Stat Exploration: Both Offense and Defense Affect Title Odds

One of the most famous cliches in basketball is that defense wins championships. That’s the type of statement that usually gets hacked to death by those wielding advanced-stats swords, but analysis has found time and again that the statement could be valid. With a Cleveland team dreaming of bringing a title home lacking defense but with a potentially killer offense, and a sleeping giant in the west in Dallas with what’s by far the best offense so far, this is as good a time as any to investigate this claim. If you’re an offense-first club, do you have a reasonable shot at a championship?

Calculating Title Odds

For an evaluation of how offense or defense leads to a championship more often, first we need to build some model of estimating title odds, which appears to be a Herculean task but can be done with a fairly simple model. I used the same model described here — binomial where the dependent variable is whether or not the team won a title — except that there’s no intercept and I account for the number of teams in the league.

Basically, I look at the team’s offensive rating and their defensive rating relative to the league average, and estimate odds for winning a title from those. Thus, we have the model results below for seasons 1955 (first year of the shot clock) to 2014:

1954-2014
Offensive rating: 0.495
Defensive rating: 0.682
(Log) number of teams: 1.72

The number being higher for the defensive rating means, yes, that defense is more important for winning a title assuming the other factors are held constant. It’s difficult to see just how much more important it is because this is nested in a logit function so I provided the graph below.

NBA title odds1

You can always view offensive and defensive ratings at basketball-reference for an idea of what the ratings actually mean — a rating of 1, for example, is just a little above average, while a defensive or offensive rating of 8 is historic. (An example of a team with a relative defensive rating of 8 and an offensive rating of 3, by the way, is the 2008 Celtics.) The separation between the lines is showing the advantage of defense given similar strengths elsewhere. This advantage grows to about 20% at the extreme end — it’s nothing to dismiss. However, a student of the NBA’s history will know one team can single-handedly distort the numbers: the Russell-era Celtics won 11 titles in 13 years thanks to consistently elite defense, even with an offense that often rated as poor. Thus, we may need to remove them from the system to see if Russell is the outlier throwing everything off.

1970-2014
Offensive rating: 0.530
Defensive rating: 0.590
(Log) number of teams: 1.67

The coefficients are much closer now, but defense still has an advantage. One may want to call off the dogs here and go home, but there’s an issue with the data available that translates into uncertainty. Namely, the coefficients have high standard errors, meaning they’re within the error range of one another and they are not significantly different. This is easy to conceptualize. Let’s say we have ten seasons of data where teams in the finals are evenly matched, and seven times out of ten the better defensive team wins. Does that mean that defense is more important? Not necessarily because we didn’t account for injuries or other factors. Using only regular season ratings, we have a high uncertainty. You can also see in the graph below that the separation is much smaller compared to the first graph, but a difference of 10% or even 5% at more modest levels isn’t something to ignore.

NBA title odds2

Weighing seasons for recency — giving more importance to the most recent seasons — yielded no new results, by the way. In fact, weighing for recency gave a slightly higher coefficient for defense for the 1970 to 2014 data.

But let’s phrase this differently. Let’s say we assume a team’s strength rating is our baseline here and we believe there’s no difference if that strength derives from offense or defense. We then test this by adding in defense as another variable to see if, all other things equal, a higher share of defense is more important. Based on the 1970 to 2014 data, defense is not statistically significant. Though, no surprise, using those prior seasons with the Celtics and defense is a significant predictor with a p-value of 0.009, but the power of this is reduced if you weigh by recency. Whether or not defense leads to titles will take more fine-grained analysis because besides the period of 1957 to 1969, there’s no significant shift in odds.

Exploring title teams

Looking at history, the average championship team has a (relative to league average) offensive rating of 2.3 and a defensive rating of 4.1. But again, the Russell-Celtics distort those numbers. Since 1970, the averages are 3.4 and 3.5, respectively. Championship teams have not been defense-heavy in a long time.

This entire issue, this axiom of what leads to championships, is particularly relevant for cases where two teams have strength ratings that are similar on paper but diverge based on offense and defense. Going back to 1970 looking for seasons where the best teams are within a net rating of 1 and one of those teams won the title, we have 12 different cases. In six of those, the better offensive team won, and in only three the better defensive team won. (There were three cases of mixed results where the title team was neither the best offensive team nor the best defensive team.)

To say that an offensive team has no chance against a great defensive team is unfounded. If you want to know if an offense-first team with electric scorers, lots of outside shooting, and one lacking size can ever win a title, they already did in 2013 with the Heat. Also, if you’re a bone-crushing defensive team, it’s no guarantee of a title either — just ask the 1994 Knicks, one of the greatest defensive teams ever.

In fact, it’s not difficult to find offense-first teams who were successful. One only needs to look at the Lakers, whether it be the Kobe-Pau team in 2009, a couple iterations of the Shaq-Kobe team, and the Magic-led Lakers of the 80’s. And the heralded 1977 Blazers, a pinnacle of basketball for most people, who were by all means an offensive club and beat a team with a similar net rating but with larger share from defense in the Denver Nuggets. Defense wins championships, but so does offense.

Historic title odds

Although the system isn’t perfect, it’s a quick way to tally title odds for every team. Since this article isn’t necessarily one for hard, quantitative results, we can play around with the data and have fun.

The least likely title teams? There’s actually a long list of teams with odds under 10% and even 3%. Using the model results for two eras (pre-1970 and post), the 1995 Rockets were the least likely championship team followed closely by the 1978 Washington Bullets and the Lakers, oddly enough, in 2001. The Lakers, of course, were the infamous title slump team that gave less effort in the regular season and ramped it up in the post-season. If there’s one thing to take away from the table below, it’s that the 1970’s were chaotic and surprises can still happen in the modern era. Also, since there were so few teams in the 60’s and 50’s, it’s difficult for a team to have odds lower than 10% and win — those Hawks were not a typical champion and lucked out from a Bill Russell injury.

NBA title odds3

For those who are curious, the 1996 Bulls had adjusted title odds of 50.7%, which seems low: Michael Jordan and what’s probably the greatest team ever would lose the title half the time? But looking at history, giants fall all the time, and it isn’t just because of injury. In fact, if we look at every team since 1970 with odds greater than 40%, we get a list of eight teams and a summed expected title share of 4.3. How many titles did those teams actually win? Four, so this is not an indictment on the model. There are no guarantees in the playoffs. We can repeat this with every team with odds greater than 20% and get an expected title sum of 19.9 and an actual title count of 21 — for a basic model, this tracks pretty well.

Looking at the most likely title teams, it’s a list dominated by the early 1970’s and the Bulls. The Bucks and Knicks were aided by a smaller league with fewer contenders, and remember the results are skewed by the strength of the opposition. If there are a couple other teams with a net rating over 7 or 8, they’ll take a big share of the odds too. Also, seasons before 1970 are ignored because every single season besides 1955, 1958, and 1968 would have been included.

NBA title odds4

Let’s take this further and build a unique leaderboard. Titles don’t always correlate to a team’s strength because you can peak at the time of Michael Jordan and come in second place, so I summed the title odds for every franchise since 1955 and listed them in the table below. The Celtics, yes, crush everyone, and that’s thanks to 7.6 title odds they gained in the Russell-era. The 76ers are surprisingly high, but they’ve been around for decades and built up a few impressive seasons in the 60’s and 50’s including a title as the Nationals in 1955. The Suns and Jazz probably like this table more than a list of championship winners — they’ve had strong contenders multiple times but haven’t been able to win. The Raptors come in last, but they’re an expansion team and look pretty good this season. What’s distressing is that the Clippers have been around since the 1970’s and barely rate above the new teams — and that’s with the gift of Chris Paul and Griffin in recent seasons.

NBA title odds5

Summary

Based on how you slice the data, defense is a little more important in winning a championship, but the results aren’t overwhelming and the 11 titles from Russell Celtics skew the model. To say that you “need” to be a top 5 team on defense is completely wrong based on past champions — for one, pure ranking is an inaccurate way to estimate odds, and secondly, what matters most is outscoring your opponent, which you can do on offense too. Let’s say this version of the Cavaliers with LeBron put out an offense 8 points better than average and a below average (-1) defense. Does that make it impossible to win a title? Of course not, and there are so few teams in NBA history who even qualify with those credentials. There are only eight teams with an offense rated over 7 and a below average defense, and while none of them won a title a couple were close: the 1988 Celtics suffered an injury to McHale for the playoffs and lost in the conference finals while the 1998 Jazz lost in 6 to the Bulls. Plus, the sum of the title odds for those eight teams is only 0.6 — they were hardly underachieving.

If LeBron and company want to bring a title to Cleveland, they won’t run into some invisible wall — having a below average defense doesn’t make it impossible to win a title. We haven’t seen it before simply because there are so few teams in NBA history with an offense great enough to overcome that handicap. If the Cavaliers are a serious contender, it will be because of how well they outscore opponents, regardless of where that strength is derived from: offense or defense.

Nevertheless, there could be small marginal gains in emphasizing defense over offense, but we need a better tool for this and one that not only looks at the playoffs game by game, but adjusts for injuries and other factors. But that’s for another time….

EDIT

A couple technical notes: I’ll provide the formula below for calculating the odds where you use relative offensive/defensive ratings (a positive defensive rating means you’re a better defense, not a worse one, so be careful there.) Also, I thought I’d address a concern some have had: the variance of offensive and defensive ratings affecting the results. I did not find that variance was significantly different for either — some have posited a lower variance on defense makes it “more” important. This was not the case.

Formula:
1/(1 + exp[ -(0.495*ORtg+0.682*DRtg-1.72*ln(teams) ) ] )

Quantcast