Tuesday, October 15, 2013

National TV Rondo

The enigmatic Rondo is one of the truly unique players in NBA history. Attached to his lore is the idea that he steps up in big games and especially on nationally-shown ones. This mythic form of Rondo, mentioned here by Bill Simmons and here by a Celtics site, is one who takes over games and uses his talents to their full potential. Allegedly, he's a triple-double machine, and instead of passively probing the defense and settling for passes he attacks the basket more.

Defining the terms

The four major networks that carry NBA games nationally are TNT, ESPN, NBATV, and ABC. Which network contributes to this mythic form of Rondo is up for debate, but it's easy to split the numbers by channel and look for patterns. If Rondo performs better with a brighter spotlight, then we should see him do better in ESPN games than NBATV games, and better in ABC games than ESPN ones.

Harvardsports already did an article on the subject, but it was limited work with a few problems. First of all, it didn't control for minutes played, much less possessions. It's not unreasonable to assume Rondo will play more in nationally televised games. The data also spanned two seasons (2012 and 2013), but in combining two seasons you assume national games were evenly distributed between the two seasons, and that's not often the case with an aging team becoming less marketable.

Also, how do we judge a player getting better or worse in televised games? You can go by points per game, or per possession, sure, but that's missing playmaking, efficiency, defense, and other factors. And the all-encompassing stats aren't perfect either. This is the age of advanced stats, but there's no agreement on which player metric to use. My solution is to break down the games by a multitude of stats to look for patterns, and then to evaluate the key question using a simplified version of WARP. I ignore defense, which is hard to capture in box score stats, and I ignore offensive rebounding -- when people talk about National TV Rondo it's usually about his offense anyway, and offensive rebounds aren't particularly relevant for point guards.

As a final note, I'm looking at both regular season and playoff games but only from the 2008 season and on, so Rondo's rookie season is ignored as I don't find it especially pertinent.

The numbers

Using a variety of sources, Rondo has been in 144 nationally televised games. 34 of those were on NBATV, but a healthy 22 of those were on ABC. The rest were on ESPN and TNT.

Adjusting for pace, the numbers are posted below. OffRat is an efficiency metric based on WARP whose inputs are shooting efficiency, usage (you get credit for carrying a bigger load), assists, unassisted shots (more credit for unassisted shots), and the league average efficiency. His numbers are mostly unchanged across the board, with the exception of ABC games. Rondo scores at a significantly higher level on better efficiency. Although he turns the ball over more, he's assisting more -- and remember, this is adjusted for pace (and by extension minutes played.) An offensive rating that's 1.6 points better is a clear separation. The chart below, by the way, didn't include playoff numbers for a different perspective.


No national TV
National TV
ESPN/TNT/ABC
ABC only
PTS/100 Poss
13.7
14.0
14.3
16.9
AST/100 Poss
10.1
10.4
10.7
12.2
STL/100 Poss
2.21
2.13
2.07
1.77
TOV/100 Poss
3.27
3.23
3.30
3.43
DReb%
7.0
7.3
7.4
8.2
OReb%
4.5
4.5
4.7
5.7
TS%
50.7
52.2
51.1
52.3
OffRat
94.6
95.1
94.9
96.2
*Seasons: 2008 to 2013, playoffs included



The difference between Rondo's offensive rating in nationally televised games versus non-televised is fairly small. In fact, using a t-test there's no statistical significance, even with the ABC-only games, partly due to the smaller sample size. However, the Celtics are more likely to face a better opponent during these games, which makes Rondo's exploits more impressive. Indeed, the competition is certainly more steep in the nationally televised games -- for ABC, the average opponent is nearly at a 5 SRS, which is contender status. Roughly speaking, there are only about four such opponents every season, and the Celtics had that strength for two of them. Using an adjustment based on the "simple" defensive rating of the opponent, Rondo's numbers are magnified. His offensive rating increases by nearly two points, and for ABC games it nearly increases by four. (Remember, the scale is points produced per 100 possessions, ignoring offensive rebounds, so it's a notable leap in efficiency.)


No national TV
National TV
ESPN/TNT/ABC
ABC only
OppDef
-0.05
-1.36
-1.73
-2.31
OppSRS
-0.33
2.15
2.94
4.76
Adj. OffRat
94.8
96.6
96.7
98.4
To determine whether or not the difference is significant, a t-test is used. But again, a few tweaks are introduced. First of all, playoff games are dumped from the analysis since playoff games are more likely to be televised and the point here isn't testing playoff Rondo. Also, the Celtics were televised at different rates at various stages of Rondo's career. Seasons are grouped together based on the proportion of televised games to non-televised. If a player wasn't televised earlier in his career, for example, his stats will look better in televised games because he was a more developed player by then.

Digging into the stats, National TV Rondo doesn't appear to be a myth: his numbers are definitely better, even breaking down into multiple seasons segments and slicing up the types of televised games. From 2009 to 2010, his numbers on national TV were better but the p-value was just a smidge above 0.05, which is the normal threshold for rejection. (P-value is basically like the probability of the null hypothesis being true, where the null is there is no difference of Rondo in national games.) The division of the seasons into only two year bins pushes the p-value higher, and if you combine seasons '09 through '12 you definitely get statistical significance (the proportion of national TV games to non-national was roughly 0.65 for 09/10 and 0.75 for 11/12, so this is still an okay comparison.)


Seasons
Non-televised average
Televised average
df
p-value
All*
2009-10
95.5
97.1
137.0
0.0581
All*
2011-12
91.5
96.6
116.4
3.78e-7
ESPN/TNT/ABC
2009-10
95.5
97.1
109.2
0.0604
ESPN/TNT/ABC
2011-12
91.5
96.8
86.6
1.95e-6
ABC
2009-12
93.9
98.5
27.8
2.67e-4
*Includes NBATV, ESPN, TNT, and ABC

Conclusion

Looking at a few seasons of data, Rondo does appear to play significantly better in nationally televised games, especially when you adjust for the higher level of defense from the opponents and when you look at games on ABC, which as a network carries more weight than a game on TNT. Adjusted for pace, he scores more points on a better efficiency, and even assists more when ABC is televising. Does Rondo just play better versus a higher level of competition? Well, that was tested too as a confounding variable, but Rondo's stats did slightly get worse versus better opponents/defenses, so the fact that his numbers improve in televised games, where there's a higher level of competition, is impressive. In any case, his numbers in general fare better versus better defenses than average, and this effect is increased when the game is nationally televised. And I'm even ignoring rebounding, where it appears he holds up even against better teams.

Are there any other explanations? Even with Rondo doing well against good defenses it's still within the spirit of the original question: Rondo rises to the occasion. Top defenses could have different strategies; they may dare Rondo to shoot and leave him open, choking off the better options. But then why do his assists not plummet and in the case of his ABC games they actually increase? He's a perplexing, unique player, and in an ode to quantum physics he changes when he's being observed.

1 comment:

  1. Nice post! Can’t wait for the next one. Keep stuff like this coming.
    Asus Laptops

    ReplyDelete