Does Preseason Matter? That Depends On Your Belief In Chance.
Plus, the one certainty (and lots of fun trivia statistics)
Hey all!
Preseason has officially kicked off for the NBA — and a few overseas teams, too — so this is the time when most media outlets start to overreact to every little thing that happens.
Whether by way of basketball starvation or clickbait, outlets will ask questions as to whether Jaylen Brown’s 33% shooting night on Friday forebodes a terrible future for the Celtics, Hunter Tyson’s high-efficiency play for the Nuggets means that he’ll be one of the best bench players in the league, or Nikola Jokic will win MVP and the title this year (he might actually pull that one off).
So, in the spirit of one of my pieces from last year in which I studied the predictive value of Summer League (it’s more than you think!), I’ve decided to find out whether or not preseason actually means anything. After compiling the data, I’m pretty confident that the answer is, uh, sort of, as you’ll see.
We’ll look at wins and losses mainly, but we’ll cover other statistics later down the line — let’s get into it.
Note: This is only going to cover team-wide statistics since 2014-15, as I haven’t had the time to scrape through the massive swaths of player data — maybe at a later date, though.
What It Means to Win
When we see our favorite teams win in the preseason, we often assume that they’ll carry it over into the regular season. However, when our team loses in the run-up to the real show, we suddenly start to claim that preseason doesn’t matter at all when it comes to wins and losses. This, essentially, is a combination of confirmation bias and self-serving bias, which both look to appeal to what we want to think about a situation. This flip-flopping isn’t exactly statistically correct, though — as it turns out, the preseason means little for win counts; at least, not on the surface.
In statistics, when you do a correlation test, a value of 0.3 or above is generally considered to mean that the two statistics you are comparing (in this case, preseason wins versus regular season wins) have a correlation of some sort. Anything below this is considered insignificant.
When you take a look at the win counts of preseason teams over the past decade, you’ll find that the correlation is 0.27, which is close to having some level of correlation but doesn’t quite fit the bill. Losses, of course, have the same (but negative) correlation, as winning a game eliminates the opportunity for a loss and vice-versa.
Yet, this low correlation doesn’t mean that preseason wins are irrelevant all the time. Case in point, the Timberwolves finished with the highest win total last preseason, which led to a historical season for the franchise. In the 2018-19 preseason, the Raptors and Warriors were 2nd and 3rd, respectively, in preseason winning percentage, and they met in the Finals that year. The biggest bucking of the correlation trend can be found in the 2019-20 season, where teams’ win totals had a 0.7 correlation with their regular season counterparts.
The point is that the correlations often shift and change due to team volatility, making it difficult to predict who will win in the preseason and, crucially, when those wins matter in the regular season. Just take a look at the volatility of the correlation values between preseason and regular season wins over time and you’ll see what I mean:
So what can we learn from the win totals in the preseason, then, if everything is so volatile from year to year? Well, that depends on how willing you are to play a game of chance. During my deep dive, I found something quite interesting related to how the NBA’s probability patterns work. To test it, I took a look at every team to finish in the top 5 in preseason win counts and tried to find out what percentage of them finished in the top 5 in the regular season. At first, I wasn’t too impressed, as it turned out to be a little under 33%, meaning that less than 1/3 of top-5 preseason squads carried it over to the regular season.
Here’s the kicker, though: That’s way better than the chances of any team making it into the top 5. Based on the NBA having 30 teams, around 16.6% of franchises would make the top 5 in any given year. Therefore, by being in the top 5 in the preseason, a team effectively doubles its chances of being a top-tier squad at the end of the regular season. Similarly, 30% of teams who were in the bottom 5 in the preseason ended up being in the bottom 5 in the regular season, a 14%-ish jump from regular odds.
If you’re a gambler (or you just like the thrill of having a chance), that’s a pretty big deal to you. Essentially, being in the top 5 means that your team has double the chance to follow that up into the regular season — even though a 30/70 split is still within the “unlikely” category.
What you may be thinking now is that preseason has far too little predictability toward the regular season, even if there is a chance that teams do follow up on their preseason successes (or failures). And, somehow, you’re wrong, but I had to do a bit of digging to prove it.
The Only Certainty
As it turns out, there is one thing that we can be certain of when it comes to preseason: Teams who finish in the top 5 never fall to the bottom 5. Literally, there’s been a 0% occurrence of this happening over the past decade. Allow me to indulge you in this highly detailed pie chart as proof:
This, actually, is pretty good information to know if you’re the type to gamble on teams to freefall in the regular season. Predicting a team that’s top 5 in the preseason to careen to the bottom 5 is exactly the type of bet that bookmakers love because they know you aren’t going to hit it. And, based on what we’ve seen over the past 10 years, they’re right.
Think about it this way. Over the past 10 years, just 14 passengers have died from plane crashes. The FAA claims that it handles 45,000 average daily flights, so let’s assume that there were a total of 146 million flights over the past decade. That means that out of all of those flights, just 0.0000096% — 9.6 millionths of a percent! Then, consider that getting struck by lightning during a decade is 0.000817%, 817 millionths of a percent.
If you were to both get struck by lightning, survive, and then die in a plane crash, you’d be the victim of a 78.4 billionths of a percent chance, which is 4,000 times less likely than winning the Powerball Lottery.
And, somehow, you’d still have had a greater chance of it occurring than a team going from the top 5 in the preseason to the bottom 5 in the regular season. Satisfied yet?
Some Other Tidbits
With this piece being somewhat discombobulated and more of a train of thought than anything, here’s a discombobulated list of some of the other correlations and notes on each:
The Three-Point Effect
Three-pointers made in the preseason had a near-0.75 correlation with three-pointers made in the regular season. This one is interesting, as it seems to show that even though coaches tend to experiment in the preseason with lineups, the game plan largely stays the same when it comes to beyond-the-arc shooting. If we see the Mavericks utilize Klay Thompson in hand-off situations to rain down threes, it’s likely they’ll continue to do so in the regular season. Three-pointers attempted had an even higher correlation of 0.85, driving the point home further.
Time To Assist + Turnover
One thing you’ll find in the correlation matrix (essentially a table of correlation values between different stats) is that preseason assists have a 0.57 correlation with their regular-season counterparts. This slightly surprised me, as the preseason often can be saturated with bench players who like to put up a lot of shots in the little time they have, similar to Summer League but less egregious. Yet, it generally holds that teams that dish out more in the preseason tend to do so in the regular season — like three-point shooting, this seems to be a coaching thing combined with certain players being assist-centric like Tyrese Haliburton.
Turnovers have a lower but still valuable correlation of 0.40, for what it’s worth. Most of this correlation comes from the youthful teams that tend to throw errant passes most often (like the Spurs last season). Nothing crazy here, but logical.
Rebounding Means Little
Curiously, rebounding means very little during the preseason when it comes to its predictiveness. Overall rebounding had barely more than a 0.30 correlation, which, if you remember, is the threshold for deeming something to be of value. While not completely useless, it’s interesting that when teams rack up boards they don’t tend to carry that over to the regular season. Could it be because more guys get garbage time rebounds than they would in regular season games? I’m unsure, but this is one of the pure counting stats that stands out.
Offense Matters, Though It Isn’t Everything
Across points (0.62), assists (0.57), and three-pointers made (0.75), it seems readily apparent that when teams are good in the preseason on the offensive side of the ball, you can generally rely on them to be good in the regular season — and vice versa. If your team struggles this year from the offensive side of the ball, I’d be willing to bet its going to be a rough season on that end. Offense and win totals, however, have a much lower correlation, so being bad on offense doesn’t mean that your squad is a lost cause.
More Info On Turnovers
While turnovers do tend to carry over a bit to the regular season, it’s important to note that it has no ill effect on whether a not can score points once the preseason is over. With an extremely small correlation of 0.006, turning over the ball a lot in the preseason doesn’t mean that the offense (see above) goes downhill during the year. So if your otherwise playoffs-caliber team is slipping up in the ball security department, you shouldn’t care much.
Fouls Don’t Carry Over
While I don’t have the foul data per game in the preseason, I do have data on free throws made/attempted, which are a good proxy. As it turns out, free throws made in the preseason don’t have a big effect on how the regular season free throws go (0.31) and free throws made don’t help either. Here’s the hidden part of all of this: if the NBA starts calling lots of fouls in the preseason to test new strategies for refereeing games, don’t expect it to become normal — at least not during that season.
That’s it for this statistically all-over-the-place piece! With preseason basketball going on, expect my own comments over the next few days/weeks. See you then!
I enjoyed the write-up - although I do think that the volatility of correlation practically suggests that there is no correlation. Generally, good teams tend to win basketball games - but the extenuating circumstances around preseason games suggests that they don't have any specifically real correlation with team success besides "this team has more talent than this team".