Run differential, .500 and a winning Mets team

By Brian Mangan, originally published on The Read Zone:


“We kind of like our team.  If you look at the run differential, we should be a .500 team. We’re not. At the same time, it doesn’t mean we should throw everyone overboard.” – Sandy Alderson

These words, uttered by Mets GM Sandy Alderson last week (in an interview with Jon Heyman for CBSSports.com) set off a firestorm in the Mets universe of newspaper columns, blogs, and on twitter. Much of the commentary was critical of Alderson, and to be fair, Alderson’s Mets deserve plenty of criticism. Was he deflecting blame? Is his head buried too deep in the spreadsheet to see that the Mets are actually “what their record says they are”?

The short answer is: No, Alderson is actually right.

Alderson’s use of “run differential” is spot-on, and deserves examination by (and explanation to) this passionate fan base. Say what you will about this team, but the concept of “run differential” (and its sister, “win expectancy”) are not just hocus pocus or advanced statistical mumbo jumbo. In fact, run differential is the simplest thing you can imagine.


Baseball is a game, and games have elements of chance in them. When chance plays a role, sometimes inexplicable things can happen. Scatter nine singles in a game and you might get shut out. String together nine singles in a row, and you might score seven runs.

Similarly, imagine flipping a coin and rooting for tails. Perhaps the first ten times you flip the coin, heads comes up seven out of ten times. If you were wise, you would look at that distribution and say “I still expect tails to come up around 50 percent of the time in the future.”

That is kind of what run differential does.  It doesn’t undo the past, but it can help you predict the future. Run differential keeps a tally of all the runs scored and runs against each team in a given season. From those, you can extrapolate a number of expected wins. Generally, a team that scores more than it allows will win more than it loses, while a team that scores the same number of runs as it allows will, generally, in the long run, end up around .500. It’s a simple concept.

But how well does it work? Luckily, smarter men than I have tackled that question. Russell Carleton of Baseball Prospectus wrote a fantastic column about the correlation between run differential and actual wins earlier this year.  He explains:

In general, I like to look for the point where the correlation reaches .70. For the uninitiated, 0 means that there’s no relationship between the two numbers, and 1 means there’s a perfect relationship between them. It’s a matter of how close you want that number to be to 1 before you feel comfortable, and you can see a chart below of how the correlation approaches 1.0 as the season goes on. The correlation hits .70 after 39 games. So, around the 40-game mark—mid-May—run differential starts to be a good predictor of what things will look like at the end of the year.

run diff 2

The more games you have played, the more accurate run differential becomes at predicting how good your team is. Which makes sense — the more games you play, the less noise there is in the number thanks to things like blowouts, and strength of schedule and streaks.

The Mets have now played over 90 games — so the correlation at this point between run differential and wins is somewhere around .86, which is quite high (note of course that the line in the chart above never reaches 1.0 — and the reason for that is because no matter how many games you play, run differential will never exactly equal the team’s record in all cases.)  Once you’ve played 90 games (almost a thousand innings!) we know who you are.

Regardless of how your gut feels about it, run differential is a damn good way of predicting final records. If you’d like some more concrete proof about the correlation between the two, check out this excellent bit of research from the Star-Ledger:

The recent history supports this claim. In the past six full seasons, beginning with 2008, 95 teams have finished the year with a positive run differential and 86 of them have also had a winning record. And three of the teams that didn’t finished at .500.  So if the Mets were to finish below .500 this season they would join a minute group of six teams from a 95 team sample to do so.

And that’s supported out this season too. This year, 15 teams have a positive run differential and only the Mets are under .500. No other team is fewer than three games over, with the Royals at 46-43.

Only six out of 95 have failed to reach .500 by season’s end. Does it happen? Sure. Is it rare? Hell yeah.


According to the research, “at the exact midpoint of the season (81 games), Pythagorean projection correlates with winning percentage [better than] actual winning percentage [.]” The Mets will probably play like a .500 team from here on out … but this, of course, is no consolation to a team that is presently 45-50. Going .500 from here on out would only result in around 80 wins.

At the end of the day, the gap between what the Mets record is and what the Mets record should be (given their true talent level) can be explained almost entirely by their record in one-run games, in which the Mets are 12-20. Stop me if you’ve heard this before: if the Mets were 16-16 in one-run games, they would be over .500 — or what their run differential says they should be.

I’ve heard all the arguments against run differential and refusing to blame their record in one-run games for the difference, but there is one that seems to have the most traction among the anti-stat crowd:

“Hey Brian, the Mets aren’t unlucky in one-run games, they’re just bad.  Good teams win those one-run games.  That’s what makes them good teams.”

Well, not exactly.  Yes, it is true that great teams have better records in one run games (but maybe not by as much as you think), but the year-to-year records in one run games deviate widely from what you might expect given such a small sample of those games. Look no further than our own Mets!

w expect

The 2013 Mets were worse than the 2014 Mets – why did they succeed in one run games?  How about the 2008 Mets, who were excellent – why did they fail?  How come the best team in baseball right now is barely over .500 in those games?  Every team in the NL West is within one game of .500 in one run games … except the lowly Padres, who at 10 games below .500 had a fantastic 17-10 record in one run games.

Sandy Alderson is right.  This team is better than last year’s team.  The 2013 Mets scored only 619 runs and allowed 684, and this year’s team is better from top to bottom, as we pointed out in our season preview.  The results so far, record aside, are good, as this year’s squad is on pace to score 635 runs but allow only 630 — a 70 run improvement.

So when Alderson says that he’s happy with the team, and he’s happy with the run differential, he is not simply buying time or deflecting criticism (although the statements about waiting for the end of the homestand to determine whether or not the Mets should be buyers or sellers is most certainly fictional).  He’s making a true statement and showing patience with this club’s turn around.

At some point, you are what your record says you are.  But not yet.


To submit a post, click here




221 comments
archer41
archer41

I can't say run differential is meaningless, but the reality is the record is what it is.  If you can't win close games, you're going to drop in the standings.  There's no way around that.  It also suggests that TC has not done a good job this season, but most of us already know that.  

Lenin AV
Lenin AV

Good teams find the way to win 1 run games ... Bad teams lose them !

Not4Nutten
Not4Nutten

Having six outfielders is not in the analysis is it!!!!!

Evil_MrM
Evil_MrM

Fascinating analysis...thanks for sharing.

username
username

I usually don't post on here but this article is the first real jump into real baseball analytic's this blog has made in awhile so I just wanted to clear something up in this article.  Instead of writing the whole thing out I'm just going to copy and paste a excerpt from Dave Cameron's latest expected run differentials post on fangraphs because he lays it out pretty well


"While pythagorean expected record — the number of wins and losses a team would be expected to have based on their runs scored and allowed — has become nearly a mainstream concept, I’ve never been a huge fan of using runs to determine how well a team has played thus far.

After all, the entire point of looking at run differential instead of actual wins and losses is because we’re acknowledging that wins and losses are affected by the timing of when runs are scored or allowed, and history has shown that run sequencing is mostly just randomness. So, developing an expected win-loss metric that removes the affects of sequencing is a good idea, but pythagorean record only goes halfway to that goal. It removes the timing aspects of converting runs into wins, but ignores the timing aspects of converting baserunners into runs. Evaluating a team by its run differential removes some of the sequencing effects of wins and losses, but leaves plenty of other parts, with no real reason why we should arbitrarily include some sequencing while taking other parts out.

That’s why I’ve always preferred to look at a team’s performance based on expected runs scored and allowed, rather than actual runs scored and allowed; this gives us the most context-neutral evaluation of team performance to date."


Pythagorean only goes half the way there and expected run differntial is the best way to judge a teams true talent we have today.  At the time fangraphs last put this together, June 26th the mets, expected run diff was -17, and the teams ex win pct was .475.  Don't know how thats changed since then tho 

James
James

So I guess what I would love to see is what was our run differential at the break the last three years?  Especially those first two years when we were > .500 going into the break and ended up in our usual 18+ games out of the playoffs.

kingman76
kingman76

This year, 15 teams have a positive run differential and only the Mets are under .500. No other team is fewer than three games over, with the Royals at 46-43.

Doesn't speak very well of Collins if they have a positive 19 run differential and they are 5 games under .500.

mjkru10
mjkru10

More of this type of analysis, please. I am as frustrated as anyone with the record the team currently has, but these numbers are starting to correspond with the eye test. This squad can pitch, and we are missing the best one of all due to injury. You look at a team like the Yankees who are searching high and low for serviceable arms. We are debating what we should do with our surplus of arms. This should make it much easier to acquire that one impact bat who can put us over the top. Keep the faith, everyone! Brighter days are ahead. We've waited this long haha.

C Nerrie
C Nerrie

I wish I could give this more time, but here's one quibble.

"Once you've played 90 games...we know who you are.... The Mets will probably play like a .500 team from here on out."

"Yikes!" I thought (even though they meant this as a positive in the sense that the Mets are better than their current W/L record).

But can you even make these statements  when the entity you're measuring is changing over time?

In our case, this is not the same team that started the year. The bullpen is redone. C, 1B, SS, RF have all begun playing consistently better (and came around at different points in the season) while 3B and CF are about the same. 

So in our case, if from the beginning of the year till now the team should have a .500 record, they should be better than that going forward based upon what we observe (and BA or OBP could be measured).

Is this just sloppiness on the writer's part? 

Am I off-base?

It would be nice to have a real statistician analyze this stuff.

Michael Leiman
Michael Leiman

Excellent article, Brian. Thought provoking, thank you. If you can, can you explain a bit more about correlation...that is, when it reaches 70, what exactly does that mean? Does it mean that there is now a 70% probability of whatever you're predicting actually happening?

Now your analysis leads, of course, to the question...why are the Mets underperforming so far. You say that all other teams with a positive differential are at least 3 over .500. As the Mets are 5 under, it suggests that the Mets are at least 4 wins shy of where they "should be." This brings me to the category (sadly) of Horrible Losses, that is, games they REALLY should have won. The first game of the Atlanta series in Atlanta, when they blew a 2 run 8th inning lead with 3 errors and a bases loaded walk LEAPS to mind. The Mets have had a bunch of these. I would say the main culprits of this  is the instability, till recently of the bullpen, it's over use and below average infield defense. These have made the Mets vulnerable to losing games that they really should have won.

Or it just may be chance!!

bbr7
bbr7

It's been said that out of all the teams that have a positive run differential the mets are the only team that aren't over .500. This is the most telling stat. In order for this to happen you have to lose close games. Those games where you have the lead late in the game and blow it. Or those games you have a runner on third with one out and can't hit a sac fly late in the game to win the game. This is why our record is bad. I hope this changes

dooley
dooley

First of all, great article! I hope you are given the opportunity to contribute more.

However, even though we have a higher run differential from last year, the question still remains, why are we still under .500? It seems that run differential describes how we are under performing (through chance?), but it doesn't seem to pinpoint the root of the problem. So, why is this team still seem to be a 70ish win team even though we are scoring more, and allowing the other team to score less? Is there another stat, or set of stats worth analyzing?

Shawn M. Milburn
Shawn M. Milburn

The sabermetrics is interesting here, but the reality as to why they are playing better is this. The pitching has been consistently good all year. Of late Zack Wheeler has been great and Jacob deGrom has been excellent. Mejia is adjusting nicely as a closer, Familia has improved and Edgin and Black have shown how starting in AAA this season has paid off.

Chris Young has been clutch in a limited role, Wright is hitting like David Wright, Duda is using the whole field and only pulling home runs when necessary. Murphy and Lagares have been consistent all year. But the most important factors have been Granderson and d'Arnaud. Grandy has been a Top 10 hitter since May and an excellent lead off hitter. d'Arnaud has been a .300 hitter since his recall from AAA and may very we'll be the guy we all were told he would be. Both guys have been delivering clutch hits for the last couple weeks. Tejada has been better than expected and Niuewenhuis and Campbell have developed nicely. Finally, Bobby Abreu's leadership and clutch hitting is a major contributor to this team. If Montero comes back up and we finally see Syndergaard and they play well, this team could be playing important baseball in September. Finishing with 90 wins is not happening, they'd have to go 45 and 22 the rest of the way, but 82 wins would be nice and a huge push for next year. 37-30 is a definite possibility, if all continues as is. I definitely see the light at the end of the tunnel

bbr7
bbr7

This is absurd. We are 5 games under 500 and more than half the season is over. Would bill parcels make excuses for a losing team with point differential? You've got to be kidding me. Management of this team is a disgrace. They just make excuses. If they ever get their act together it won't make up for all this garbage we have had to see. Just stfu and start doing something about it

Brian P. Mangan
Brian P. Mangan

@username Thanks for the comment, and I did read that over at Fangraphs.

The research seems to say that expected run differential is slightly more accurate than run differential at predicting wins (just like Baseball Prospectus's third-order wins), but yet, I cannot seem to wrap my head around the concept.

I realize that's a little unfair of me to say (since I'm expecting the readers here to believe the research that I believe) but to project wins -- already a conceptual thing -- on something that's also, itself projected, is a step too far for me to want to accept at this moment.

Mets third-order win pct now is .497, and pythag is .523 (both better than real life .474).

username
username

so basically while the mets have been unlucky in converting runs into wins, they have been lucky in converting walks, hits, ect, into runs

James
James

Interesting stats, we are 10-6 in "Blow Outs" with a +26 Run Differential, while 13-20 with a -7 Run Differential (obviously) in 1 run games.  So if you remove our blowouts and one run games we have an exactly even Run Differential (since we are currently +19 all from the above mentioned games) and we basically .500 in those games (23-26 in the above games, while 22-24 in all non-blowout/non-1 run games.


To answer my own question (thanks Baseball Reference).  We were terrible in Run Differential both before and after the break in 2013, but in 2012 we were +20 before the break and -80 in the second and in 2011 we were +11 and -40.  We are a really really really bad second half team.  Hope that improves.

rf04
rf04

@C Nerrie It's certainly true that many players have begun playing better - but that's just more evidence compiling that clarifies what the team is. In the same way that the Mets are a team that's +19 but 5 games under .500, sometimes because of some strange luck you look at the statistics and find that Curtis Granderson is a .150 hitter with very little power. It could continue that way, but probably won't (and didn't).

The point of the article (and the statistics) is that OVER TIME these things USUALLY balance out. Although there are times when you sign a guy who hit .267 with 36 HRs to a 4 year deal for a ton of dough and he hits worse than Mario Mendoza.

Things like that appear to have a 99% probability to only happen to Mets fans. That is an actual statistic (in that 99% is really a number). So it's entirely probable that if any team could end up +120 and have a 70-92 record, the Mets will be that team.

Brian P. Mangan
Brian P. Mangan

@C Nerrie 

If misleading, it may be becuase the sentence was poorly written: "Once you've played 90 games...we know who you are.... The Mets will probably play like a .500 team from here on out."

What I mean is that after 90 games, the run differential is getting to a point where it is very close to reflecting the true talent of the team.

And yes, teams change over the course of a year -- I'm willing to bet that the Angels with Mike Trout were certainly better than their April 2012 run differential.  But you'd have to be wearing pretty rose-colored glasses to think that the 2014 Mets are somehow a different team post-break than they were pre-break.

They're hitting better now, a little, but so many things happen in the course of a season that it comes out in the wash.  I think they're a little better now than in April/May, but we'll see.

rf04
rf04

@Michael Leiman Think of tossing a coin. The odds are ALWAYS 50/50 for heads or tails before each toss. But once you start tallying your results, it will take a while before the results actually show 50/50. After 3 tosses, the closest you can get is 67/33 (and possibly even 100/0) but because the sample size is so small it doesn't actually tell you very much if anything. Eventually, the randomness begins to converge approaching 50/50.

If you keep running the same experiment with the same variables, over time you approach the truth of the odds. In baseball - while the result is either WIN or LOSS, the variables in the experiment (the team) are a lot more fluid and changeable than a coin. So what that graph says is that according to previous data about baseball team W/L records compared to their +/-, after about 39 games (when the curve hits 70%), there is very little variation between the # of actual wins and the # of wins the +/- seems to predict the team would have.

It's not perfect... but it boils down to - don't get too excited about any small stretch of games (like an 8-2 home stand). But you can likely judge a team by a stretch of 39 or more games according to the +/- correlation.

dooley
dooley

I am not sure if this helps, but in science you generally use a cutoff of 0.9 or higher where there is a ten percent probability of the correlation being due to chance alone. Of course they are dealing with studies that have an enormous social impact, such as medication, and not baseball

Brian P. Mangan
Brian P. Mangan

@Michael Leiman Thanks! 

I can't explain correlation much better than the excerpt ... but as it approaches 1, the correlation gets stronger.  At 1, the correlation is exact, while at 0 there is no correlation at al between the two things.

So for instance, there is some correlation between the weather getting hotter and higher run scoring in MLB, so maybe the correlation there might be .80. I can't speak to the percentages in between, but statisticians all have their own personal preferences before they say something is correlated.  Carelton likes .70. 

As for the Mets record, its probably a combination of random chance and a couple bad collapses. 

Here's a new interesting thing, too.  Look at the MLB standings... there isn't a single team with negative run differential ahead of a team with positive run differential in any division: http://espn.go.com/mlb/standings

rf04
rf04

@cyrus Not entirely true... you're describing W/L record in close games. The good thing that this stat says about the Mets is that they're generally losing, but not getting blown out. So, YES - that means in some games they are getting close but not cashing in with the big hit. But being in close games means being close to winning. The worst of all worlds is losing games and losing badly. That implies there is very little your team is doing well. The Mets are showing they're not so far off from the talent levels of other teams because they generally play close games.

bbr7
bbr7

These type of games that I am describing, the ones the mets predictably lose is EXACTLY what separated the good teams from the bad teams. And that's a FACT

Brian P. Mangan
Brian P. Mangan

@dooley Yeah... I think that is the great unexplained question that remains, and I don't know the answer yet.  For teams that deviate from their expected wins -- why do they do that?

I suspect the answer is something like 90% chance and 10% managerial decisions, chemistry, bullpen.  It may have been researched, so maybe someday I'll be able to get into it.  I really wonder, also.

Robert Hess
Robert Hess

@cyrus  I agree.  Here's run differential in a nutshell ---If in each game you score more runs than the other team than you've said it all.  The Mets have lost way too many games by one or two runs.  Then they'll win a few games by scores of 6-1 or 7-2 and it scews the representation of run differential.  The absolute bottom line is Wins vs. Losses!!!

Greg Baker
Greg Baker

@cyrus This comment right here is why we can't have nice things.

Brian P. Mangan
Brian P. Mangan

@James One interesting fact is that blowout wins correlates better with team talent than one run wins!  It's in the research presentation linked above.

Good teams win more one run games than they lose, but by a small margin.  Then win a lot more blowouts than they lose.  It makes sense.

James
James

@rf04 @C Nerrie And because its too easy not to say, Terry Collins will be that manager.

C Nerrie
C Nerrie

Thanks for responding.

I did understand you to be saying that after 90 games the run differential comes close to accurately reflecting the talent of the team.

If we had nothing else by which to judge the future performance of the team, then we could go with that and confidently say, "The Mets will probably play like a .500 team from here on out."

Or if we observed no change in players or their performance, we could confidently say that.

But if Trout joined the team, or if a couple of stars became injured, or if half the players were playing better than they did during the first part of the season, we can no longer so confidently predict the future using past run differential.

I took a statistics class from a man who really loved it, and I respect statistics. 

But with something like a baseball team - which changes - we need to be alert to things staring us in the face that may outweigh the statistical prediction, don't we?

(Though I gather from your rose-colored glasses statement that you don't see what I see from the team....)

Once again, thanks for the response.

Michael Leiman
Michael Leiman

@rf04 @Michael Leiman Thanks very much for this explanation; very helpful to my rational brain. On the other hand, I'm a METS fan and am wildly excited and hopeful about the 8-2 home stand!!

Brian P. Mangan
Brian P. Mangan

@dooley Thanks!  Despite all of this I am not a statistician by trade... so that insight helps a lot.

Michael Leiman
Michael Leiman

@Brian Mangan @Michael Leiman Wow! Well, despite not fully understanding correlation, I'd say this one approaches 1!!

Thanks for your thoughtful response, Brian. It's good to have a non emotional, non finger pointing discussion about the Mets. Further, I've looked at some of your prior writings...sorry I didn't know about it earlier. Two things stood out: Excellent piece about the use of CY and how it reflects, oh, lets' say, a lack of awareness by our manager. Similarly regarding your thoughts on Cowgill. Baseball is a game of failure, but it seems that, at least in these cases, the chances for failure were increased. And then, of course, the fans go crazy and beat up on the "failing" player.

Still, I'm not a fan of CY and felt that Campbell should be given the chance to play regularly in left. Until Neuwinhaus came along and starting hitting more consistently and with more power than I thought possible. So now I'd play Kirk, at least against righties.

Like your stuff a lot, Brian; still, sorry, CY!

James
James

@cyrus Yes it is a "FACT", definitely in all caps, except that is exactly covered in the article how we were a bad team for years that won in 1 run games (and with a terrible bullpen no less), so it seems much more a "FACT" that a lot of this is randomness instead of a FACT.


That doesn't mean that there isn't some correlation, since we are humans after all that feeling that we "always" blow 1 run games has to get into players heads, which is why Grandy's HR against the Braves was so very very huge.

zoddie
zoddie

@cyrus Did you read the post?! 


If your hypothesis that your 1-run record is exactly what separates the good teams from the bad then why do teams 1-run record change so much from year to year with teams that are largely the same?


Are the 1st place Nationals a good team?  Because they have a losing record in 1-run games.

jmjmjm
jmjmjm

Don't good teams also lose 4-3 instead of 4-2, thus hurting their one-run record?

rf04
rf04

@Brian Mangan @dooley It's the human factor. If you could reduce winning baseball games down to a perfectly predictable statistic, there wouldn't be much need to play the games.

bbr7
bbr7

Exactly. That's how it's always been and how it always will be.

Brian P. Mangan
Brian P. Mangan

@Greg Baker Hahah yep.

@cyrus If you read the article, or the other stuff I write, you'd know that I'm quite critical of the Mets and the management.

At this point of the season, run differential is a better predictor of final record than current record is. That's just a fact.  They can't undo being 45-50, but it is good news for the future hopefully.

James
James

@C Nerrie I tend to agree with Brian, after 90 games we have seen the team's talent.  Yes some players are playing better, I think Grandy will be better in the second half (more consistent) than he was in the first, but we will get some negatives to that as well (deGrom getting shutdown for example and Legares having a slower second half).  I think this is a .500 or slightly better, if the team can keep focused on baseball and trying to actually win, something Collins has been TERRIBLE at as a manager then we should have a decent second half.


Brian P. Mangan
Brian P. Mangan

@Michael Leiman @Brian Mangan Thanks for the super kind comments!  Make sure you follow @readzonenow for all our updates, and I'll probably be here again too.

As far as CY-- you can't undo the past, so his ship may have sailed.  I'm glad he was able to deliver a couple of clutch home runs for us though.

bbr7
bbr7

Of course the nationals are a good team. They are first in the league in era and they are in first place. It's that simple. And I'm not going to go into my hypothesis. What I was trying to say is my hypothesis is characteristic of the mets. And why am I saying hypothesis? It's the way it is. If you watch losing teams and winning teams they have certain key characteristics, which is one of many. Nevertheless I hope the best for the franchise. But I refuse to be optimistic untill they break out and prove me wrong. I did read the post ( to answer your question ) and of course I think it's valid. But the bottom line is wins and losses. And I've had enough of the latter. I've also had enough of trying to over analyze what we have. We have what we have. A team that is 45-50

bbr7
bbr7

I understand what you are saying. And on a small level I do think they are improving. But I see it on a different level. You are analyzing the hell out of this. Again, they are improving. But overall they just aren't getting the job done. It's that simple.

zoddie
zoddie

@cyrus I think you are missing the point that it is instructive for the future.  The data is important because it can instruct an organization's future, which is why it is important to Sandy.  If the Mets had a -47 run differential, like the Phillies do, we would know we are very, very far away from competing (like the Phillies are).  That our chances in 2015, without a MAJOR change in personnel, would be pretty terrible.


But that's not the case.  What's done is done and the Mets are under .500.  This team has not done some little things to get wins (like sac flies, etc.) but that doesn't mean they won't next year.

bbr7
bbr7

Ok well we will see

bbr7
bbr7

Furthermore, I hope they end the season respectfully. I hope they spend money this offseason and spend it wisely. I can see it happening. Best of luck to them and the fans