Pages

Sunday, June 24, 2012

Stop tanking, it doesn't work


Not to brag too much, but I wrote a post recently that looks at whether the NBA’s winners benefited from tanking. It also examined the perennial losers and showed some anecdotal evidence that they’ve done plenty of tanking, intentional or otherwise, to little effect.

I would not really be doing my job (not the real one, the blog one) if I didn’t at least look at some numbers given that the blog is called “Sports + Numbers,” so let’s look at some numbers. AThis post is about evaluating sequences of results. For example, we may look at where teams finish in year 0 and then track subsequent performance. Performance here will be measured mainly in terms of winning percentage and sometimes in terms of playoff or lottery position (e.g., 13th ranked winning percentage of the playoff teams will be considered 13th in the playoffs, while second worst winning percentage of the lottery teams will be considered 2nd in the lottery).

My main goal is to answer the following question: If a team finishes lower in the league than they otherwise would, does it have a demonstrable effect on future performance?

Housekeeping

I will be looking primarily at the period beginning with the 1984-85 season because that is the one in which the NBA introduced the Draft Lottery to make tanking less attractive. At various points I may also take a look at the period beginning with the 1997-98 season, because that year’s draft class was the first to be subject to the individual rookie wage scale that, as I have noted, is important.

The Big Picture

The first picture we will look at is team performance in future years based on league rank in year 0. For this analysis, year 0 will be any season from 1984-85 through 2001-02. This will allow us to look at ten years going forward from any given season so each year will have the same 482 teams in the data set.

Sample: 1985 to 2002
The picture is a bit complicated, but I love heat charts so I couldn’t resist posting it. One thing to notice is that the deeper reds and blues – noting more significant lows or highs, respectively, in winning percentage – dissipate rapidly in a wedge starting with the middle teams and widening in each subsequent year. This is what people are saying in general when they describe mean reversion.

A second observation here is that there is fairly strong persistence of outperformance in the NBA. This makes intuitive sense in a league dominated by superstars and dynasties, where top players changing teams is rare enough to warrant an hour-long special on ESPN[i].[ii]

A final, more anecdotal, observation from the graph is that there are odd clusters. The 9th through 14th ranked teams have relatively poor performance in years four through eight. Alternatively, the 15th through 19th ranked teams seem to do abnormally well during those periods.[iii]

What this suggests to me is that we need to ditch league rank and look at relative playoff/lottery rank. This will put a firm difference on the last playoff team in and the first out, as the first out should get the benefit of being in the lottery while the last in should get only the benefit of a swift beating by the top seed in their conference (unless the top seed is the Dallas Mavericks in 2007, Seattle Supersonics in 1994 or the Chicago Bulls in 2012). Using playoff/lottery rank will have the added benefit of normalizing worst in the league, which was 23rd in 1985 but 29th in 2002 due to expansion. In the playoff/lottery rank, the team with the worst winning percentage will always be first in the lottery and the best will always be first in the playoffs.

Sample: 1985 to 2002
Now we’re starting to get there.

Does Tanking Pay?

There are two distinct kinds of tanking at issue here. Dropping from the worst playoff team to the best lottery team or dropping from a bad lottery team down to the worst lottery team. Any conclusions derived from this data come with the massive caveat that two teams finishing in the same position can be very different. An 8th seed in the Eastern Conference could, hypothetically, be an aging team on the way down while the 8th seed in the West is an up-and-coming young team that just clawed their way into the playoffs for the first time. While those should average out in a large enough data set, this is not that data set. The post-1998 sample in particular is rather smallish for this exercise but, because it is interesting, I will keep using it.

Sample: 1985 to 2002

Teams that finish as one of the lowest-ranked playoff teams appear to perform better than those that finish as the highest-ranked lottery teams through the next six or seven seasons based on the 1985 through 2002 data. In the post-1998 data set, however, the high lottery teams outperform the low playoff teams in the medium run, recovering from an initially-depressed winning percentage to exceed the low playoff teams in years 3 through 5. The strategic implication here would be to tank, if possible, when faced with a seventh or eighth seed in the playoffs, but the differences we are working with are relatively slight, three years away and making the playoffs is worth a couple more games of revenue even in a sweep.

Sample: 1998 to 2007
What about for reaaaaaaally bad teams?

This data set can also inform our strategic decision on whether to tank to go from a 5th to 8th position in the lottery to a 1st to 4th, increasing the chance of winning the top pick.

According to the data set starting in 1985, the worst teams in the league – those in the 1-4 range of the lottery – revert to .500 at roughly the same speed as the next worst cohort. This holds up in the post-1998 data set as well. Since the combined cut doesn’t show us anything, we need to take a look at a finer cut of the data to see whether being the worst team and having the best shot at the lottery makes any difference.

Beginning year 1985 to 1997, worst 8 teams only

Beginning year 1998 to 2007, worst 8 teams only

These two charts look at the performance of the 8 worst teams in the league in both the full lottery era and the post-1998 lockout era. What seems clear to me is that there was a significant underperformance of the worst team in the 1985-1997 era. The absolute worst team stays below the second worst for the full five years, but jumps above them in years four and five of the post-lockout set. The team with the seventh worst record outperforms significantly, but I would hardly go to an eighth place team and tell them to fall only one spot.

In lacking a clear finding for the most current sample, the question of tanking answers itself. Without giving a clear improvement in subsequent performance, is it worth it for teams to spend up to half the season putting backups on the floor and pissing off season ticket holders?

Going back to my original question, I have a very hard time answering that teams do themselves any favors competitively by finishing worse than they otherwise would. When you figure in the short- to medium-term damage done to their fans I would argue that tanking is a poor strategy for building a team.



[i] Wait, most players don’t use an hour-long special on ESPN to explain changing teams? Oh well, it’s still relatively rare.
[ii] I’m still upset about the whole LeBron thing.
[iii] Seriously, can we go back in time and take away his Finals MVP? If it doesn’t go to the Refs I think Shane Battier has a better case.

No comments:

Post a Comment