Why is Run Scoring Depressed in September, and What Just Happened in August?

Expect to see less of this in the month of September. (via Cathy Taylor)

Expect to see less of this in the month of September. (via Cathy Taylor)

While monitoring the size and location of the strike zone this season on a monthly basis, of particular interest to me has been the resulting run environment across major league baseball. Runs have dried up in recent years as the strike zone has expanded, and some of my previous work has estimated how much of the dip in run scoring is directly attributable to the larger zone.

This relationship has remained largely consistent over the past half-decade or more. As the bottom of the zone continues to fall, runs and walks drop along with it, while strikeouts creep higher. It is with this as a backdrop that we come to the month of August 2015. Consider the following visual of average runs scored per team per game by month over the past five seasons:


After three full years elapsed without teams averaging as many as 4.30 runs per game over a single calendar month, major league teams averaged a whopping 4.49 runs per game in the just-completed month of August. This may not sound like a big difference, but you can see from the image above just how unprecedented this is in recent years.

The high octane month has powered the yearly offensive numbers to a point where run scoring in 2015 is now almost assured to be higher than in 2014. Here are the latest data:

Year Strike Zone Size (sq. in) Strike Zone Size Below 21” (sq. in) K% BB% R/G
 2009 435  0 18.0% 8.9% 4.61
 2010 436  6 18.5% 8.5% 4.38
 2011 448 11 18.6% 8.1% 4.28
 2012 456 19 19.8% 8.0% 4.32
 2013 459 30 19.9% 7.9% 4.17
 2014 475 47 20.4% 7.6% 4.07
2015* 478 50 20.2% 7.5% 4.20
* At the end of August 2015

Here are the accompanying visualizations of the strike zones to date in 2015, from the umpire’s perspective:

* As of the end of August 2015

* As of the end of August 2015

While this August was unusually productive for offenses around the league, as part of my last strike zone update at the end of July I noted that at least in recent seasons, September tends to play out with notably fewer runs scored than the season average. Here’s a table that shows the differences:

Year K% BB% R/G
2009 0.3%  0.1% -0.11
2010 0.8%  0.0% -0.17
2011 0.8%  0.1%  0.13
2012 0.7%  0.1% -0.09
2013 0.8%  0.1% -0.12
2014 0.3% -0.4% -0.16

I wanted to determine why September has been less kind to offenses than other months. Given the extreme outburst of runs in the August that just wrapped up, the analysis can serve the secondary purpose of attempting to understand why that was the case.

Let’s consider possibilities one at a time.


Could it be as simple as the typically cooler weather in September cooling off  bats? It has been shown that higher temperatures allow fly balls to travel farther, logically leading to more home runs, which sounds like a good way to score more runs (unless you think home runs are rally killers!).

Here is an image showing average runs scored per team per game against average monthly temperature over the past five seasons:


While there are certainly some higher-scoring hot months and somewhat lower-scoring cool months, on the whole there is virtually no relationship between monthly temperatures and runs. Even if temperature did play a telling role in runs scored per game, on average both April and May have been notably colder months than September. August 2015 was the fourth-warmest month of the past five seasons, for what it’s worth.


We already know that strikeouts have been up in September. Could it be that when hitters have been making contact in the final month of the season they’ve also been less fortunate than earlier in the year when pitching and defense are still being fine-tuned?

Here’s a look at batting average on balls in play (BABIP) over our five-year period of study, by month:


From the image we see that BABIP plays no role in the variation in runs scored per game from month-to-month. Even if it was impactful, it turns out that BABIP is typically lowest in the first half of the season, with July, August and September all having higher BABIPs that April, May and June.

While it doesn’t stick out, August 2015 did have the second-highest BABIP of any month over the last five seasons, at .302. In fact, June through August 2015 comprise three of the six highest BABIPs by month since 2011. It’s interesting to note that despite defensive shift totals setting new records year after year, defensive efficiency across the major leagues is certainly not getting any better on the whole.

Another metric that we can lump under the “luck” category is home runs per fly ball.


It would appear we have made an impactful discovery; HR/FB percentage drives a notable amount of the monthly variation in runs scored per game. September has seen the second lowest HR/FB percentage of any month in the season over the past five years, higher only than April.

This also goes a long way to explaining the anomalous month of September 2011 in the table above, when run scoring was actually higher than the season average for the only time in the PITCHf/x era. September 2011 was unusual in that it saw a HR/FB percentage that was the second highest monthly rate of that season.

Finally, the runs-aplenty month of August 2015 can be seen in the top right corner with the highest HR/FB rate of the last five years by a fairly wide margin. From this analysis it is apparent that when batters made contact last month, they fared better than normal both as far as reaching base when the ball stayed in the park and having the ball exit the yard when it was put in the air.

Out of curiosity, I checked the correlation between average monthly temperature and HR/FB percentage, and found just a small correlation (R-squared of 0.04). In reality the more complete effect on a baseball in flight would be air density, in  which temperature plays just a part, along with elevation, air pressure and relative humidity. Perhaps if the other elements of air density were readily available, a study of runs scored per game vs. air density would prove more fruitful than purely temperature.

Pitch Velocity

We know that pitch velocity is up. Here is a look at monthly runs per game and average four-seam fastball velocity over the past five seasons:


There is a visible correlation, although in this case I would not say necessarily without further study that this is causation as well. Velocities have been increasing at the same time as the strike zone has been falling, and the latter may be much more of the cause of this trend than the former.

Velocities are certainly higher from June onward in the season, but July and August see higher speeds than September, so regardless this relationship doesn’t go too far toward explaining the lack of offense in September.

Nonetheless, a look at velocities by month shows that August 2015 is an outlier in the top right corner here once again; Four-seam fastballs were a full 0.2 mph faster on average than any other month in the last half decade. Another startling way to break this down is to note that before last month, teams scored on average fewer than 4.20 runs per game in all 11 months where the average fastball velocity was over 92 mph. Then all of a sudden last month four-seamers were thrown harder than ever, and teams scored almost 4.50 runs per game. What a crazy month!

Strike Zone Size / Average Pitch Height

The method that I use for measuring strike zone size is not ideal for measuring on a month-by-month basis; sample sizes of two months or greater give far more accurate results. Given that we know during this five-year stretch the strike zone growth was overwhelmingly being driven by a falling bottom of the zone, we can make use of a proxy for zone size, borrowing an idea from Rob Arthur.

The proxy is to look at the average height of all pitches thrown on a monthly basis during our period of study. Pitchers adapt quickly when the strike zone is being called lower, and throw more frequently down in the zone.


The lower that pitchers have been allowed to throw and get away with it via positive reinforcement from umpires, the better it has been for run prevention. Except, of course, for August 2015, which once again stands out, only this time in the top left corner. Pitchers kept the ball lower last month than any month in the past five seasons, yet teams scored more often than in any of those same months. Is this really just a “lucky” month for offenses where they benefited from better than expected results when making contact? Or could this, along with the noted streak of months with slightly better BABIP results, be the start of a sign that teams and batters have started adjusting to the new era of more strikeouts and a lower strike zone?

On the whole, there is very little difference in average pitch height among the months of the season, so while this is a factor, it is really explaining why offense has been down on a year-over-year basis more so than a month-to-month basis within a season.

Batter-Pitcher Familiarity

We know that the Times Through The Order (TTOP) penalty exists: Batters in the aggregate perform better each subsequent time they face the same pitcher in the same game. This effect may be due to the pitcher tiring, the batter getting familiar with the pitcher’s arsenal, timing, and release point, or some combination of those things and other aspects of the game.

We know that September is a month where active rosters can exceed the 25-player limit that is in place from April through August. This means that most teams add call-ups, some of whom are debuting in the majors.

I wondered if the TTOP extended to the career level. In other words, on the whole do batters gain an advantage after having faced a particular pitcher many times compared to those first few looks? Anecdotally, it feels like often rookie pitchers baffle major league lineups the first time through, when advance scouting reports have to try to fill the void of personal experience for big league hitters.

To examine this question, I looked at wOBA by major league career batter-pitcher times faced. The sample for this was all plate appearances since 2011 where the first batter-pitcher match-up at the major league level occurred no earlier than 2011. (Actually, I know only that each batter-pitcher match-up did not occur between 2007 and 2010, so it is possible a few pairings faced each other five years or more prior). Of course this is not perfect; hitters may have faced pitchers in spring training, in minor league ball or even college or high school. But I feel it is a close enough approximation to address this question.

Here are the results of wOBA by batter-pitcher times faced at the major league level:


There certainly appears to be a trend suggesting pitcher familiarity is advantageous to batters not just within a game, but at the career level. Of course some of this trend is in fact just the TTOP, but this extends to multiple games over even multiple seasons against both starting pitchers and relievers.

It is important to understand that the graph above includes results from all batter-pitcher match-ups. Less-talented, fringe major league hitters and pitchers would tend to make nearly all of their major league match-ups on the far left of the graph, since they wouldn’t stick around in the majors long enough to face the same opposing player multiple times. Conversely, batter-pitcher match-ups that number in double digits likely suggest that both the batter and pitcher are talented enough to have survived in major league baseball long enough to have faced the same opposing player on so many occasions.

Another visualization that agrees with the fact that batters gain an advantage from additional looks throughout a career is this one that offers strikeout rate:


Some of the advantage gained by hitters may be due to the fact that as pitchers age, velocity wanes. Conversely, plate discipline for batters generally holds steady or can improve as they age.

It also lends credence to the move to relief pitcher specialization. More pitchers throwing fewer innings means pitchers can  throw harder and potentially have to face the same hitters fewer times over a career, both seen as positives by this study.

Are there really more major league debuts in September than other months? As far as pitcher debuts go, April and June have more on average than September. For hitters, on average there are more debuts in only June than September. There is some logic to April, June and September being the most common months of the season for players to debut. April debuts would be players who make the team with solid spring training performances, June debuts would be prospects held down until the Super-Two deadline passes, and September debuts would be call-ups added when rosters expand. Given that June and September have been the lowest scoring months of the past five seasons, perhaps we’re on to something!

Unfortunately, when we look at the average batter-pitcher time faced for all plate appearances over each month of the season, the relationship doesn’t hold water for September:

Month Average Batter-Pitcher Time Faced*
April 4.02
May 3.61
June 3.60
July 3.97
August 3.70
September 4.41
* Of all PA where first batter-pitcher match-up in the majors occurred 2011-2015.

September actually has the highest average time faced of any month within the sample used for this study. Of course there are competing interests here; while many new players are added to rosters in September, players who have been on big league rosters all season have now had April through August to amass more plate appearances against rival pitchers. The sample again also excludes all plate appearances where the batter and pitcher faced one another in the majors before 2011, so the calculated average is incomplete insofar as it does not include these battles of more senior players.

It also must be considered that teams still in contention are likely to stick with their most productive players down the stretch, and these players are often “veterans” who “have been there before.” These players would tend to have faced major league pitchers a number of times each, in particular with all of the intra-division play that occurs in the final weeks of the season.

Talent Level

One final theory to the lower run scoring environment in September is that as stakes are higher for teams in playoff contention, managers may decide to play most talented players on a daily basis. Similarly, the bullpen may get shorter, with the top relievers of the year getting a higher percentage share of innings pitched than they had earlier in the season.

This would have to be studied by using preseason projections of a fairly all-encompassing metric for all players (say wOBA for batters, wOBA allowed for pitchers, for example). With these in hand as proxies for “talent level,” the average of these metrics could be calculated by month and then a relationship with runs scored per game could be explored.

Alas, I could not easily find projections for one all-encompassing metric for both batters and pitchers from one projection system for all years 2011 through 2015. So this last theory is left as an exercise for the reader.

Putting It All Together

I would make the case that the recently concluded month of August was the most unusual month of baseball in the past five seasons. Pitchers threw harder than ever, to a strike zone that was lower than ever, yet batters responded by hitting home runs on a higher percentage of fly balls than any other month in the past five years and a compiling a higher BABIP than all but one month over that time period. The result was the highest scoring month since the strike zone began to fall.

The question of why September is a dry month for offenses hasn’t fully been answered, but perhaps some headway was made. Home run per fly ball percentage tends to be lower in September than the yearly average, and this behavior appears to drive a fairly significant amount of the variation in monthly runs scored per game. Fastball velocities are slightly higher than average, which on the whole helps to suppress offense. There are more player debuts than normal in September with the expanded rosters, and initial batter-pitcher match-ups favor the pitcher.

Yet it it inconclusive whether in the aggregate September really led to more of these match-ups where the batter was relatively inexperienced with the hurler on the mound than in other months of the season, given that most hitters had gathered more experience with the set of major league pitchers over the season. A further study of air density and player talent level by month would be interesting to uncover other potential theories not addressed in this study.

References & Resources

Jon Roegele is a baseball analyst and writer for The Hardball Times. He was nominated for a SABR Analytics Conference Research Award in 2014 and 2015. Follow him on Twitter @MLBPlayerAnalys.
Newest Most Voted
Inline Feedbacks
View all comments
8 years ago

I have a very non-scientific theory about the overall HR and run-scoring bump in August. Perhaps it was fueled primarily by two teams that got significant psychological and talent bumps at the beginning of the month as the trade deadline passed and their GMs made a flurry of moves. Those two teams — Toronto and the Mets — are also the highest run scoring and HR-hitting teams in the majors in August. Toronto being first in both cases, the Mets being second. Now Toronto is likely always going to be one of the highest run scoring, HR-hitting teams given their lineup and home park, but for the Mets, that’s just the opposite. The Mets prior to August were one of the worst offensive teams in all of baseball. So the Mets alone probably did more to bump up the overall numbers in August relative to the rest of the year than the Blue Jays did. But both teams fueled the bump.

As a fan who watched the Mets closely in August, I can attest that the team was playing on a psychological “high” all month long. The moves made by the front office at the end of July seemed to infuse the team with extra energy, confidence and purpose. Combine that with the fact they played at the end of August at Coors and Citizens Bank Park and – bam! – you have a recipe for explosive run scoring. Had they played those same series at Citi Field, I doubt the run scoring would have been half as much.

In terms of this overall trend and all teams, maybe it also has something to do with there being a lot of younger pitchers in the majors this year (seems that way to me) and they’ve been hitting innings limits and career highs in innings pitched which started to come to a head in August? Even if FB velocity didn’t wane, maybe their fine control did, as well as the break on their breaking pitches? This situation would of course be alleviated in September with expanded rosters and the opportunity to rest these starters more or skip their rotation turns entirely. Just a thought.

8 years ago
Reply to  Metsfan

I would throw the Cubs offensive surge in there as well. In fact just the Mets and Cubs surge in offense, both of whom had struggled offensively prior to Aug. could help acct for it.
There was an unprecedented amount of deadline trades this year. That also may have something to do with it.

8 years ago

This was a fascinating read.

Mike Green
8 years ago

Here’s one possible explanation for the lower run-scoring environment in September. The expanded rosters allow managers to give many pitchers short outings more in certain games without loss of effectiveness and without wearing out the bullpen. Take last night’s game here in Toronto. Trevor Bauer struggled quite a bit in both the first and second innings, and trailed 4-0 when he was pulled in the second. Ordinarily, Francona would have gone longer with him, but with a 13 man bullpen, he was able to run out 6 or 7 pitchers for 1 inning each without using his high leverage relief. He could not have done this in August; if he had pulled Bauer, he would have to run out a couple of pitchers for 2-3 inning stints. The approach worked.

Many teams now have interchangeable parts at the back end of the bullpen. Some are sitting in triple A for most of the season, ready to give a good low leverage inning at a moment’s notice.

8 years ago
Reply to  Mike Green

I like this theory. Also, expanded rosters allow for a lot of rookie position players to be brought up, and on the teams that are already out of playoff contention these players have a chance to get a lot of play time. Naturally, they don’t usually do very well.

8 years ago

You totally glossed over this, but how in the world is there no relationship between BABIP and run scoring? That seems totally counter-intuitive to me.

Voros McCracken
8 years ago
Reply to  Ben

Where should I start?

8 years ago

At the beginning.

Eric the Clown
8 years ago
Reply to  Ben

The variance in BABIP is so small, compared to HR/FB. And BABIP doesn’t include home runs. I would imagine the effect of BABIP is simply dominated by the effect of HR/FB.

8 years ago
Reply to  Ben

For one, because BABIP is inversely-correlated with HR/FB.

8 years ago

That’s definitely not the way you want to measure the impact of temperature… You need to figure out the relationship between temp. and run scoring on a per park basis. With some parks, it is huge; with others, higher temps. correlate to high humidity, wind blowing in, and other run-suppressing variables.

Taking one-month chunks of an unknown set of parks and calculating avg. overall temp. is not going to show what the true effect would be if we lowered the temp. by 10-15 degrees at all parks and measured the impact. I’ve done the research, and temp. is, in fact, a significant part of why Sept. (and October) have reduced run scoring.

Pirates Hurdles
8 years ago
Reply to  MP

High humidity is not “run suppressing”. Water vapor is lighter than molecular oxygen or nitrogen gas.

Jon Roegele
8 years ago
Reply to  MP

Thanks for the comment. I think we are basically saying the same thing. You’re saying that as temperature changes, many other conditions change as well; some of these make positive contributions to the offensive environment, some negative. What these factors are could certainly depend on the location where the game is being held.

In the article I mentioned a couple of times that while temperature on its own does not seem to drive change in the offensive environment, looking at the larger picture (I mentioned air density) could very likely yield different results. I would not be surprised if they do result in more of an impact on run scoring; and you say your research has shown that to be true.

Is your research on this topic in a public location for which you can share the link?

Whether it is weather or not
8 years ago

Very respectable analysis. As you astutely noted, temperature is not the only weather variable that could have an effect. In fact, humidity and air pressure both can be huge factors. I believe that at Coors Field they started adjusting the humidity of the balls to make them softer thereby hoping to decrease home run production.
On the other side of humidity, contrary to how we feel, humid air is less dense than dry air– which should reduce spin on pitches and increase both pitch speed and batted ball distance.

A quick web search did not provide any summary of average barometric pressures, but I will keep looking.

Peter B
8 years ago

There’s a lot to go through there, but one thing that stands out to me is about the “times faced” analysis. Pitchers decline over their career while batters improve and then decline, and so you’d expect wOBA to rise and K% to fall as the “times faced” increases: the pitcher is (on average) declining in skill with every one of those matchups.

8 years ago

Please note that “impactful” is not a real word but jargon that has migrated from bad business writing. Though I suppose this is in some sense a losing battle.

Jon Roegele
8 years ago
Reply to  Hedgehog

That’s interesting! I admit I did not know that, but just read a little about it. It is in all the dictionaries now, for what it’s worth.

8 years ago
Reply to  Hedgehog

Stay strong. I’m a copy editor. I fight those battles every day because somebody has to.

While I’m at it, when did every damn mundane thing become an “event”?

M. Fiers
8 years ago
Reply to  bucdaddy

…or historic?

8 years ago

Water vapor is indeed less dense than dry air; however, the difference between humid air and dry air is so small that it has negligible effect on the flight of a ball (because 99% of air is N2 and O2, a change in relative humidity from 0% to 100% reduces air density by less than 1%, and of course the humidity range in the real world is even less than that). When it comes to air density, Coors is all about altitude, not the humidity.

What humidity does affect is the ball itself and (to a lesser extent) the bats. Damp ball and damp bat produce a less elastic collision than dry ball and dry bat, and that means less force on the ball thus less distance in flight. That’s why Coors has humidor for the balls; that’s why some batters like Ichiro take care to not let their bats get damp.

Moreover, as a ball absorbs atmospheric moisture it gets slightly larger and heavier, which may make it easier for pitchers to grip, and slightly more difficult for batters to drive; however, once they do drive it, a higher weight means it will have more momentum to overcome atmospheric drag. But by far the largest effect humidity has is on the elasticity of the ball — or, more precisely, its Coefficient of Restitution (COR). This has been investigated experimentally by Alan Nathan et al, and I quote from their conclusion:

If the relative humidity is increased from 30 to 50%, the cylindrical COR decreases by 0.024, which is about 4.5%. We assume a similar decrease occurs at the higher speeds of the ball-bat impact. For a typical Major League bat, pitch speed, and bat speed, we estimate a decrease in batted-ball speed by about 2.5 mph, corresponding to a decrease in fly ball distance by about 14 ft. Adair estimated that each percent change in fly ball distance changes the probability of hitting a home run by about 7%.4 Taking 380 ft as a typical home run distance, a
reduction of 14 ft corresponds to a reduction in home run probability by about 25%

8 years ago

Joe, Joe, Joe (sigh)….

Of course, temperature and specifically air density correlate HEAVILY with HR/FB and run scoring in general. You know that! Have you ever watched a game when the temperature is 40 degrees versus when it is 95? Have you ever seen the Vegas totals? Come on, everyone knows that the difference in run scoring between those two extremes, for example, is enormous. Why would you even say that run scoring and temperature have no relationship?

The reason you are getting no relationship in your chart is because you are mixing up parks. You would have to control for the park to see the correlation. Do the same chart but on one axis do the difference in temperature between the average temp at that park and the game time temp. Then on the other axis, do the difference between runs scored for that game and average runs scored per game in that park.

And, as you also know, correlation does not tell us the magnitude of the effect. It just tells us whether there is an effect and how much noise there also is. The relationship could be 1 run per every degree with a very small “r” or it could be 1/100 of a run per degree with a very high r. You also know that! The “r” is also a function of the underlying sample sizes.

In fact, the relationship between temperature and run scoring is around .15 runs per 10 degrees. Not exactly linear (it flattens out especially in colder temperatures) though.

Jon Roegele
8 years ago
Reply to  MGL

I controlled for park like both you and an earlier commenter suggested, and posted the resulting graph here:


It showed an effect of 0.03 runs per degree. And was statistically significant. So yes the effect is there 🙂

If I look at the (game temperature – average game temperature per park per year) and average these per month, I get the following:

April: -10.14
May: -3.32
June: +2.35
July: +6.18
August: +4.66
September: -0.04

So September is basically the “average” month of the season as far as overall temperatures are concerned. Of course this doesn’t mean that it is “average” as far as air density, or wind, etc.

It’s this last point that made me think that temperature wasn’t having a large impact on September scoring vs. the rest of the year.

8 years ago
Reply to  Jon Roegele

I saw that. Good job! I was still surprised at the low r, but that doesn’t really mean anything other than there are a lot of other reasons, including randomness, that run scoring varies.

If you did nothing else but group games together according to temperature intervals, you would get a much higher “r” even though the relationship is the same.

One has to be very careful in attaching any meaning to a correlation when the underlying sample sizes for each element in the distribution is small, in this case, one game.

For example, I don’t imagine that that correlating BA for players one AB at a time is going to yield a correlation much above zero, but year to year, it will probably be near .5.

8 years ago

OK, it hasn’t been especially warm in August and the strike zone hasn’t changed. That pretty much means it’s probably just a lucky month, which happens all the time. There are a lot of months in the history of baseball.

Jetsy Extrano
8 years ago

The “times faced” graph is intriguing, but I’d be fascinated to see it with correction for the issues that you and others point out.

To correct for in-game TTTO, is there enough data to split it by first time in the game, etc.? Or subtract the overall TTTO numbers to compensate.

For aging curves, could also compensate for average aging… or look at the subset occurring within a certain number of years limit, and compare.

For selection, filter the sample so the pitcher population is the same across the graph.

I’d be surprised if there’s much left over, but it would be a neat find!

8 years ago
Reply to  Jetsy Extrano

Right, the reason for the weird results in the “familiarity” graph is just what Joe alluded to – different pools of batters and pitchers. As with temperature and run scoring, you can’t do it that way.

8 years ago

I was wondering when somebody was going to notice the scoring bump, but the only reason I thought of it at all is that it seems like for awhile now every night somebody scores 15 or 18 runs.

I’m going to toss out that maybe it was just a number of outlier games like that that did it. Like last night, KC and Wash both scored 15 runs. So is it that all teams went from playing 4-3 games to playing 5-4 games, or is it that most teams are still playing 4-3 games but a few put up a number of 15-3 games, and that made it look like scoring OVERALL was up?

8 years ago

Is there any research on whether lower strike zone is umpire-created or batter-created? Seems to me that the post-steroids era has a bunch of absolute hacking going on at the plate with batters. Noticed it about 10 years ago – and has become even more so. I know this used to be a stereotype about Latin players in the late 80’s and early 90’s (swing at every pitch as if it means losing machismo to let a pitch go by). And the steroids era definitely changed many batters approach – swinging hard for the fences. None of which has really changed with a new generation of batters who only ever saw the power bats as kids.

Did umps start calling strikes lower because batters were swinging down there anyway? Or was there some ‘decision’ somewhere to start calling lower strikes?