What does small ball look like?

The Western Conference playoffs have become defined by lineup matchups. In knocking off the San Antonio Spurs last round, the Oklahoma City Thunder found success pushing towards the bigger end of the lineup spectrum, often playing Enes Kanter and Steven Adams together. Kanter’s defensive shortcomings were less of a problem against San Antonio’s bigs, particularly David West and Tim Duncan, neither of whom are particularly mobile in space or have three-point range.

In the Western Conference Finals, Billy Donovan has zagged somewhat, still getting great production from his bigger lineups but also using some small ball groups to great effect. The Golden State Warriors’ “Death Lineup,” a small ball group featuring Stephen Curry, Klay Thompson, Harrison Barnes, Andre Iguodala, and Draymond Green has been completely neutralized by the Thunder’s smaller units.

A few weeks ago, I looked at the minute-weighted height of each team by position, focusing in on the Warriors and Thunder as this series was getting started. Today, I wanted to slice things slightly differently and look at relative size by specific lineup.

The table below plots every lineup that played at least 100 regular season minutes by their average height and their average position (using individual position estimates from Nylon Calculus) with each position represented by the standard numeral (1-point guard, 2-shooting guard, etc.). The size of each mark reflects the total number of minutes that lineup played.

The average position of every single lineup sits in a range between 2.5 and 3.5. A mark of exactly 3.0 would represent a traditional position distribution, one player from each of the traditional position categories. The fact that every lineup falls within such a narrow range means that small ball, or big ball, is almost always the result of swapping out one player for another who usually plays one position up or down the scale. Essentially, we’re looking at lineups with a small forward at power forward, or two centers. We are rarely seeing lineups with three point guards, or three centers.

However, this is a slightly skewed perspective. The position estimates at Nylon Calculus are meant reflect the positions where each player is actually receiving minutes. So if we assume that Draymond Green’s natural position is power forward, his actual position estimate comes in slightly higher (it’s 4.2) to account for those minutes that he plays center in the “Death Lineup.” That means that the spread of average positions for these lineups is being slightly condensed and may actually be slightly wider if we just thought of players as representing the skills of a single, natural position.

Because of that noise in the system, I’m going to just use the average height of a lineup for the rest of our analysis, as that seems to be a better proxy for what a team is adding or gaining in terms of skills by the design of their lineups. This next view looks at the point differential per 100 possessions for each lineup, by average height.

 
This data is also slightly warped by circumstance. Lineups with a small number of minutes played are more subject to random variance and less likely to have point differentials that accurately reflect their abilities. However, lineups with more minutes played are likely to be better — if a lineup isn’t good it generally won’t play. Still we can see some trends begin to emerge.

There is a slight trend towards smaller lineups being more effective, however this is skewed heavily by the Warriors and doesn’t account for how many minutes each lineups has played. Personnel dictates that most teams fall primarily on side of the spectrum — bigger or smaller than average. There are a few teams, like the Indiana Pacers, who have a very good big lineup and a very good small lineup. We can also see teams like the Cleveland Cavaliers and the Golden State Warriors who appear to be much more effective the smaller their lineups are.

This next view takes the same plots and separates them into offensive and defensive efficiency.

 
The small lineup the Thunder have had so much success with against the Warriors — Russell Westbrook, Dion Waiters, Andre Roberson, Kevin Durant, Serge Ibaka — doesn’t appear in our sample here because they played just 46 minutes together in the regular season. However, it’s worth noting that even calling them a small ball lineup may be a stretch. While the average position of that group works out to 2.6 — definitively small ball and what would be the smallest lineup by position in our sample — their are relatively big position-by-position. By listed heights on Basketball-Reference, that group has an average height of 78.6 inches which is only slightly below average and about a half inch taller than Golden State’s “Death Lineup.”

It’s also fair to assume that even that is undershooting things as Durant is only listed at 6-9, Roberson at 6-7, and focusing on height doesn’t include the more functional effects of wingspan, which would be an even more dramatic advantage for this group. Of course, one has to be cautious of this:

The full season data is fun to play around with but even a 100 minute sample, which limits us to just a fraction of the five-man units that actually played together, may still be too small for meaningful analysis. And stretching that analysis to what’s happened in just four games in the Western Conference Finals puts us on even shakier ground. Still, whatever the reason and whatever variables are at play, the Warriors are not getting the results they usually get from their secret weapon. That’s a big reason the Thunder are in a position to eliminate them tonight.