What is WAR?
Wins Above Replacement (WAR) is an attempt by the sabermetric baseball community to summarize a player’s total contributions to their team in one statistic. You should always use more than one metric at a time when evaluating players, but WAR is pretty darn all-inclusive and provides a handy reference point. WAR basically looks at a player and asks the question, “If this player got injured and their team had to replace them with a minor leaguer or someone from their bench, how much value would the team be losing?” This value is expressed in a wins format, so we could say that Player X is worth +6.3 wins to their team while Player Y is only worth +3.5 wins.
Calculating WAR is simpler than you’d think. If you want the detailed (yet very understandable) version, check out the links at the bottom of the page; Dave Cameron does a good job of walking through the process step-by-step. The short answer, though, is that as follows:
● Offensive players – Take wRAA, UBR, and UZR (which express offensive, base running, and defensive value in runs above average) and add them together. Add in a positional adjustment, since some positions are tougher to play than others, and then convert the numbers so that they’re not based on league average, but on replacement level (which is the value a team would lose if they had to replace that player with a “replacement” player – a minor leaguer or someone from the waiver wire). Convert the run value to wins (10 runs = 1 win) and voila, finished!
● Pitchers – Where offensive WAR used wRAA and UZR, pitching WAR uses FIP. Based on how many innings a pitcher threw, FIP is turned into runs form, converted to represent value above replacement level, and is then converted from runs to wins.
WAR is available in two places: FanGraphs (fWAR) and Baseball-Reference (rWAR). Both statistics use the same framework, but are calculated slightly differently and therefore sometimes show different results. The above explanation is for fWAR; see the section below on rWAR for more information on the differences between the two iterations of WAR.
Context:
League-average WAR rates vary. An average full-time position player is worth +2 WAR, while average bench players contribute much less (typically less than +1 WAR). Average starting pitchers also are worth around +2 WAR, while relief pitchers are considered superb if they crack +1 WAR.
For position players and starting pitchers, here is a good rule-of-thumb chart:
| Scrub | 0-1 WAR |
| Role Player | 1-2 WAR |
| Solid Starter | 2-3 WAR |
| Good Player | 3-4 WAR |
| All-Star | 4-5 WAR |
| Superstar | 5-6 WAR |
| MVP | 6+ WAR |
Also, here’s a breakdown of all the players in baseball in 2010, courtesy of Justin Bopp from Beyond the Boxscore.
Things to Remember:
● Since there is no UZR data for catchers, the fielding component for catcher fWAR is calculated using two parts: the Stolen Base Runs Saved (rSB) metric from the Fielding Bible, and Runs saved from Passed Pitches (RPP). This accounts for a large portion of a catcher’s value, although pitch framing is not yet included in WAR.
● WAR is context, league, and park neutral. This means you can use WAR to compare players between years, leagues, and teams.
● It is possible to have a negative WAR. In fact, the worst fWAR any player has had since 2002 is Neifi Perez from the Royals, who posted an incredible -3.1 wins in 2002.
Links for Further Reading:
Intro to WAR – Big League Stew
Background on WAR – Offense
- Part 1 – Batting
- Part 2 – Fielding
- Part 3 – Positional
- Part 4 – Replacement
- Part 5 – Converting Runs to Wins
- Part 6 – Dollars
- Part 7 – Additional Info.
- Part 8 – Team Context
Background on WAR – Pitching
- Part 1 – Introduction
- Part 2 – FIP
- Part 3 – Replacement
- Part 4 – Run Environments
- Part 5 – Converting Runs to Wins
- Part 6 – Park Adjustments
- Part 7 – Calculations
Common Misconceptions – The Book Blog
4
What bull !
This is just another attempt by stat geeks to take away the one thing that matters more than anything and try to make someone seem better than they are , or some others appear lesser .
When you play the game for money , winning is the only thing that matters .
If that’s not a direct quote of Leo Durocher in “Nice Guys Finish Last” , it’s close .
The problem with stat freaks is that they can make the numbers say what they want them to say .
I see people on TV who supposedly know something saying Mariano Rivera is the greatest pitcher of all time .
Really ?
What an insult to pitchers of the bygone era who carried their teams like Sandy Koufax and Bob Feller by pitching 9 innings and making 40 starts plus some relief !
And they’re just the tip of the iceberg .
I read that Jack Morris probably won’t make the Hall of Fame because his WAR isn’t high enough .
How many of the pitchers who have a higher WAR were the aces on three World Series winning teams ?
Since when has the result on the field stopped mattering ?
It’s a made up stat that shouldn’t have anything to do with anything .
The numbers often DO lie .
It’s as simple as that .
John Ogrin
The rules for determining when a pitcher is credited with a “win” is a “made up stat,” so perhaps winning shouldn’t have anything to do with anything, by your logic. Baseball has ALWAYS been a game of stats. The issue here is that this is a stat that doesn’t fit your personal needs, which is fine. But understand that those of us who enjoy the statistical analysis of baseball are trying to gain a greater understanding of “the result on the field.”
It’s not like these stats are being taken from a video game. They are descriptive of what is really happening out there. If you build a team with high WAR players, guess what, you’re going to win!
Then show us, empirically, how the numbers are wrong. You don’t have to make up a stat, but to disprove empiricism, you need to show your work.
You can’t argue with Jon Heyman masquerading as John Ogrin.
You know really all statistics are made up.
I could go through every part of your comment but the others said it very well and I just don’t think it really deserves anything more than what I offered above.
Statistics are all made up, this is true. However, outside of the baseball universe the most important thing in statistics is an understanding that stats are only estimates of true parameters. What makes statistics useful is knowledge of just how reliable or unreliable a particular variable is at describing reality. This means you must quantify uncertainty. I haven’t seen any information about calculating uncertainty for WAR or most of the other sabremetrics for that matter. Without uncertainty data any stat is pretty much useless.
LOL wat? this makes absolutely no sense
very good I think, But I really wanted a stat that puts defense contribution into batting statistics
Just stumbled upon this by way of another article, which explained that wraa has it’s roots in this… so cool!
Rating pitchers is even trickier than rating batters. Clemens has been called the greatest pitcher since ww2, but he never played for a bad team. Kaat and Blyleven played for quite a few bad teams.( Kaat more than Blulevn). They finished about 70 wins or 3 a year behiond Clemens, don’t even try to tell me that the teams they played on didn’t cost them that many win.s I used to hear that a great player means about 3 wqins a year over an average player. Seems about right to me. WAR seems better suited for batters than pitchers. Personaly, O try to take a starting pitcher’s win lost % out of his team and then compare it to his team without him Guys Like Walter Johnson, Carlton, and the 47 year old miracle, now hurt in Philly, look pretty dran good when you do that. As for hitters, what if he is replaced by a majot league regular ubnstead of a minor leaguer? Kind of blows WAR, in that insrance. But, as you say, use mpre than 1 method of evaluation.
If you add up the WAR for all the players on a team, does it add up to the teams number of wins? If not, is that possible?
It does, to a degree. Add up all the WAR for a team and then add an adjustment of around 43.5 WAR and that should do it.
Steve,
I just looked up Jim Rice on Fangraphs and Baseball-Reference, and the differences are significant: 41.5 rWAR compared to 56.1 fWAR. There are slight differences, even seasonally, in every column, from fielding (expected) to positional, replacement, and batting (unexpected to me). Can you explain or provide a link to the differences in how the calculations are made?
Thanks,
Ben
I’ve included lots of links in the article above — I recommend checking them out, as they’re my main resources and should answer all your questions.
In particular, I recommend checking out the multi-part WAR series provided in the ‘Links” section, as well as this piece by Sean Smith (linked to in the rWAR/fWAR section): http://www.baseball-reference.com/blog/archives/6063
That should do the trick – best of luck!
WAR doesn’t seem to be an accurate assessment of a player because it is comparative to their backup. So if a player’s minor league backup is very good statistically, it will bring his WAR down. But another player of the same position on a different team has a backup who is awful statistically so his WAR goes up, and now he seems like the better ML player, even though he probably isn’t.
@vj, that’s not true. Replacement level is a universal level for each league. I’m pretty sure fWAR has a universal replacement level for both leagues. It doesn’t matter who is on a player’s team when calculating WAR.
I hate to say it, but Baseball Reference does a much better job of calculating WAR.
This will be a fantastic blog, could you be involved in doing an interview about just how you created it? If so e-mail me!
Does WAR (among batters) in any way take into account the number of plate appearances a player gets?
From what I understand, it doesn’t, so someone who bats leadoff is automatically much more valuable than an identical player who bats 9th. Am I missing something?
Yes, replacement is on a per PA basis.
I’m sure this is a simple answer, and I think you may have answered this. But this is prorated for a whole season. So currently McCutchen is a 4.6 WAR, and if he plays at this level all year he’ll stay at 4.6, and have a 4.6 WAR at the end of the year. This is why you add yearly WARs to get a career WAR.
I might’ve answered my own question, but want to make sure. Thanks
I didn’t ask a question.. If I compare 2 players at 4.6 and 2.0, the 4.6 is worth 2 more wins for the whole year?
Thanks
It’s not a rating, so it’s not saying McCutchen is playing at a 4.6 WAR level. It’s cumulative, so if he keeps playing at a high level he will accumulate more Wins above replacement. Conversely, if he begins to play poorly, he can see those wins decrease. Let’s say he ends the year at 6.0 wins and you compare him to someone at 2.0 wins (as you said in your comment), McCutchen would be worth 4.0 additional Wins Above Replacement over the player with 2 wins. Next season everyone starts back at 0.
Now that there is a new version of Sierra on this site, is it quite possible that WAR for pitchers is going to be adjusted to be based on that instead of FIP?
Seriously? I think that the people who come up with these statistics and those who put faith in them only have an understanding of baseball afforded them by their Playstations.
The problem with WAR is that it is based on some arbitrary numbers. The positional adjustment is completely arbitrary – who determined the a CF is worth exactly 1.5 wins more than a first baseman. This is an arbitrary adjustment (obviously, since the positional adjustments just happen to be in increments of exactly 0.5 wins … weird). Also, the final number is determined by adding up statistics of different units. Who determined what significance wRAA and UZR have in relation to each other? Who determined that they can just be added together – molding different units into one single stat? Not only have defensive statistics, including UZR, been shown to be faulty in and of themselves (defensive metrics seem to differ depending on who is calculating them), but the hubris to think one can extrapolate the UZR to determine “wins” and just arbitrarily add it to a completely different statistic (wRAA) to determine a final definitive number is a model that would make any good economist’s head spin.
As an example, Brett Gardner plays left field, and Curtis Granderson plays center – if they switched positions, which they could (both have played both positions) – suddenly, their value as players changes. Just by moving to center field, Brett Gardner is worth one more win as a player. It is one thing to suggest (arbitrarily) that the Yankees would be one win better if Gardner played center and Granderson played left – However, Brett Gardner didn’t change – he is still Brett Gardner. Therefore – using the stat to determine BRETT GARDNER’S ABILITY is flawed. To take it to the next extreme – if Brett Gardner played center field this season, his WAR would be about 5.8 – and if Granderson played left field, his WAR would be about 4.3!!! Suggesting that just by switching positions, Brett Gardner would be more than a win more valuable to the Yankees than Curtis Granderson!!!
positional adjustments are not arbitrary actually. You should go out and research it a bit.
This is sports we’re talking about here, right?
Is there any verification method for WAR? For example, is there any way of estimating the uncertainty involved with the statistic? If not, it seems somewhat hand-wavy. I see the logic in its development, but it’s impossible to know how reliable the numbers are without some verification method. What has been done, or what is the rationale for believing in WAR?
First of all, thank you very much for this wealth of information! Fantastic site & article!
However, I had a quick question for anybody out there…
How is it that if we add up all of the players WAR on a team, that number doesn’t reflect the total number of wins over the baseline of 48?
For instance, the St. Louis Cardinals players added up to 28.1 WAR. However, the team finished with 86 wins, which is 38 wins over that base level for a replacement team. If WAR is an accepted method to determine an individual players’ worth, shouldn’t the entire teams’ WAR reflect record?
“If WAR is an accepted method to determine an individual players’ worth, shouldn’t the entire teams’ WAR reflect record?”
WAR is based on runs created/prevented, and sometimes teams win more or less than their runs scored/given up would suggest. That difference between a team’s runs scored/allowed and their record can’t be attributed to a specific player. There’s no intent or skill there – it’s just something that happens.
So there’s no actual “worth” in it. It’s not something that I would pay extra for (or less for), if I were a GM, because I have no way of knowing if that over- or underperformance is something that an individual player “takes with him”, like his hitting skill or his baserunning or his defense.
I do wonder if there’s something else throwing off the calculation, though. A 10-win variation is a little suspect. I’d be interested in knowing what else would cause the difference.
You forgot to add in the pitcher WAR.
Thanks Jesse, I appreciate it. Actually, I’m doing an investigation into this. If you take the runs created and compare it to wins over the baseline of 48, the MLB average for 2010 would be 7.24 runs per win over replacement.
Therefore, I would say that the ten runs per win method is at fault in this case. However, I would like to compare this with other years to test this.
And Andrew, I did take into account pitcher’s WAR.
Dear Fangraphs,
Please join the 21st century and use SIERA for WAR calculations.
Thank you.
Can anyone tell me why WAR for pitchers and position players aren’t two separate stats? I’m always running into people using WAR to compare pitchers to position players and to me it doesn’t make much sense. Pitching WAR is derived using a different methodology than that employed for Offensive WAR. Just because you named them each the same thing doesn’t mean they equate…
Pitchers’ wins and losses shouldn’t even be a part statistical analysis. Consider this: One pitcher “scatters” seven hits but gives up no runs. The opposing pitcher throws a no-hitter and loses (yes, he can – a walk, a sac bunt, a stolen base, a sac fly); who pitched the better game ? What did the “winning” pitcher do to get the win? With the exception of strikeouts and walks; everything a pitcher accomplishes is solely the result of his defense. In nearly every no-hitter I’ve ever seen, there has been at least one incredible defensive play keeping the no-no intact. One more thing: When looking at a pitcher’s record, does anyone ever look at who he faced in each of his outings? If a pitcher is frequently up against the other team’s ace, he will probably not do as well as when he is frequently up against the other team’s number four or five guy. Is this ever accounted for in the stats?
2011 Tyler Clippard 88.1 IP 18 Runs 1.2 WAR
2011 Jonathan Papelbon 67.0 IP 29 Runs 1.2 WAR
Correct me if I’m wrong, but isn’t there a glitch here? The question asked is: “If this player got injured and their team had to replace them with a minor leaguer or someone from their bench, how much value would the team be losing?” But why ask “if” (or at least ONLY “if”) in regard to events that have already transpired? Where does it ask “Did”? i.e. “DID the player often make a replacement player relevant?” If player A is considerably better than a replacement player, but only plays about 125 games in the average season, how is that NOT relevant. In other words, Babe Ruth on the DL is NO better than, well…me. And who cares how good or bad a replacement player is if the player being compared plays 162 games?
This is probably a really dumb question, but is there a reason that only FA dollars being spent are used to determine how much each win above replacement is worth in dollars? Why do we not use total MLB salary or 25 Man roster salary? Wouldn’t this give us a better understanding of the total value of the Win and not the value of this player simply in relation to the other FA’s available?
To: John Ogrin. I know exactly what you mean that every geek thinks their Bill James. To be honest I have never heard any human being, even at Yankee Stadium, (in all three of them) where I have been watching BB games since the 1950s say Mariano Rivers is the best pitcher ever.
As a matter I have never any say he was the best Yankee pitcher ever. Just to nit pick. Have to nit pick to get ready for pitchers and catchers. Even after the Giants huge win yesterday when the Giants got revenge for 1962. In 2008 we got revenge for 1961.
I’m still unconvinced about WAR. I may be off slightly on this number, but, for example, I read that Justin Verlander had a WAR of 8, impressive I guess. So that means he is eight wins better than a minor league or waiver wire player. Really? So your top minor league pitcher is going to win 16 games? (Verlander won 24). Hardly. Maybe I’m confused about what WAR really means, but I can’t believe that someone who won the pitching triple crown with that record is only eight games better than a replacement. Set me straight. By the way, I am not anti-saber. I’ve enjoyed reading Bill James since the early 1980s and believe much of what he writes.
If he received the same amount of defense and run support that Verlander did, he’d do quite well.
Remember than pitcher Win/Loss records are not entirely up to the pitcher but have a lot to do with the team behind them. Would he have won 16 games? Probably not, but that’s not a testament to his ability, but rather the team’s ability to help him out on defense and score runs.
The MLB “wins” statistic is not equivalent to the “wins” in WAR. I’m mangling the stats a bit, but, as an example:
Verlander’s W-L stat for 2011 is 24-5, but behind him, the team went 25-9 in 2011. So using Verlander’s WAR, a replacement’s team would have won ~ 8 fewer games, or would have gone 17-17. That sounds about right to me.
You can’t factor in the no decisions, blown saves, and come-from-behind wins that are part of the traditional MLB statistic, which is why the thought of an average SP getting 16 traditional “wins” is tough to think of, but having a decent (or above average) team win 17 games in which that average pitcher starts is not.
(PS the Cardinals went 18-15 in games in which Jake Westbrook, with a rWAR of 0, played.)
Exactly. No other pitcher on the team made it to 16 wins. And before acquiring Doug Fister; Phil Coke, Jacob Turner, Charlie Furbush, Duane Below, and Andy Oliver all saw time as the 5th starter. In 23 starts they combined for 4 wins. Turner is the teams top prospect and Coke, Furbush, and Below are all still on major league rosters, so it’s not like these aren’t legitimate replacements. I understand that wins/losses is dependent on a lot more than a pitcher’s individual performance, but it’s not entirely irrelevant.
Verlander also had 4 “tough losses” and no “cheap wins”, the Tiger’s defense was horrible, and his run support was average (59th in the league). So I don’t think you can say that his record was the result of being on a powerhouse team, a la 2010 Phil Hughes.
Still new to the advanced statistics so this was very interesting. However, I am a little confused by the constant use of WAR in the player projections for 2012. While they give a good idea of the player’s value and production, the WAR figure does not tell me what “kind” of player he is. I think I need to see the counting stats in addition to just WAR.
I want to kill so many Jews
WAR is a joke for one major reason. The 10 runs equals 1 win stat. Too arbitrary as if a win has to be a blow out to count. Also, SABR has no stats for clutch play which decides winning and losing in that not all runs or wins are equal.
I know that you won’t read this but here’s your explanation on those two things:
Using the “pythagorean” equation, we find that runs scored and runs allowed correlate very strongly to a team’s winning percentage and we find that about a 10 run swing in either direction equates to about 1 win. So, that’s why if a players is worth 10 runs above a fictional replacement level player, the player is worth 1 WAR.
SABR does indeed have stats for clutch play.
http://www.fangraphs.com/blogs/index.php/get-to-know-wpa/
There’s also this:
http://www.fangraphs.com/blogs/index.php/get-to-know-clutch/
WPA correlates with a player’s WAR as a we should expect good players to hit to their talent level. Good players tend to affect the game (makes obvious sense) in a better manner for their team.
Clutch doesn’t correlate with anything. Its all over the map. There’s no consistency from year to year for players. Basically, it says that there is no such thing as clutch but there is such a thing as small sample sizes and confirmation bias.
What is it good for?
What I don’t like about WAR is that it’s still a ‘counting stat’ in that if you play more games in a season, your WAR is likely to be higher than someone who played fewer games unless that player was much better. For example… if a player plays first base for 160 games and averages .05 WAR/game (season WAR of 8), he’ll appear to have the same value as a player who played 107 games and averages .075 WAR/game. Particularly for teams that have a lot of depth and cash flow, even a backup level player is usually capable of providing a WAR above 0.
I understand that there’s obviously some value in simply playing additional games, just being a body on the field and showing up to work, but I still think WAR should be presented additionally as a “per game” stat, rather than counting stats.
Furthermore, any stat that deems Mark Teixeira a reasonably good hitter must be fundamentally flawed (jk).
Of course! Glad to be of help.