Why I Saw the Heat, Knicks, Thunder, and Clippers on Sunday

If you ever watch NBA games on TV (especially on ABC), you might have thought, like I do, that the same small set of teams play each other in different combinations each weekend. And it's only a matter of guessing whether the Lakers will be playing the Knicks or the Celtics, and if they aren't, whether the Heat will be. Being a Milwaukee Bucks fan, I'm especially sensitive to this kind of thing.

Broadcasters and the NBA can only show so many games, so there will invariably be some criteria applied to decide which games get nationally televised on a Sunday afternoon. A team's success or star power are two obvious criteria to consider, but perhaps most obvious is each city's television market size. Airing games with the L.A. or New York teams will attract audiences in the nation's two largest TV markets, in addition to whatever other audience those teams attract across the country.

The success of the Oklahoma City Thunder complicates that simple formula, though. The Thunder have one of the leagues biggest stars in Kevin Durant, made the NBA Finals last year, and have their home in the country's 44th largest TV market. A quick look at the broadcast schedule on ABC, ESPN, and TNT shows that the Thunder get as much airplay as any other team. The table below shows the teams with the most scheduled national TV games and their market ranks.

This data suggests that market size is a big factor, but a team's performance can earn it big TV time. Perhaps though, team success is correlated with market size so that bigger cities have better teams. If that were the case, broadcast decisions based mostly on performance would still result in the big-city teams getting more airtime.

A simple test of correlation shows that's not the case. The number of television households in a team's home TV market has no significant relationship with a team's wins in either the 2011 season (Pearson's r=.04, p=.83) or the 2012 season as of March 3rd (r=.18, p=.34). That is, being based in a big city does not generally mean winning more games.

So what are the relationships between team success, market size, and the number of nationally televised games?

I built a (multiple linear regression) model to test these relationships, and I included the total number of scheduled TV games for the 2012 season, the number of TV households in each team's home market, the total wins in the 2011 season, and two other variables as measures of team success. 

The first of these is an index of 2011 playoff performance. On this index, the league champion Heat get a 1, the Thunder get a 2, those eliminated in the conference championship round a 3, and so on down, including a rank for those who missed the playoffs. This might be a better measure of success than the win records from the shortened 2011 season. 

The other success variable is the number of wins during the 2012 season as of March 3rd. This is intended to represent the potential for success that each team showed when the TV games were scheduled before the season started. Of course, using this measure assumes some predictive ability of the TV schedulers, but some prediction of success based on trades and retirements is possible, and I think it makes sense to include such considerations in this analysis.

The results of this analysis show that the playoff performance rank index is the strongest predictor of TV games (b=-2.69 p<.01). A decrease of rank by one (e.g. from champion to runner-up or from first round loser to non-playoff team) means 2.69 fewer TV games. This relationship makes sense from a programming standpoint because the top-finishing teams got the most exposure during the 2011 playoffs and had the most momentum coming out of them. As the table above shows, the top-finishing Heat and Thunder are among the four NBA teams with 25 scheduled TV games.

The wins to-date for the 2012 season also have a significant relationship with the number of TV games (b=3.88, p=.018). The model shows that each additional win during the 2012 season relates to about 0.4 more TV games. This suggests that, in addition to the other performance measures, TV schedulers' predictions of 2012 performance have a strong relationship with teams' number of national TV games.

Finally, getting to audience size, the model initially showed significant relationships for both market size and 2011 wins with the total number of TV games. When I considered the interaction between audience size and 2011 wins though, the independent effect of each disappeared. That is, it is the combination of a team's TV market size and its wins last year that has a significant relationship with TV scheduling (B=.021, p=.056). In this sense, winning a lot of games or being from a big market is not as important as winning a lot of games and being from a big market. It is that combination that has a significant relationship with more nationally televised games.

So that's sort of a first hack at this. Some factors left out here are the specific networks and days on which games are aired. Another potentially important variable is game time, because the desire to air late games on cable might increase the air time for West coast teams. The model considered above is pretty strong though, and it is able to account for about 79% of variance in the number of TV games among all the NBA teams.