Roster Expansion and September Hitting

Over the last few weeks, several players who have had pretty lousy years have gotten hot at the same time. For instance, here are the September lines for some various players:

Ichiro Suzuki: .424/.452/.593, 207 wRC+
Carlos Pena: .293/.473/.512, 172 wRC+
Gaby Sanchez: .311/.436/.556, 170 wRC+
Justin Smoak: .345/.419/.527, 165 wRC+
Justin Upton: .312/.365/.558, 147 wRC+
Russell Martin: .246/.358/.491, 136 wRC+

Ichiro was essentially give away by the Mariners at the trade deadline (at his request), a sign of just how far he’d fallen from his time as the franchise icon. Pena, Sanchez, and Smoak all played themselves out of jobs earlier this year. Upton struggled to the point that Arizona made him available for trade talks, and is widely expected to move him this winter. And, while Martin has shown decent power for a catcher, his average has hovered around the Mendoza Line all season.

When you mention that these kinds have had a good run the last few weeks, there are generally two responses:

A. Small Sample Size, which, well, yes, of course it is.

B. September hot streaks should be discounted because of roster expansion, as inferior non-MLB pitchers are taking the hill and skewing offensive numbers around the sport.

There’s no arguing with the first point, as any monthly split is going to be SSS, and as such, subject to massive swings in variance with minimal predictive value. The best hitter in August (min 50 PA) was Joaquin Arias, after all. Trying to ascertain anything from samples of 50-100 plate appearances is a fool’s errand.

However, on the second point, I’ve never actually seen anyone show that offense in September spikes up dramatically after rosters expand. Intuitively, it makes some sense that more bad pitchers should lead to more runs, but at the same time, more pitchers available leads to more match-up opportunities and more aggressive bullpen usage, which should lead to fewer runs scored. So, I decided to look at the last 10 years, and see whether the sport saw any kind of noticeable change in September offense.

2003 0.264 0.333 0.422 0.328
Sept. 0.262 0.330 0.416 0.325
2004 0.266 0.335 0.428 0.330
Sept. 0.261 0.330 0.419 0.326
2005 0.264 0.330 0.419 0.326
Sept. 0.261 0.330 0.413 0.324
2006 0.269 0.337 0.432 0.332
Sept. 0.265 0.332 0.427 0.328
2007 0.268 0.336 0.423 0.331
Sept. 0.274 0.343 0.435 0.339
2008 0.264 0.333 0.416 0.328
Sept. 0.266 0.336 0.420 0.331
2009 0.262 0.333 0.418 0.329
Sept. 0.262 0.333 0.408 0.325
2010 0.257 0.325 0.403 0.321
Sept. 0.250 0.319 0.390 0.314
2011 0.255 0.321 0.399 0.316
Sept. 0.255 0.322 0.405 0.319
2012 0.255 0.319 0.405 0.315
Sept. 0.254 0.320 0.398 0.314

In seven of the last 10 seasons, league wOBA in September has been lower than the season average overall, this year included. There were jumps in both 2007 and 2008, but those were followed by equally large declines in 2009 and 2010, so it’s hard to say that there’s any kind of trend toward increasing offense in the final month of the season.

It’s possible that looking at league wide numbers is obscuring a real effect, and some guys are generating a real benefit from facing pitchers in September that they wouldn’t get to face earlier in the season. However, for every team that is out of the race and giving experimental looks to guys called up from the minors, there is another team that is desperately trying to win every game they play, and now has the advantage of a deep bullpen of situational arms with which to play the match-ups. Over a month’s worth of playing time, these things probably don’t even out for every hitter in baseball, but there simply aren’t enough scrubs throwing lots of innings around the sport to substantially alter the run environment in the final month of the season.

For the guys closing out bad seasons by finishing strong, we should be skeptical of their performance because it was relegated to a single good month, but we probably shouldn’t be any extra skeptical of that performance simply because it came at the same time as the rosters expanded. There might be a correlation between larger rosters and the timing of when these guys got hot, but that doesn’t mean that the larger rosters caused the hot streaks.

Print This Post

Dave is the Managing Editor of FanGraphs.

17 Responses to “Roster Expansion and September Hitting”

You can follow any responses to this entry through the RSS 2.0 feed.
  1. Julian says:

    Aren’t the numbers skewed against offense, too, because there are also “inferior non-MLB” hitters are batting? Wouldn’t it be better to look at AVG/OBP/SLG/wOBA for hitters that were already in the majors before September?

    +29 Vote -1 Vote +1

    • Tyler says:

      This is smart.

      Vote -1 Vote +1

    • Tomcat says:

      Exactly, this is the problem with the “steroid era” being blamed for all of the crazy numbers of the late 90s early 00s it ignores the elite pitching numbers of the Pedros, Clemens, and Madduxs of the world. Individual HOF type players like Barry Bond’s had great seasons and there were occaisonal Bret Boone or Brady Anderson types but league wide it isn’t like HR/PA spiked. This study ignores that rookie hitters and batters are being added at the same time. Looking at the Sept splits of elite batters seems like a better plan.

      Vote -1 Vote +1

    • AustinRHL says:

      Yes, this seems blindingly obvious. I have no idea why Dave didn’t do this.

      Vote -1 Vote +1

    • AustinRHL says:

      On the podcast today, Dave explained his reasoning as being, basically, “well, September callups are mostly pitchers, so much more playing time goes to the pitchers than to the hitters.” Obviously, the strawman-like point he’s arguing against is still basically wrong, and he’s probably right that Major League hitters don’t see their numbers go up much in September. But I just don’t think he has any excuse for making such a gigantic assumption.

      Vote -1 Vote +1

    • Matthias says:

      Maybe looking at the splits for only the hitters on teams with a winning record on September 1 (or within X games of a playoff spot) would show us something…

      Vote -1 Vote +1

  2. Ian R. says:

    Interesting study, but the data might be skewed due to the inclusion of the September call-ups themselves. It makes sense that a bunch of minor league players getting a fair number of plate appearances against major league pitchers would cause offense to drop, especially if you believe that the gap between AAA and MLB pitchers is bigger than the gap between AAA and MLB hitters.

    I’d like to see two more versions of this study, one showing offensive performance for the league excluding hitters who were added in September (but including PAs against pitchers added in September), and one showing pitching performance for the league excluding pitchers who were added in September (but including PAs against hitters added in September). I’m guessing that data would be a lot harder to compute, though.

    Vote -1 Vote +1

  3. drewggy says:

    Needs to be re-run for only players who had 200 PA (or some number) pre-September.

    Vote -1 Vote +1

  4. Taylor says:

    Good piece, and the previous commenters make helpful suggestions for next steps. I would add that looking at measures of variance/dispersion would help to contextualize these numbers. Simple chi-squared tests can indicate if the variance is significantly different in September than in the rest of the season. If variance increases but the mean stays the same in September means are the same, then you have a decent indicator that the “win at all costs” and “play the kids” approaches actually are happening, and are canceling one another out.

    Vote -1 Vote +1

  5. DTF_in_DTL says:

    Expanded rosters also mean fewer plate appearances to go around. Doesn’t pinch hitting and a lack of consistent abs tend to drag down offense as well?

    Vote -1 Vote +1

  6. Ian R. says:

    It’s also worth noting that injuries may have a bigger impact in September than in other parts of the season, for two reasons: Many good players (especially good, young pitchers) are shut down or given extra days off in September to preserve their health, and many other good players play through injuries for teams in pennant races instead of yielding to a healthier (and potentially more effective) backup.

    Vote -1 Vote +1

  7. PackBob says:

    Thank you, thank you, thank you.

    Many just seem to assume that September call-ups create offensive advantages and that teams out of the playoff picture have the tendency to dog it, giving a huge advantage to playoff teams that play them. Ask Detroit about that idea with reference to Minnesota.

    One thing doesn’t change a bit: on any given day a bad team playing well can beat a good team playing poorly. Scheduling inequities have a much bigger impact on standings than September call-ups, which overall may have little to no impact.

    Vote -1 Vote +1

  8. Adam Stein says:

    Observational data, not backed by any statistical analysis.

    If there’s any effect, it’s going to be very very small probably too small to see. Half the teams are in playoff contention, at least at the start of September. A lot of non-contending teams seem to feel an obligation to play their best team against contending teams. For certain in the first group, and likely in the second group, those teams are using expanded rosters to improve the quality of their pitching. As Dave noted, playing matchups, and I’d think resting their best pitchers in blowouts and I think getting quality innings from players whose ETA is “next year”. For every crappy starting pitcher who gets hammered, there’s an Erasmo Ramirez who is pretty good in part because none of the hitters have seen him before.

    So really you’re talking about games in which both teams are out of playoff contention and one happens to be using a AAA pitcher. While a few pitchers get shut down in September, it’s not like every team is shelving their top 3 starters and best couple of relievers to go with a AAAA or AAA pitching staff.

    It’s just small sample size.

    Vote -1 Vote +1

  9. channelclemente says:

    What a silly article.

    Vote -1 Vote +1

    • Baltar says:

      I’m with you, cc.
      I’m embarassed that I actually read the whole article and the comments.
      I guess I was really bored.

      Vote -1 Vote +1