Bruin Sports Analytics
Offensive/Defensive Weight Part 2: Wrapping Up a Complicated Relationship
By: Ian Geertsen
(With an offensive/defensive weight of 14, the 2003-2004 Detroit Pistons are the most heavily defensively weighted championship winner since the 1983-1984 season—the image above shows defensive cornerstone Ben Wallace matched up against the late Kobe Bryant during the 2004 NBA finals)
Check out Part 1: Offensive/Defensive Weight Part 1: Correlations Between Orientation and Success
Welcome back! We’ve spent a while looking at the correlation between offensive/defensive weight and winning, so it seems like a good time for a change in perspective. Now, we will look at average offensive/defensive weight values of different groups in different scenarios.
Average Offensive/Defensive Weight
In the visualization above, we can see the average offensive/defensive weight among playoff teams from each decade. This data paints a familiar yet different picture from what we have seen in the previous data. Just as before the 80s are the most offensively weighted decade, as the average playoff team in this decade had an offensive/defensive weight of -0.333. The 90s remained negative but with a smaller value, although the 00s show a massive shift towards defense, as the average playoff team in the 00s had a offensive/defensive weight of 1.056. At its simplest form, this graph shows that playoff caliber teams in the 80s and 90s tended to be slightly offensively weighted, while playoff caliber teams in the 00s and 10s show the opposite.
This graph shows the average offensive/defensive weight, this time among championship teams. We can see that the eighties champions are heavily skewed towards offense, thanks in large part to the Showtime Lakers who averaged an offensive/defensive weight of -9.33 between their ‘85, ‘87, and ‘88 championship seasons. While the 90s-dominating Bulls averaged an offensive/defensive weight of -2.5 in their six championships, heavily defensively weighted seasons from the Rockets in ‘94 and Spurs in ‘99 help push the decade to a positive average offensive/defensive weight. The Lakers won three more championships with offensively weighted teams in the 00s, but only one other team that decade managed to win a title with an offensively weighted team—it was the ‘06 Heat, who won with an offensive/defensive weight of -1. With heavily defensive seasons from the championship Pistons, Spurs, and Celtics, the 00s also come out with a positive average offensive/defensive weight. Amazingly enough the average offensive/defensive weight for champions in the 10s comes out as exactly zero. This decade was characterized by dominance of well rounded teams, as 6 of the champion teams in the decade had an offensive/defensive weight between -1 and 3.
Now, let us take a similar approach but with a new set of criteria. Looking at all teams which won over fifty games since the 83-84 season, of which there are 262, we can see that each decade saw an offensively weighted average to varying degrees. As we would expect, the most negative offensive/defensive weight was found in the 80s, although unexpectedly the second most negative weighting was seen in the 00s. This, like many other pieces of evidence, shows that teams who reach a high level of success tend to be very slightly offensively weighted.
This time, let us repeat the exercise above except this time only looking at teams with over 60 wins. Of the 54 teams to earn more than 60 wins since the 83-84 season, the average offensive/defensive weight actually became more negative than when looking at teams with over 50 wins. Compared to the previous exercise, both the 80s and 90s saw average offensive/defensive weights that were even more negative, while the 00s and 10s saw a flip from negative to positive average offensive/defensive weights. These two effects mostly cancel out, leaving the average offensive/defensive weight among all decades similar to the one above.
Stepping away from wins, we can see a different side of the picture when we look at teams who had a regular season SRS of at least 7. Of the 51 teams to accomplish this feat the average offensive/defensive weight was -0.627, or about half of the magnitude that we found when looking at teams with over 50 and over 60 wins. This criteria is anomalous in that it is the only perspective we have used so far that does not have the 80s as the most offensively-weighted decade, as this time the 90s have an average offensive/defensive weight of -1.615, over four times the magnitude of the average value from the 80s. Also, the 00s return to being the most defensively weighted decade in this analysis, and the 10s are once again appear to be offensively weighted. Also, despite having a similar sample size to the data on teams with over 60 wins, this data is much more compact, with the largest magnitude of average offensive/defensive weight being just -1.615.
Average Offensive/Defensive Weight Analysis
Raising the criteria for success usually results in the average offensive/defensive weights becoming more extreme in magnitude, especially on the level of individual decades. It is unclear, however, how much of this effect can be attributed to the data itself and how much is simply from the reduction in sample size. This set of graphs, just like all of the data, reveals conflicting information and patterns; although not consistent, this data on average offensive/defensive weight also suggests that the aughts and tens had a relatively stronger correlation between defensive weightedness and success, while the opposite is true of the eighties and nineties.
It is interesting to note that, when looking at all decades, playoff teams and championship winners saw a positive average offensive/defensive weight, while teams with over 50 wins, over 60 wins, and an SRS over seven all averaged a negative offensive/defensive weight. The playoff teams category had the most open criteria and therefore had by far the largest sample, coming in at 592 teams. The next most open category in terms of selectivity was teams with over 50 wins; with 262 teams, this category had less than half that of the playoff teams category. Cutting away so many teams—teams which achieved 50 wins or less—led to a negative average offensive/defensive weight, which suggests that raising the criteria to include fewer, more successful teams would give you a group of more offensively weighted teams. Moving the criteria from over 50 wins to over 60 wins follows this same pattern, as the average magnitude of offensive/defensive weight was larger in the greater than 60 wins sample than the greater than 50 wins one, albeit by a very small margin. This pattern was not upheld, though, by the SRS greater than seven group of teams. The greater than 60 wins category was made up of 54 teams, while the SRS greater than seven category had 51 teams; because of their similarly high level criteria, their similar sample size, and the fact that many teams will overlap in these categories, one might expect these two categories to exhibit similar results. This, however, did not end up being the case. The average offensive/defensive weight among teams with an SRS over seven was -0.6275, a value which is just over half the magnitude of the greater than 60 wins category. This category is unique in that it has the 90s as the decade with the most negative average offensive/defensive weight—all other categories had the 80s—although it was similar to the consensus in that it had the 00s as the decade with the most positive average offensive/defensive weight—the playoff teams, championship winning teams, and teams with over 50 wins categories also found this to be true. Finally, we have the championship-winning teams category. Despite having the smallest sample size and quite a bit of variation between decades—three of this category’s four decades exhibited average offensive/defensive weights greater than one—this category actually ended up with the smallest average offensive/defensive weight between all decades. I find this result to be fascinating; the fact that so many different teams with different playing styles over different generations all balance out to give this group the most neutral average offensive/defensive weight is kind of poetic.
Given its sample size of just 51, the SRS over seven category exhibited surprisingly compact results. The SRS over seven group had just one decade with an average offensive/defensive weight magnitude over one while the wins over 50 category had two, wins over 60 had three, and the championship winning teams had three. Like the SRS over seven group, the playoff teams category also had just one decade with an average offensive/defensive weight magnitude over one. Similarly, when observing each of the five categories’ largest magnitude decades we can see that the SRS over seven category had the second smallest, behind just the playoff teams group. This checks out for the playoff teams category, as they had by far the largest sample and their data should therefore be less volatile. The SRS over seven category had the second smallest sample, though, so the fact that the data from this group seems to be relatively compact leads me to believe that the SRS over seven category may be a more accurate depiction of the relationship between success and offensive/defensive weight than comparisons involving just wins.
(The San Antonio Spurs put up more seasons with a regular season SRS over seven than any other team in the 2010s. The Spurs led the decade with four seasons, Golden State was second with three seasons, and OKC third with two seasons.)
In this section, we will be looking at the average wins of different groups of teams, starting out by simply looking at offensively and defensively weighted teams. As we can see, teams which are offensively weighted won on average more games than defensively weighted teams in every decade for which we have data, reflecting the previous trend that offensively weighted teams tend to find slightly more success than defensively weighted teams. Also following previous patterns, the difference between average wins was greatest in the 80s, when offensively weighted teams won an average of 2.724 more games than defensively weighted teams. The decade with the closest average, the 00s, saw offensively weighted teams win an average of just 0.608 more games. Over every year since the 83-84 season, offensively weighted teams have won an average of 1.498 more games than defensively weighted teams. All in all, this graph shows a very similar pattern to what we have seen in previous data; the eighties show the biggest advantage for offensively weighted teams with that advantage diminishing in the 90s, diminishing still in the 00s, and increasing slightly again in the 10s.
As we are now looking only at playoff teams, clearly the average win totals are higher for both groups. Just as we saw before, however, each of the four decades reveal an advantage for offensively weighted teams. Also, using this criteria once again saw the 80s have the largest difference in average wins, as well as the 00s as having the smallest difference. On the whole, out of all teams over all four decades offensively weighted playoff teams have won an average of 1.153 more games than defensively weighted playoff teams. But will this pattern continue when we look at championship winning teams?
It is interesting to note how the average number of wins for playoff teams was lower in the 80s than in the other decades by over three wins. This reflects the fact that, until the modern expansion era began in the 88-89 season, 16 of the 23 total NBA teams made the playoffs; this led to the pool of playoff teams being overall less competitive than in future years as a larger percentage of below average teams were included. This data suggests that including these teams led the decade to have a higher correlation between winning and offensive weight, although why exactly this effect is shown I am not sure. It is hard to determine exactly how much impact this has on the data, although the fact that the 80s saw the most offensively weighted average offensive/defensive weight when looking at teams with over 50 wins and over 60 wins suggests that the 80s would likely be a very offensively weighted decade even if less teams were let into the playoffs.
Just as we saw when looking at average offensive/defensive weights, setting the criteria to only include championship winners leads to the most neutral results, as the overall difference between offensively and defensively weighted championship teams was just 0.097 wins. In fact, both the 80s and 90s saw a difference of less than 0.3 wins between offensively and defensively weighted championship teams. This set of criteria also saw two decades where defensively weighted teams had more wins on average than offensively weighted teams.
Now, instead of simply looking at teams which are offensively or defensively rated, we will look at heavily offensively weighted and heavily defensively weighted teams. Kind of a mouthful, I know. For these purposes, heavily offensively weighted teams are teams with an offensive/defensive weight less than -10, while heavily defensively weighted teams are teams with an offensive/defensive weight greater than 10. While the previous exercises included many teams who were close to neutral in their calculus, this only looks at teams who have a somewhat extreme weighting in either the offensive or defensive direction, and allows us to compare these groups in the context of average wins.
As we can see, heavily offensively weighted teams have averaged more wins than heavily defensively rated teams, winning an average of 1.146 more games since the 83-84 season. As we’ve seen before, the difference between offensively and defensively weighted teams is largest in the 80s, although this set of data is unique in that this difference defies previous standards. Heavily offensively weighted teams in the 80s won an average of 4.55 more games than heavily defensively weighted teams during the decade, a fact which I find hard to believe. This set of data also reveals that heavily defensively weighted teams averaged more wins than their offensively weighted counterparts in the 00s; this is consistent with the results of the previous three graphs, as each had the 00s as the decade where the defensively weighted teams performed the best in comparison to the offensively weighted teams.
When looking at heavily offensively weighted playoff teams we can see that the 80s once again saw the largest advantage for offensively weighted teams, as heavily offensively weighted playoff teams won an average of 3.8 more wins than their defensive counterparts. This was the only set of average wins data, however, that did not have the 00s as the best decade for defensively weighted teams, as this data saw heavily defensively weighted playoff teams win an average of 1.533 more games than heavily offensively weighted playoff teams during the 90s.
In case you are wondering if there are enough teams present in each group for these findings to be legitimate, since the 83-84 season there have been 106 heavily offensively weighted playoff teams and 120 heavily defensively weighted playoff teams. Unfortunately the sample of heavily offensively and defensively weighted championship teams was not large enough to conduct an analysis within this criteria.
Average Wins Analysis
The data presented on average wins was heavily skewed in favor of offensively weighted teams. Not only did the offensively weighted teams average more wins in each of the five categories, but when looking at individual decades within those categories, the offensively weighted group averaged more wins in 16 out of 20 possible decades. This section may have provided the most definitive evidence we have seen yet, because this data makes it clear to see that offensively weighted teams simply average more wins than defensive ones. This fact rang especially true in the 80s and 90s, and the 80s actually saw the largest difference between offensive and defensive weighted teams in four of the five categories. While the results may not have been as extreme in later decades, this pattern was still present through the 00s and 10s as well. The 00s clearly came out as the decade where defensively weighted teams performed the best in comparison to offensively weighted ones, although the defensively weighted teams still only averaged more wins than offensively weighted teams in two of the five categories.
Similarly to what we found with the average offensive/defensive weight data, offensively weighted teams saw the largest advantage when more teams are included. As we raise the criteria for inclusion—and the number of teams in each group decreases—the advantage offensively weighted teams have over defensively weighted teams decreases, to the point at which this advantage is only slightly present when just observing championship teams.
Looking at all of our data, it seems apparent that, if there is a correlation between success and either offense or defense, the correlation is likely with offense. This is especially true when we include more teams, and therefore more lower quality teams, into our calculus; as we raise the bar for which teams are included, such as shifting from looking at teams with wins over 50 to wins over 60, the correlation between offensive/defensive weight and success tends to get less negative. As we can see from the chart below, the majority of statistical tests we have run point to offensive weightedness being more conducive of winning than defensive weightedness. In fact, 13 of the 16 analyses we ran found this result to be true, although many of these correlations and differences are very small. The three analyses where defense comes out on top, though, are among playoff or championship teams, suggesting that defensive weight may play a more prominent role at high levels of success. For instance, the average offensive/defensive weight among all championship winning teams was 0.2905, and the correlation between offensive/defensive weight and SRS was 0.0437.
It is also interesting to note that categories only looking at playoff teams gave the most negative correlation between offensive/defensive weight and success, while at the same time the average offensive/defensive weight among playoff teams is positive. This would suggest that, while the average playoff team since the 83-84 season has been slightly defensively weighted, there may be a very slight correlation between becoming more offensively weighted and becoming more successful among these teams. The fact that offensively weighted playoff teams have averaged 50.36 wins while defensively weighted playoff teams have averaged just 49.21 wins also backs up this finding. When looking at our data on average wins, we can see that offensively weighted teams have averaged more wins than defensively weighted teams in all five categories. While these differences are quite small—two out of the five categories saw win differences of less than one and no category saw a difference larger than 1.5 wins—this evidence shows that offensively weighted teams simply win more games, no matter how you slice it. It is worth noting that the closest difference between average wins was among championship winners, where offensively weighted champions won just 0.1 more games than defensively weighted champs. This once again suggests that, at the championship level, defensively weighted teams tend to look the best in comparison, and the difference between offensively and defensively weighted teams is usually the smallest.
Looking instead at how these statistics have changed over time, we can also see interesting patterns from the historical context. Throughout this analysis we have consistently seen the 80s, and to a lesser extent the 90s, as having the strongest relationship between offensive weight and success. As you can see in the chart below, the decade that saw the largest advantage of offensively weighted teams over defensively weighted teams was the 80s in 12 of the 16 analyses we ran. On the flip side, the decade that gave the largest advantage to defensively weighted teams over offensively weighted teams was the 00s in 10 of the 16 analyses. As we have seen, the general trend appears to be that the relationship between offensive weight and success was strongest in the 80s and weakened in the 90s; the relationship between defensive weight and success was at its strongest in the 2000s, and by the 10s the relationship shifted back again.
(Spearheaded by superstars Kareem Abdul-Jabbar and Magic Johnson, the 1980s “Showtime” Lakers reached championship heights through a very offense-centric style of play)
We can’t discuss this trend, though, without bringing up the externalities that undoubtedly helped to shape it. For one, the NBA saw rapid expansion during the 70s—from 1970 to 1980 the NBA increased from 14 to 22 teams—and had a subsequent period of expansion in the late 80s, going from 23 to 27 teams over the span of 1987 to 1989. Basketball played during the expansion years was likely of lower quality overall as each team became slightly worse with each expansion draft—there were three expansion drafts in the 80s, resulting in the creation of five new teams. Compound that with the addition of teams that immediately slot into the bottom of the wins category and you simply get a lower quality league. The expansion of the NBA, in addition to the illegal defense rules and other rules changes made by the league, likely contributed to the extremely offensively-biased results seen in the 80s compared to other decades. Likewise, the removal of illegal defense rules in 2001 was likely a contributing factor to the defensive weighted results seen in the 00s, and the hand-checking rules put in place in 2005 may have factored into the 10s being less defensively weighted than the 00s. It’s impossible to say exactly what impact these rule changes may have had, but any basketball fan knows how much these rules have shaped the game, leading them to undoubtedly affect data like this.
In the following chart (last one, I promise!) we instead look at which decades were closest to and farthest from the average value across all decades. While the previous chart gives us valuable information on absolute values in the data, this data tells us information about the decades in relation to the overall distribution of the statistic. From this chart, we can see that the 80s still appear to be the most extreme decade, being the farthest from average in 13 of 16 analyses. This time around, though, it is the 10s that appear to be the most centered decade, as the 10s value was closest to average in 10 of our 16 analyses. This shows that, while the values from the 00s tended to be closer to zero and were the most neutral between offense and defense, the 10s were the closest to the center of the distribution.
Given that the 10s decade was closest to the average of all decades, and considering that the 10s appear to be more offensively weighted than the 00s, it appears that the current trend we are on would lead teams in the future to become more offensively weighted. The style of play and pace of the league may also predict this change; as we saw earlier, faster pace of play may be loosely correlated with more success to offensively-weighted teams, and the NBA’s pace of play has been increasing rapidly in recent seasons. The average pace of play among the last 5 NBA seasons was 97.60 while the average pace for the five seasons before that was just 92.64. Given this dramatic increase, as well as the pattern the league has followed in recent history, I would predict that the league will continue to see very slight increases in the correlation between an offensive weight and success. But, as we’ve seen many times today, things are rarely so simple.
Sources: Basketball Reference, NBA Rules History, Wikipedia,
GitHub Repository: https://github.com/Iangeertsen/Offensive-Defensive-Weight.git
Google Sheet: https://docs.google.com/spreadsheets/d/1ZXeTIywze6KVGtpvxbcWkD1qOyBvJ e29VDvUFf7ud58/edit?usp=sharing