Title: Does Preseason Matter?
Date: October 29, 2012
Original Source: Hoopdata
Synopsis: My debut at Hoopdata looked back at the last six years of data to see if a team’s preseason performance is correlated with its regular season performance. I also looked at whether a perceived improvement in preseason can predict a regular season improvement as well.
Preseason doesn’t matter. Or say they say. And so Ray Allen says.
This axiom has been long held, and it makes some sense. Players on new teams are learning to play with each other, coaches are experimenting with schemes and line-up configurations, and stars are playing smaller minutes at lower effort levels to ensure a healthy start to the regular season.
At the same time, it feels like there is anecdotal evidence of signs from the preseason being strong indicators of performance changes in hindsight.
Looking at player stats, Kevin Pelton of Basketball Prospectus recently wrote a piece that deemed the preseason to be neither completely meaningless nor completely useful. Pelton summated his findings and his feelings as such: “It matters, but it’s not nearly as meaningful as a player’s track record. What the numbers indicate is that we ought to approach the preseason, like anything else, in Bayesian fashion. When the results are consistent with what we already know, they strengthen our conclusion.”
As a Toronto Raptors supporter, I’m more interested in present at what preseason results can tell us about team performance. My thought was that if I looked at it, I’d find largely the same result as Kevin did with players – there will be some correlation between preseason success and team success, but a lot of it will just strengthen our preconceived notions. That is, the strong preseason play of Chicago shows that they can be good even without Derrick Rose in the interim, while the poor play of Orlando confirms their status as cellar dwellers for this year.
With teams, there are more confounding factors in winning percentage than with player stats, as outlined above. But still, I wanted to see if the 6-1 record the Raptors posted this preseason is a sign of things to come or just a red herring in a small sample pre-season where the Raptors were the rare team playing all of their regulars a fair amount.
I pulled pre-season data from 2005-06 to 2010-11 (I left out last year’s preseason since the lockout shortened it to two games per team), and compared it to regular season performance from 2004-05 to 2010-11. What I was looking for was to see if pre-season record correlated with regular season record. I also wanted to see if changes in performance level (that is, a team performing much stronger in the preseason than they had in the previous year’s regular season) correlated with changes in actual year to year winning percentage.
Not surprisingly, the correlation between pre-season winning percentage and same-season regular season winning percentage wasn’t that strong, with a correlation of just 31.7% and an R2 of just 0.1. So yes, preseason numbers are correlated with regular season numbers, but the relationship is not strong. This data and this chart tell us little about what to expect from surprising preseason performances, since they can easily be chalked up to the randomness of a small sample size or the caveats outlined earlier such as player usage and team experimentation.
I also wanted to see, though, if just changes in performance could be signalled by improvements or setbacks in the preseason. I took the difference between a team’s preseason record and their regular season record in year n-1 and compared it to the change in their regular season records. My hope was to see if a big jump in the preseason signalled a coming improvement in regular season performance or vice versa for teams that would struggle. This time, there was a slightly higher correlation at 38.9% with an R2 of 0.15. This still isn’t strong, of course, but it at least shows that preseason changes aren’t entirely random.
So no, the preseason is not entirely irrelevant. If your team was unexpectedly good, you may have reason to be hopeful. But if your team was unexpectedly bad, there is plenty of room for you to reasonably ignore that fact (Lakers fans) since these relationships are anything but pronounced.
For this year, the team’s with the biggest drop in performance from 2011-12 regular season winning percentage and 2012-13 preseason winning percentage were as follows. This doesn’t mean they’re doomed by any means, but there is reason for concern, or at least reason to go back and examine the preseason game tape or box scores to see if there were some obvious reasons. Reasons like Metta World Peace saying the Lakers’ 0-8 record is just the team showing love to their fans.
In terms of reasons for optimism, the Raptors and new additions Kyle Lowry and Jonas Valanciunas (among others) top the list after a 6-1 preseason. Despite the objections of DeMar DeRozan, even the most negative projections are showing some amount of improvement for the team (John Hollinger’s 33-win prediction would actually be a gain of four wins given last year’s prorated record). While they’re unlikely to be an .857 team, my prediction of 38-44 seems completely reasonable as only two of the 14 teams in this sample to have a preseason winning percentage of .857 or better finished below .500 that year.
Here we see that the Warriors, Kings, Sixers and T-Wolves also have some cause for cautious optimism. Of note is that of the top-20 teams in this sample in terms of preseason improvement over the previous year, only three failed to improve their actual regular season record from the season before.
So yes, feel free to get a little excited about a strong preseason. Don’t go overboard and bet the mortgage on Vegas over/unders, but safely bet your friends your team improves on last year’s record at least a bit. As for the laggards, well, there are plenty of counter-examples to this relationship, so take the advice of Metta and Ray and just cross your fingers.