Deep Dives: Matchups and Sub Patterns Continued

A quick follow on to the previous looks at starter versus bench lineups here and here. I was wondering about substitution patterns, and to try and see if any team consistently was able to find favorable situations where they had starters playing against opponent reserves. More on those results in a second, but first the general.

Watch enough NBA games, and you can easily pick out the pattern of play. Usually, teams stick with their first units for the first five or six minutes of the game, until the first timeout. After this, subs start to get mixed in, leading to a sort of dip in overall level of play spanning the last few minutes of the first quarter into the start of the second. Then by the under-nine media time out in the second quarter, the starters tend to filter back into the game. The same pattern is observable in the second halves of games, though to a lesser extent, as teams try to set up their closing lineups[1. NBA teams end the game with the same five which started a surprisingly tiny proportion of the time, 128 times total in 2014/15, though this of course reflects foul outs, late game defensive or free throw subs and so on to a degree.] for the stretch run.

Looking at play-by-play data from the 2014/15 season, here are the average starters in the contest by game minute, counting down from 48[2. I chose to ignore overtime periods for now as those unplanned for situations are much more about winning the game at hand than rotations for the majority of coaches and teams.]:

If the numbers for the 4th quarter look low, it’s because the above include all games. Looking at the 4th quarter of just games which ended with a final margin of less than 10[2. Not the perfect measure of close games, but it will do for a quick look.]:  

Even in close games, the number of starters on the floor is lower than expected. My intuition is that this reflects the degree to which teams tend to play slightly (or greatly) smaller closing lineups than they start with, starting from the top where both Golden State and Cleveland routinely finished games with their starting centers, Andrew Bogut and Timo Mozgov respectively, on the bench in favor of smaller, more versatile players. Additionally there is probably some “feel” going on here, with reserves who have had particularly strong games left on in place of a starter having a less than stellar game. All 30 teams largely followed the same pattern with some variation either way.

One instance which stood out was San Antonio. Among the better teams in the league, the San Antonio Spurs had a fairly striking point of differentiation, going to their bench about a minute earlier and a full reserve “more deeply” around the under six minute media timeout[1. San Antonio averaged 3.1 starters in the game between 6 and 5 minutes left in the first quarter, the only other team under 3.75 was Philadelphia and starter/bench distinctions don’t really apply to recent vintage Sixers teams quite like they do the rest of the NBA.], while bringing the starters back into the game in numbers significantly earlier in the second quarter. Perhaps relatedly, the Spurs were second in the NBA in second quarter point differential last season.

In terms of application, this only strengthens my general theory that teams have opportunities to better maximize the contributions of aging stars by moving them into bench roles. A Dwyane Wade-type coming into the game with about four minutes left in the first quarter seems like a scary thought for many bench-heavy units to deal with, all the while making it easier to conserve Wade’s minutes, especially in the first half, just as an example. Of course, there is a certain first-mover advantage here, as if this tactic becomes generally adopted, the advantage would dissipate.