Do NBA Preseason Games Matter?
October, 2007

 

Every year during the NBA Preseason we hear phrases like: “Who cares about our record, the preseason means nothing”, or “We’re 6-1, this is going to be a good year!”

Of course, technically, preseason games don’t matter one bit.  The goal of each team, each season, is to win a Championship.  To win a Championship, you must first get there, and that means making the Playoffs, which is directly based on regular season records, and absolutely nothing to do with preseason records.  So, any team can theoretically win zero preseason games, and then go on to win a Championship.  Conversely, any team can win all of their preseason games and then go on to flop in the regular season.

With the 07-08 preseason now coming to a close, what I would like to explore with this article is just how much we can take from preseason games. However, rather than just making casual observations, I will attempt to take a scientific look at the numbers.

Before we look at the numbers though, I would like establish what types of numbers I will be working with and a few notes about them.  I will be working with correlations, which, as quoted from Wikipedia, “indicates the strength and direction of a linear relationship between two random variables.”  A correlation will range from –1 through +1 and is stronger the closer it is to -1 or +1, and weaker the closer it is to 0.  Further, a correlation of -.67, for example, is the same strength as a correlation of +.67 -- it just signifies a different relationship.  In the instance of +.67 the two variables show a positive correlation, or as in this case, high preseason winning percentages are paired with high regular season winning percentages and low preseason winning percentages are paired with low regular season winning percentages.  Conversely, in the case of -.67 the two variables show a negative, or inverted, correlation in which high preseason winning percentages tend to be paired with low regular season winning percentages and low preseason winning percentages tend to be paired with high regular season winning percentages.

I began by compiling preseason and regular season records starting with the 06-07 season and going back through the 99-00 season.  With this I was able to get a pretty good sample size (n=235), and stop at the disaster lockout season in 98-99.

The overall correlation between preseason records and regular season records over the course of those eight seasons is .35, which is considered a moderate positive correlation and evidence that preseason games might not be as completely worthless as some would like to suggest.  Below is a table showing the correlation between preseason and regular season records year by year, and the total over the course of all seasons, for all teams.

Over the course of these seasons we see a relatively stable correlation, with the exception of the 02-03 season in which we see a very strong correlation, and last season, 06-07, in which we see all but zero relationship between preseason and regular season records.

Well, those are some nice numbers and all, but if you’re like myself, you’d like to see some more.  Surely there are some stronger numbers in there somewhere, so let’s see what we can come up with.

I decided to divide regular season records into three categories and see how they correlated with their corresponding preseason records.  The categories are essentially bad teams, average teams, and good teams.  These were determined solely by the number of wins, with “bad teams” being those with 35 wins or less (n=77); “average teams” being those with 45-36 win (n=75); and finally “good teams” being those with 46 wins or more (n=83).

What we see here is interesting:  Teams who posted bad and average regular season records, tend to have posted quite similar records in that years preseason.  However, teams who did very well in the regular season tend to show very little correlation with their record in that preseason, incidentally even showing a slight negative correlation.  This may surprise some initially until you consider the circumstances.  First, teams who do very well in a given season typically did well in the prior season, which means on average they likely had a longer season than other teams, and a shorter off-season.  Secondly, teams who do very well also, on average, tend to be older veteran teams, and with that come veteran legs.  When these things are viewed in combination with one another, it’s really no surprise to see the relationship being rather erratic, showing little to no correlation.  These are teams built for the Playoffs, so for them, preseason games, and sometimes even early regular season games, matter very little in the grand scheme of things.

It’s nice to be able to look at a team’s regular season record and see how it correlates back with it’s preseason record that year, but what would be even better is if we could use preseason records as an indicator to look forward to the upcoming regular season.

For this, I divided preseason winning percentages into three categories: bad, those with a winning percentage of .375 and below (n=84); average, those with a winning percentage of .400-.571 (n=74); and finally, good, those with a winning percentage of .625 and above (n=77).

What we see is less than promising for anyone hoping to make some bold predictions using preseason records.  Each class of preseason winning percentages, individually, correlates only mildly with the upcoming regular season, and with no real significant differences between each class.

Moving on, I decided to go even deeper and draw out the very best and very worst preseason records and examine how they tend to correlate with their corresponding records in the upcoming regular season.  For this, I determined the best preseason records to be those with a maximum of one loss, which left us with a total of 14 teams (n=14), and, as you might guess, the worst preseason records were determined to be those with a maximum of one win, which left us with a total of 10 teams (n=10).

Upon drawing out the very best and very worst preseason records, we can now see a much stronger correlation.  The very best and the very worst records in the preseason tend to have a relatively strong correlation with the record that team will post in the regular season, however, the small sample size here should be noted.

For a more detailed look:

One interesting note is that, at least as far back as the 99-00 season, no team who has lost, at the most, one preseason game, has gone on to win less than 42 games in that upcoming season.  Teams who would fall into this category this year would be the Atlanta Hawks and the Orlando Magic.

As for teams who fall into the second category (the worst preseason teams), this year we have three teams: the Philadelphia 76ers, Cleveland Cavaliers and Miami Heat. Incidentally, the Cavs were also on this list year and the proceeded to go on to win 50 games and play in the NBA Championship. Also, as further illustration that anything can happen, this marked the first year since at least the 99-00 season in which a team (Miami) failed to win a single preseason game.

The last thing I would like to take a look at is how each team’s individual preseason record has historically correlated with their regular season record.

With that said, here are the team by team correlations between preseason and regular season records, from 99-00 through 06-07:

The first thing that sticks out to me is that, with the exception of the Miami Heat, all of the teams who have won an NBA Championship (Lakers, Spurs, Pistons) in this time frame all show little and/or a negative correlation with respect to preseason and regular season record.  This is consistent with the above table which shows that teams who win 46+ games (good teams) show a slight, but negative, correlation with preseason records, while teams who win <46 games (average and bad teams) show a relatively strong correlation with preseason records.  Also, I can’t help but noticing, right after I mentioned Atlanta above, here they are topping the list with the highest correlation between their preseason and regular season record.  In each of the past three seasons, the Hawks have shown improvements in both their preseason records and corresponding regular season records, and yet again this season, have improved their preseason record.  Perhaps this is the year the Hawks finally turn the corner.  It would certainly be consistent.

In closing, I want to again note that none of these numbers suggest any direct causal relationship, and should not be interpreted as such.  For example, the Hawks are by no means guaranteed to go on to win at least 42 games this season, however, if they didn’t, based on their preseason record, they would be the first team not to. Hopefully, at the very least, this article provides some numbers to go along with all the talk that surrounds preseason games and their value.