The Question of Momentum

After clinching the division on September 21st, the Twins finished the season with losses in 8 of their final 11 games and were outscored 72-41.  Twins fans are used to exhilarating regular season finishes followed by playoff letdowns (see 2006 and 2009), but this unusually lackluster ending to the regular season has caused some to worry about the Twins' playoff chances.  Of course, it's easy to point out that the Twins were resting some of their key players (Joe Mauer, Jim Thome, J.J. Hardy) for many of those games, but the pessimist might point to the 9 total runs scored in the Toronto series, when much of the playoff lineup was intact, as cause for concern.

Normally, I would dismiss an abstract notion such as momentum, especially in this case where it was sacrificed in order to get the Twins as healthy as possible for the playoffs.  Rather than simply assuming momentum was irrelevant, I decided to see if the data agreed with me.  The question I sought to answer was whether momentum--measured by performance in the final 10 regular season games, as it was the easiest data to find and collect--has a significant impact on a team's performance in the Division Series.  Some other data that I collected that I thought might be correlated was overall regular season record, whether the team had home-field advantage in the playoffs or not, and the approximate clinch date for the team.  (It would have been too hard to find actual clinch dates, so I used games ahead of a playoff spot at the end of the season. It's not perfect, but hopefully it's good enough.) I used data from the last nine seasons, from 2001-2009.

The first thing I noticed about the data set is that only one other playoff team won as few games as the 2010 Twins did in their last 10:  last year's St. Louis Cardinals went 2-8 in their last 10 games and were promptly swept out of the NLDS.   One data point hardly constitutes grounds for any sort of conclusion, though. It seems obvious that playoff teams would rarely finish so poorly in their last 10 games, since those teams have to be pretty good in the first place (except for maybe the 82-win Padres of 2005).

The next two things that stood out to me are unrelated to the main theme of this post, but I felt they deserved mention anyways.  The first is just how good the 2001 Seattle Mariners were.  In Ichiro's first year (and MVP year), the Mariners won 116 games and were a full 31 games away from worrying about a playoff spot.  The next farthest ahead was the 2008 Angels, who won 100 games while taking advantage of a weak division in which no other team even went .500 (the A's finished with 79 wins).  The other interesting note is that there has not been a 5-game Division Series since 2005; of the 16 division series in the past 4 years, not one has reached a fifth game.  Really, the baseball postseason as a whole has been lacking for a few years, with most of the recent World Series reaching only four or five games.

Okay, now back to the issue of momentum.  The first problem I ran into was how exactly to quantify Division Series performance.  My two ideas were either a binary variable for advancing or not (1 for advance, 0 for not advance) or just a variable denoting the number of wins in the Division Series.  I started off by testing if just wins in the last 10 games had any correlation with the variables of interest, but the regression showed no such relationship.  The R-squared for both the binary variable and Division Series wins was under 2% with P-values well outside the area of significance.  I didn't want to just stop there and conclude that momentum is meaningless, though.

The next step was to check if any of the other data was correlated with success in the Division Series.  Regular season winning percentage, for example, had a P-value of 0.041 in a regression with Division Series wins, so regular season wins could at least be a significant factor in postseason performance.  Note: it was not correlated significantly with the binary variable 'advance', and I decided at this point that a binary variable was not really the best option for a dependent variable, so though I kept checking it to be sure I won't be mentioning it any further.

So it seems that, as I presumed in the beginning, the last 10 games represents a small sample size that has little bearing on Division Series performance, especially in a case such as the Twins' where results are skewed due to worse players getting extra playing time.  Nonetheless, I ran through a few other regressions with the other variables (games ahead and home field) to see if any interesting patterns emerged.  However, none of these variables had anything close to a significant impact on the number of Division Series wins.

In fact, the only thing last 10 wins, home-field advantage, and games ahead did have in common was that they were all correlated with regular season wins.  Thus, better teams that win more games tend to win more of their last 10 games, have home-field advantage more often, and be farther ahead in the standings.  There are certainly reasons you can give as to why the Twins might not win this series, but their sluggish finish to the season should not be one of them.  The only thing that definitely helps a team win games in the Division Series is being as good a team as possible, so the Twins did the right thing by resting their players (and maybe costing themselves a few wins in the process) in order to be at their best for the playoffs.