Defense wins championships. So does offense, but that’s not the cliche. It’s defense that puts great teams over the top and turns the elite into the immortal. And the numbers bear that out — in the last 10 years, every team that made the NBA Finals had an above average defense.
Not all championship defenses are created equal, however. Pop quiz! Which defense was better: the 2011 Dallas Mavericks, or the 2013 Miami Heat?
A quick perusal of your favorite statistical website provides an easy answer. It’s the Heat, of course! In 2013, Miami allowed 103.7 points per 100 possessions, via Basketball-Reference; the 2011 Mavericks, on the other hand, had a defensive rating of 105.0. 103.7 is less than 105 (MATH!). The Heat win.
It’s not quite that simple, however. Each season is an entity unto itself, and a year-to-year comparison can be a difficult prospect as a result. The further separated in time the respective season, the more odious the task. There’s a litany of reasons for such compartmentalization, but the commonality is change. Teams change across the board, from rosters to ownership and everything in between. And the league changes, too. Different rules come into vogue as others slip into history. Different points of emphases take their place in any given year. Different strategies and tactics take route, as defenses become more sophisticated and offenses adjust their shot selection and means to such ends. Sum the variables and every season represents a different scoring climate, where no two numbers are similar — no matter that they might be the same.
Back to those championship teams. In 2013, the average NBA team scored and allowed 105.8 points per 100 possessions. In 2011, that number was significantly higher (107.3). So was Miami really the better defensive team? Their raw defensive rating was lower, sure, but it came in a league that was less conducive to the ball going through the rim. Fortunately, we can use the average rating* in each year to figure out just how much better each team was relative to the environment in which they played.
*Since everything sums to zero with 30 teams, the “average” team has the same offensive and defensive ratings. Any references to “average rating” is this league average for both offense and defense in the specified year.
Simple subtraction would get you one number, but a more accurate representation of the distance from the middle of the pack would be a percentage. And that’s just as easily calculated; we simply take the raw individual team defensive rating, divide it by the league average for that specific year, and multiply the result by 100. The product is a defensive efficiency number that’s adjusted to the league average for the season in which the team played. It measures what percentage of the league average scoring a team allowed; the lower the number, the better.
Take that 2013 Miami Heat team. They allowed teams to score 97.92% of the league average points per 100 possessions. The 2011 Mavericks, conversely, gave up 97.86% of the league average. The raw numbers, then, which indicated a clear advantage for Miami, were misleading. When we account for the context of each season the 2011 Dallas Mavericks had a slightly better season defensively than the 2013 Miami Heat did.
The same can be done for offense, too, of course, except in that case, the higher the number, the better. In a nutshell, this is what I’m calling ORtg+ and DRtg+, shorthand for “league-average normalized offensive/defensive rating.” It’s a concept that’s long been around in baseball, and it’s existed in spreadsheets and APBRMetrics forum threads for ages. But it’s generally not recognized on a larger scale, and I think that’s largely because there is no searchable database for these numbers.
That problem is fixed. Basketball-Reference has offensive and defensive rating numbers for each team going back to 1973-74, and they have the league average for each season as well. I pulled all of that data to come up with ORtg+ and DRtg+ numbers for each team in each season going back that far. The spreadsheet is available here; you can sort it by season, individual year/team combinations (i.e. the 2011-12 Charlotte Bobcats), franchise histories (sorry, Sonics fans, but the Sonics are labeled “Thunder” here), raw offensive and defensive ratings, adjusted offensive and defensive ratings, adjusted net rating (a simple subtraction of DRrtg+ from ORtg+), teams that made the Finals, and parameters for the last 20 or last 10 completed NBA seasons.
Which is fun! The data in a spreadsheet is revealing in a lot of ways. But it doesn’t quite do the scale of this project justice. Luckily for me, all-around blogging superstar Ian Levy is a wunderkind with the visualization software Tableau, and he put together this fun little chart:
Those dots represent the 1,073 individual team/season combinations in the dataset. The y-axis is DRtg+; better defenses are to the bottom, worse defenses to the top. The x-axis is ORtg+; better offenses to the right, worse offenses to the left. Quadrant IV, in the lower right, is where all the best teams hang out — good defense, good offense. And squares represent teams that made the NBA Finals.
The visualization is also sortable by Franchise, ranges of ORtg+ and DRtg+, individual team/season combinations (tip: hover over the title bar in that checklist that says “Specific Team and Season” to reveal a toolbar with a search option to more readily find individual combinations — the format is “City Team – Year”), and specific seasons. And they can be combined for different multivariable parameters, too!*
*Note: because Ian is a busy man, the visualization is only updated through the All-Star Break. That only affects 2013-14 teams, and marginally at that. The spreadsheet data, on the other hand, is accurate through games completed on 2/25/14.
I’ll be diving deeper into the data in a series of posts over the coming days and weeks, but first, some quick-hitting notes on what most stood out from this project:
- Truly elite teams on either offense or defense are roughly 5% better than league average (meaning an ORtg+ of 105 or greater, or a DRtg+ of 95 or lower). Related — the standard deviation for NetRtg+ is 4.43 percentage points.
- The very worst offenses in modern league history are worse, percentage-wise, than the very worst defenses. The worst defense in this sample was just over 8% worse than the league average that season; the worst offense, by comparison, was 11% poorer than league average scoring in its respective season.
- The Jordan-era Bulls are ridiculous, outliers among outliers. But there are a couple other teams of more recent vintage whose impressive regular seasons were overshadowed by a failure to win the title, including certain Celtics, Thunder and Cavaliers squads. Check the data for specifics; I’m not giving away any spoilers here.
- Certain franchises have historical narratives that fit the data, which makes for some interesting individual franchise graphs. The Suns are one, with a propensity toward offense:
And San Antonio’s chart is just impressive for its long-standing success:
There are certainly some caveats with this data. There’s some variance between Basketball-Reference’s efficiency calculations and those of other websites, so certain teams separated by just decimal points might flip-flop rankings depending on the source data. It’s not “true” adjusted efficiency, either, which would take into account each team’s opponents in a given season as well. Basketball-Reference’s Simple Rating System does a better job of capturing that adjustment in the aggregate, but it isn’t parsed into individual offensive and defensive components. And that was the impetus here — a way to historically compare teams across different eras and even different years, and to further the discussion of normalizing statistics to make them more useful. Advanced statistics are all well and good, but what’s the point if we use them in rudimentary ways?
Again, a major tip of the hat to Ian, without whom this project wouldn’t have been nearly as fun to produce. And much credit and thanks to Basketball-Reference.com; without their invaluable database, this idea would have been dead in the water.