with what his teammates did. What you are doing here is the equivalent of saying whoever leads the league in RBIs is the best hitter, since he got his hits “when it matters”.
The events may have been random, but they mattered. I’m proposing that what matters is what should be taken into account when deciding the MVP award. I’m trying to re-capture the narrative from our stultifying numbers.
RE24 credits (and debits) batters for getting hits (and not getting hits) when they matter (or don’t matter) in a much fairer way than RBIs (or Runs Scored) do. You can normalize the opportunities between batters, too.
I suppose what you’re suggesting is a variation on RE24 that gives more/less credit to batters based on what happens after their plate appearance. A retrospective credit/debit based on what subsequent batters did (such as driving them in). How would you do that in a way that wouldn’t devolve into Runs Scored?
So I thought about it a bit and came up with this. Use RE24, but when one or more run(s) score(s), you debit/credit the run scorers and the RBI man so that their total contributions are equal.
Here’s an example:
No one on, no one out: 0.5 runs expected
Single: 0.9 runs expected, 0.4 RE24
Single, runner goes to third: 1.75 runs expected, 0.85 RE24
Double scores both runners: 1.15. runs expected, 2 runs scored, 1.4 RE24
You then take all three participants’ total RE24 (2.65) and divide by three (0.88) and replace their initial RE24’s with this new figure. So the first guy who singled has his RE24 more than doubled based on what the batters after him did (0.4 to 0.88). The RBI man loses more than a third of his RE24 but still has a bit more than he would have had with the bases empty (0.88 vs. 0.65). If he is subsequently batted in by a batter, his RE24 goes up. If not, it stays at 0.88.
Would something like this work better for you? Has anyone proposed it before?
BTW, Cliff, Peter left a specific reply for you in the comment below this one. Don’t know if you saw it.
Tango – In each of your four cases the player has a run value added of 4 runs for hie 4 home runs during the game. I think the better question to ask is ” Is there a compelling reason for the player’s game value to be any different than that. In the 17 years that I played baseball through college and later in adult leagues I can’t remember a single plate appearance that I went into where I didn’t want to win that confrontation with the pitcher by not making an out no matter what the score was. I am not saying that I wasn’t aware of the score and that certain baseout situations and score differentials didn’t require a slightly different strategy , but for me it was almost always take a look at where the defense was playing me and try and hit a hard line drive where they weren’t. And I think that almost all players that I played with felt the same way. I guess that is why I feel strongly that game score should play no part in evaluating what a batter accomplishes in his at bat.
And to answer Cliff above it is also why I feel strongly that it is important to take the baseout situation into account by using run value added (RE24). A player is placed in a certain position in the batting order by his coach knowing that in that position he is likely to bat in certain baseout situations more than others. It is his responsibility to his team to make the most of the situation in which he is placed. He knows that if he is batting third, fourth, or fifth he will be batting more times than average with men on base an men in scoring position. But Run value added balances that extra opportunity by placing a higher initial expected value on those baseout situations and the batter must still perform better than average to get a positive run value added for his PA. Batters that can consistently do that will have run value added totals higher than their linear weights and batters that don’t will have linear weights lower than their run value added.
I think there’s a case to say that run value is the metric we should use to evaluate player performance – I happen to be in the camp for the most part, and so RE24 is excellent for that (or RE24/boLI if you want to normalize plate appearances).
But for those who want to use win value as a basis to evaluate player performance, we’re left with WPA, which depends very heavily on the sequence of game events, or WPA/LI, which normalizes out all the leverage from WPA. I think what Dave is proposing here is a happy medium between these two metrics, in that it reflects the base/out situational value added (from RE24) but scales it to be more reflective of win value. So if your preference is win value, the GameRE24 metric appears to strike that middle ground.
Peter:
Your answer is clear that all you care about is runs, irrespective of the impact or timing of those runs to the game at hand. Given that, then it’s clear that RE24 is what you should use.
If someone wants to argue that the impact or timing of those runs matters, then it’s certainly a valid position for that person to take.
Hmm. ” My opinion is that you should eliminate all unrepeatable and luck-based skills, retain repeatable skills only.” I don’t know that I agree with that. The value a player creates resides in what he actually accomplishes, not in what he should have accomplished given his skill set.
For example, say Jason Kipnis goes nuts with the bat next year, and hits .402 with 55 HR, leading the Indians to a WS title. He’d almost certainly win the MVP, even thoughI think we all would agree that’s a HUGE fluke, and not repeatable. But Cleveland would still have that championship even if he then hits .275 with 15 homers in 2017, and that title has value.
Sure, if you’re evaluating trading for a player or signing him to a contract, you want to look at repeatable skills and not luck, but when it comes to giving out awards, I’m inclined to look more at what actually happened.
I am a big fan of the RE24 stat. I think it’s a great “RBI-like” indication of production. Like all stats it has it’s strengths and weaknesses. I am NOT a fan of correcting it for “leverage”. Yes, some guys get lots more opportunities than others. That somewhat misses the point of the stat though.
Lets say you put 8 terrible hitters around a pure league average hitter. What would you expect his RE24 to be? If you said something other than zero, you’re wrong. His leverage is very low which multiplies his production… but his production is average. Similarly if you put our pure league average hitter in a lineup with 8 studs, his RE24 will still be, you guessed it, zero.
Now, that’s not to say leverage has no effect. If we put a GOOD hitter in a lineup with 8 scrubs he’ll put up a positive RE24 but it won’t be as much as you would expect given the GOOD hitter’s stats. Similarly, a BAD hitter’s RE24 will look better than you may think it should on our 8 scrub team. That seems to indicate leverage compensation but wait! 8 scrubs need the help more than 8 studs and the offensive contributions of our hypothetical GOOD or BAD player are not the same towards a W or L on the 8 scrub or 8 stud teams! Multiplying by leverage ruins RE24.
The best way, I think, is to look at a triple crown of production (stop me if you’ve heard this before) and focus on the individual’s contributions on a RATE basis, on a PRODUCTION basis, and on a RESULTS basis. Basically OPS+, RE24 (NOT de-leveraged), and WPA. I think many people on here are absolutely right saying correcting RE24 is really trying to get back to WPA for no reason. I would NOT like to see RE24 and WPA have an effect on the hitter’s batting contribution of WAR but if you want to talk about stats that mattered to a team’s success or failure, I think they’re both important, and both measure correctly what they’re getting at.
Thanks. I understand what you’re saying about de-leveraging RE24, but our goal here is to judge individual players who have had different types of opportunities presented to them. So if we lose some of the accuracy of RE24 in the process, the goal of being able to better compare players seems worth it to me.
It’s kind of pathetic how hard you’re trying (and failing) to devalue Harper’s epic season. It’s even more pathetic how hard you’re trying to prop Rizzo up.
Darn! Because that is so what I was trying to do.
Reading through the article, that is how it felt.
It seemed as if you weren’t going to stop until you had something that showed that Harper shouldn’t be the MVP, or at least that he shouldn’t be a slam-dunk MVP. While that may not have been your intention, it is how it came across (at least to me).
If I had really wanted to devalue Harper’s season, I would have just quoted his WPA and then sat down. This article wasn’t *about* Harper. He was the example I used to explore something I’ve been thinking about for seven years.
a eskpert said… that making solid contact in those vital situations is more indicative than the outcome. Yes and no. The guy who has the kind of bat control to find the holes should not be penalized for it. The guy who pounds the ball into the shift and makes outs consistently should not be rewarded for merely making contact when the defense can play him successfully. The guy who is in between is in between. That is about the ground balls and line drives. Fly balls are different, of course. There is not much defense against a home run, and a clutch sacrifice fly is also worth something. But again, the guy who hits the fly balls that end up easy outs without advancing runners should not be rewarded.
I live near the DC area now so I see a lot of Nats games, and since moving here I have noticed how little Bryce hits when there are guys on base but just thought maybe it was just me.
With that said, We should also try and look into how your lineup effects your situational hitting. Because Bryce hasn’t had a real consistent lineup around him or anyone besides Escobar hitting well most of the season, this could have an effect as well.
Maybe being the only real threat has made him press more in run scoring situations, not to mention the possibility of pitchers approaching him differently as well with men on. Granted I do believe that some guys are just simply better in certain situations, I don’t believe this is something that should be ignored.
Just a thought, I am a Mets fan so I have no reason to defend Bryce lol.
Emmanuel, did you not notice that Harper’s OPS in RISP situations was 1.035? YMMV, but that doesn’t look to me like a guy who’s “pressing” and hitting “little” with men on base.
Studes, fascinating work. Both your math skillz and patience with comments are far beyond what I could manage. [I don’t think the commenter who called your argument “logically incoherent” was trying to be insulting, but I know I too would have felt insulted by the remark, and would likely have reacted much less pleasantly than you did!]
Thanks, Michael. I agree he wasn’t trying to be insulting. I actually don’t mind straight insults, cause then you know where the commenter is coming from. But I have a strong dislike of people who pretend to be offering logical arguments, hiding behind condescending words like “logically incoherent” and “honest debate,” when they are just using their a priori beliefs to justify anything they say. People like that hurt the level of Internet discourse everywhere, so I tend to react against them pretty strongly. Character flaw of mine.
First, thats alot of nonsense coming from the both of you. Part of what hurts Internet discussion is when posters take needless offense to an argument they present as if they’re the ones under fire. No one was personally attacking you, Mr. Studeman, so stop acting like a victim. If you thought that you were correct, all you had to do was show why, instead of derailing the conversation in the manner that you did. That’s it.
Second, no; what I said was not an “a priori” belief. It’s a simple, fundamental principle of reason, one that is absolutely true. Just looking at the game at the margins has that blindspot that I wrote about, and it affects how we figure out player value.
Thanks Mr. Studeman, this is exactly the tweak to WPA I was looking for in the MVP discussion. People discount RBI (for valid reasons) but it did describe an actual even that had value.
I think some of the pushback is that a lot of sabre discussion is designed specifically to go beyond actual in game value provided and see how good a player is. Looking at a pitcher that has given up a ridiculously high babip indicates they are better than their boxscore number would indicate. But, it also means they’re providing less value than you’d guess by their talent.
Thanks, Zach. That is exactly my perspective, too.
People make the same mistake all the time. RBI is not a useless stat, but it’s utterly useless *for evaluating talent*.
At any rate, the argument has been had a thousand times. You take Slugger because he’s better. The better player is the goal. Situations involve ifs ands and buts. Ability doesn’t. The minute you make things more difficult than they actually are is the minute you have done poorly as a team architect. Even if the purpose is selecting between the two you want to keep it simple. Saying so and so is valuable *if* this or that is a major red flag.
You’re talking future performance. And you take Slugger only because we’re all fairly convinced that clutch by and large isn’t a skill. There is no such real person as Eddie Situation.
What this comes down to is a HUGE semantic issue – are you valuing story or talent? Just because something “should” happen doesn’t mean it will, and it’s possible that despite Harper’s better raw output, it just so happened that Votto by one measure was more integral to winning games than Harper this season.
So, I braved the comments section, what a great debate! And I have another philosophical wrinkle for all the great minds here. I want to talk about the runner in the run – the guy who actually scored.
The discussion so far has revolved almost exclusively around the person who drove the run in – the batter. And there’s an issue with this that was brought up – it’s an opportunistic occurrence.
So here’s a question going back to core of “Value = Winning”. The reason situational hitting is opportunistic is BECAUSE OF ANOTHER HITTER.
Going back to old stats, I’d like to argue that Run Scored could be “valued” more. Part of crediting Rizzo for driving in a run is crediting his teammate for being on base in the first base.
So – here’s the stats question – is there a stat out that records situational runs scored? Wouldn’t that be fascinating? Who’s managed to get on base and then get driven in the most in the most situationally important games?
And yes – there’s a very big and nasty wrench with this – the home run. That most effective of run scoring techniques where you do all work and drive your own darn self in. And that makes matters so deliciously more complex.
Thanks, Mac. I think you understand this, but RE24 bridges the breakout between Runs Scored and RBI’s. You can come at it from either angle.
As to your specific questions, I don’t know of any stat like that. 15 years ago, my very first Internet baseball article was a measure of how often batters got themselves into scoring position, but that’s only part of what you’re talking about.
It looks like the comments have dried up, so I thought I’d add a little post mortem. First of all, I’d like to thank everyone for their terrific feedback and comments.
One person asked how often the difference between wRAA and RE24 is an issue. I reviewed league leaders in the two stats in the 1980’s and the short answer is: about half the time. For one example, Wade Boggs led the 1988 AL in wRAA at 63, but only had 45 RE24. OTOH, George Brett had 41 wRAA, far behind Boggs’ 63, but racked up a league-leading 54 RE24, handily ahead of Boggs’ 45. The voted MVP, Joe Canseco, was a close second in both categories.
So I am sticking to my proposition that situational hitting, as embodied by RE24, ought to be a primary consideration–not secondary consideration–in MVP voting. Based on the comments, I have some work to do with the close game adjustment.
And Bryce Harper has had a pretty good week and now leads the National League in RE24 by a good margin, 78 to Joey Votto’s 68. So, yeah, Bryce Harper for MVP! At least, with two weeks to go…
This is obviously late, but this discussion did an excellent job of seeing how stats come from different perspectives. It took a little bit of reading before I understood that Dave was aiming for the entire game, ex post facto, as the proper contextual perspective for value. There is a good argument here.
However, with all of the confusion over the topic, I wonder if the idea of “context neutral” stats drives part of the confusion. WAR, obviously, considers the batter apart from the context of people on base or game situation. This is not properly context “neutral” but a particular, chosen context. To be clear, this one context makes sense, and WAR is an excellent measure, the best comprehensive statistic I’ve seen yet. The term “context neutral” suggests, however, that any other context–the context of the entire season or the game or the particular people on base–is wrong rather than simply different. It might be better to call it a “context indifferent” stat rather than a “context neutral” one. Neutral makes it sound as if there is an Archimedian point on which all statistics must stand. Indifferent attempts to just say that the statistic is indifferent to the particularities of in-game situation, game score, or seasonal standings.
It was fun to think about the presuppositions of different statistics. Thanks, Dave.
Why do we give some players extra credit for what the players who came before them did, but not give extra credit to other players for what those who came after them did?
The guy who walked to lead off the 9th inning in a close game is just as responsible for the win as the guy who sacrificed him in later in the inning. Maybe more so, if someone made an out in between. You’re only giving credit to half the responsible players.
Oh my. I feel like we covered this pretty well in the comments, as well as in the attached articles. But let me summarize how RE24 works.
Each batter is given credit for his contribution based on an expectation of what will happen after him. The expectation is set at average. In other words, it’s not true that RE24 only gives credit to half of the responsible players. It gives credit to all players, based on average expectations of what might come after. The guy who walked to lead off the ninth is given credit, based on assumption that league-average events will happen subsequently.
If and when a subsequent batter exceeds those expectations, he is given credit for doing so. The batter who walked is not given extra credit if the subsequent batter exceeds expectations, but why should he receive credit for something a subsequent batter did? This also works in reverse, by the way. If a subsequent batter makes an out, the one who walked is not debited for that. The system you’re suggesting would work both ways, I assume, by demeaning the value of that walk if subsequent batter all make outs.
Personally, I think this is the fairest system for doling out credit for specific runs scored. Other people disagree, I know, but I haven’t heard of a better system that matches actual runs scored per game.
But why use an average later at-bat to judge the value of the walk instead of what actually happened? From earlier parts of the discussion, it looks like this analysis is all post-hoc anyway (we think what’s important is what wound up happening, not the players’ ability to make it happen consistently on any basis going forward). That walk turned out to be really important to winning the game, just like the later flyout turned out to be really important. Shouldn’t the leadoff hitter get credit for the R just like the batter gets credit for the RBI?
Trying to crystallize this thought… if what we’re trying to do is to quantify the actual value of an event to the team given what actually happened in the game, why don’t we go all the way? This stat basically says a table-setter’s value is whatever we would have expected regardless of the table-cleaner, while a table-cleaner’s gets higher if the table-setter did a better job.
That only makes sense if we’re trying to quantify the “grit” or “clutchness” of RBI guys (only) or something similar. It’s about mental state, not value. If we want the actual historic value of the event, let’s apply this same idea to *all* events and include actual performance of players before and after.