Freelance Friday is a project that lets us share our platform with the multitude of talented writers and basketball analysts who aren’t part of our regular staff of contributors. As part of that series we’re proud to present this guest post from Mika Honkasalo, a 21-year old NBA enthusiast from Helsinki, Finland who is studying computer science and mathematics. Mike enjoys big men who can pass and players who can shoot off of screens, find him on Twitter @mhonkasaloNBA.
The playoffs are different from the regular season. Not only are the stakes and level of competition higher but the the game actually changes in some fairly significant ways.
Pace tends to drop around 4 percent, more free throws are taken and teams commit fewer turnovers.
The structure of how you need to win also changes. Unlike in the regular season where the ratio of wins to losses matters, in the playoffs you’re only counting the number of wins you get, since it doesn’t matter how many you lose as long as you get to four in seven before the other team does.
The premise of this article is to find unique skills, tendencies or traits that playoff teams exhibit that allow them to overachieve their expected win totals in the playoffs, the search for playoff success. The charts below will show different statistical categories relative to league average and plot them against how they fared against estimated playoff wins.
I built a regression model with weighted values to predict expected playoff wins based on a few different factors including what your conference looks like and how good you actually are to get rid of as much noise in the data as possible.1
The sample size includes the last eight playoffs, which I concluded to be a big enough data set. Mainly because if certain patterns aren’t strongly hinted at in the last eight years of playoff and regular season data, they really aren’t tendencies that provide test cases for modern NBA teams. It’s not a huge sample meaning there’s some noise and findings can’t be taken at 100% certainty, but eight years felt like a good balance between useful test cases and variance. Here’s what I found:
Defense Beats Offense
To start I looked at sort of the low hanging fruit and cliches of playoff basketball, the most common of which is “Defense wins championships”. Surprisingly, not only does defense at the high end of the distribution really help, having a great offensive rating doesn’t help you at all, in fact it is harmful to your expected playoff wins.
Offensive rating actually has a negative correlation with expected wins, even at the high end of the distribution among the very best offensive teams. If you have a top-5 ranked offense, on average you will lose one win from your excepted total.
The difference between being a top-5 offense and having a defense comparable to league average compared to the opposite is a net minus of 3.2 wins. That’s almost one round in the playoffs, which is a massive difference. In fact, in the top quartile and in the bottom quartile of overachievers, there are 75 and 55 percent more top-7 defensive teams than offensive teams.
As a side note, pace doesn’t really impact the equation one way or another except at the extremes of being either really fast or really slow. 60 percent of playoff teams play at a below average pace, and the best place to be is generally in the middle of that area. There’s also some double counting involved here since many of the top offensive teams are fast paced teams. So you can’t play at a snail’s pace or just run up and down (each amount to a loss of -1.2 games) but otherwise you are fine.
The More Superstars the Better
The image above shows the impact of having multiple top 12 players based on Win Shares, which on average amounts to an additional 2.1 wins compared to just having one of those players.
Building a team around one star actually negatively impacts a teams performance vs. expected wins by a win. This doesn’t change even if you have only one of the top-3 players available in that year, bringing statistical validity to the notion of building around multiple stars, though the impact is significantly smaller than the offense vs. defense comparison.
The fact that multiple stars improve your odds of “beating the odds” shows something about teams being able to key in on a single player, which brings flashbacks to the 2011 Eastern Conference Finals where the Miami Heat did anything and everything defensively to get Derrick Rose out of rhythm.
If you don’t have any Top-12 players the result is close to a net plus-minus zero.
Here’s a really interesting one I was excited to learn about — the average age of the top contributors on playoff teams and how that impacts over or underachieving in the playoffs. This is almost the holy grail of conventional basketball wisdoms and the stats bear out some pretty fascinating stuff.
Of the top 16 of teams that outperformed their expectations, 15 were in the top 10 in age that year. There are mitigating factors that should skew these numbers such as bumps in minutes played in the postseason for older players, but 15 of 16 can’t be an accident.
Funnily enough, before reaching the very top end of the distribution, youth actually helps. I’m not sure if it has to do with fresher legs or if younger teams improve more in-season but the impact is beyond the scope of error enough to be statistically a valid observation that can be said with confidence.
Experience matters if you want to “defy the odds”.
Identity is Everything
The chart above shows the impact of assists on outperforming playoff expectations, but I could have chosen one of nearly every statistic that’s tracked and available on Basketball-Reference.com from defensive rebounding to steal rates and rim protection. What’s notable is that being exceptional in nearly every metric2 helps you outperform expectations.
It doesn’t matter which one, most are a net plus (though there are relative differences between them).
Some notable categories in relative order where team is in Top-5:
1. Assists +1.4 wins
2. Defensive Rating +0.9 wins
2. Effective field goal percentage +0.8 wins
3. Rim Protection (amount of attempts and percentage taken into account) +0.7 wins
3. 3-point percentage +0.5 wins
5. Defensive Rebounding +0.3 wins
6. Lower Turnover rate +0.2 wins
7. Offensive rating -0.9 wins
The most astounding observation is in the case of effective field goal percentage, which creates the inverse effect of offensive rating (a +1.7 wins compared to offense). This points to the fact that execution of offense matters more than how many points you are able to produce per 100 possessions. In a weird sense it’s not how many points you score, but how many buckets you make (adjusted to 3’s of course). Which is an amazing insight that I don’t really know what to make of. My guess it has something to do with the general makeup of teams that are able to execute good offensive sets to generate scores, especially in clutch situations, which helps them excel in a number of other areas. One could point to the Oklahoma City Thunder as a counter example of a team that sometimes struggles to run impressive sets at the end of games which hurts them.
Most of the content of this article has followed conventional basketball wisdom. Defense, experience and superstars are valuable. The biggest deviation from that comes in rebounding, which equals a net zero when you take into account both sides of the ball. Elite rebounding doesn’t seem to matter in the playoffs, or at least in a way which yields high dividends when compared to what is already expected from a team. Even in the extreme where take the top half of rebounders and compare it to the bottom half, there’s literally no difference.
Assists are actually the most important simple stat (Go Spurs!) when looking for teams that have a chance to outperform in the playoffs. The takeaway here is that whatever you do well, you should strive to do even better when the playoffs start. Identity is everything.
Though there are some things that are relative valued higher than others, which can be taken with a grain of salt due to sample size, being exceptional at one or two things and then building focusing on and getting them to translate in the playoffs is more important than anything.
In the top quartile of teams in expected wins, the ones that didn’t have any stat in the top 3 are expected to win 2.4 games fewer than teams with at least one– the value of over half a series.
*Link to SQL Schema in HTML here. All Stats per Basketball-Reference.com. Charts Created on Zizhujy.com
- Regression Model Standard Deviation 2.74 in Expected Playoff Wins. For reference being Top 5 in Simple Rating System amounts to +0.2 wins, meaning the advantage of just being very good is small and gives me confidence in the model. ↩
- Exceptions being offensive rating, offensive rebounding, shooting 3’s (attempts, not percentage) and pace ↩