More than half of the teams in the NBA make the playoffs year. This makes a lot of the drama at the season’s conclusion about seeds. Having a 3 instead of a 4 or a 6 instead of a 7 could dramatically alter your team’s playoff chances…right?
The notion of a seed’s value is tricky. A lot of fans probably overrate the value of having a top spot: the commonly cited stats show that the top couple seeds almost always end up winning the Finals. However, this probably has more to do with the quality of those teams: if a team with the talent of the ’96 Chicago Bulls for whatever reason just sneaked into the playoffs, they would probably be only slightly less likely to win a championship. But the best teams usually have the best records and therefore the best seeds.
Therein lies the problem with using playoff results to determine seed value: they’re biased. We have to find the true value with a different method, starting from scratch.
There are two primary advantages to having a higher seed:
- The team is more likely to face an easier schedule.
- The team is more likely to have home-field advantage.
Solely for the purpose of making things difficult for me, the NBA essentially has two distinct seeding rules, one for each of these advantages. The first advantage is governed by the conventional seeding, in which division winners are guaranteed one of the top 4 seeds. However, the NBA bases home-field only on regular season record.
The other problem is that the first advantage is dependent upon the makeup of a particular conference in a particular year. This year’s Western Conference, for instance, has an unusually small difference between the best and worst seeded teams. Additionally, for any particular year, the difficulty of a schedule can be as much about the particular match-up problems in a series as it is about the overall quality of the teams.
With that in mind, I gathered historical data for the average team quality at each seeding over the last six years (post-realignment). In 2005 and 2006, the seeding rules were slightly different, so I reseeded the teams as if today’s rules were in effect (ESPN’s past standings already do this…someone should probably fix it). To measure team quality, I used Basketball Reference’ SRS statistic and converted it into winning percentage. Here are the average team qualities at each seed in the East, 1-8:
.714, .673, .601, .534, .541, .530, .500, .461
And in the West:
.682, .645, .650, .621, .631, .581, .577, .570
No surprises there; the West has been a much deeper conference over the last half of the decade.
Onto the methods: I’ve devised two possible ways to determine seed value.
- Create a completely neutral bracket: randomize the seeding and run a very large number of trials. Then examine how the chances of the original #1 seed change when it is given a random seeding. Find this value for all 8 teams.
- Replace one of the 8 with an average playoff team and see what its probability of making the Finals is. Repeat for all 8 teams.
I used the first method for this article, and will examine the second one later. I estimated home-court advantage at 60% for a single game. For simplicity’s sake, I (partially) solved the seed problem by making the 4 vs. 5 match-up have a neutral court. After a million random trials for each conference (using odds ratios to determine the probability of each matchup), here are the results:
The results are unsurprisingly different in each conference: while the top East team is the only one to actually benefit from seeding, the top three West teams all get a bump. The distribution in the East is counter-intuitive: you might expect the second-best team to benefit from ordered seeding, but in the trials the difference was negligible. How is this possible? One explanation: the #2 and #3 seeds are the only teams capable of knocking off the# 1 seed with any kind of regularity. In an ordered seeding, the #3 seed will never play the #1 seed while the #2 is still in contention. But in a random seeding, there is a chance that the (original) #1 and #3 will play each other before #2 plays either of them. This gives the #2 a fighting chance to make it to the Finals without having to face the juggernaut.
The value of one seed relative to another also merits consideration. Repeating the graph at the beginning:
The units are the percent change in probability. A value of 1.00 would mean that there was no difference between ordered and random seeding for that particular team. A greater value indicates a benefit from ordered seeding, and a value less indicates a disadvantage. From this graph, we can see that the Eastern Conference has two relative plateaus: 3-4-5 and 6-7-8, while the Western Conference is relatively flat after the sharp drop from from 3 to 4.
The problem with this particular method is that it isn’t necessarily indicative of brackets in which the talent distribution isn’t the same. The value of one seed over another is highly dynamic as the season progresses. However, this model might serve as a reasonable baseline for early and mid-season estimates. Over the rest of the playoffs, I’ll be looking at other effects of seeding that might shed more light onto more specific situations.