martes, 15 de abril de 2014

It does exist an objective alternative to the NBA coaches’ ranking provided by the ESPN.

Coauthored with Fernando Herencia

The ESPN web has recently published an article where it was provided a ranking for the NBA coaches of this season. In doing so, they asked a panel of experts to “rate each team's coach in terms of their performance in guiding the franchise to overall on-court success, both in the short and long term”. In the explanation of the methodology by the ESPN it is said that “there's still virtually no objective measure to rate the guys who draw up the plays and manage the egos apart from titles and rings”. It is true that there is not an objective measure to evaluate the performance in the long term since they are based on expectations, however it is not true that there is no objective measures to analyze the coaches’ performance. The most popular is that based in frontier methods, in which for each coach is attributed in some way the quality of his rooster and subsequently it is evaluated the difference between the actual results and the best results possible for his squad. The lower the difference between them the better the coach performance is. This methodology was used to evaluate NBA coaches´ performance in this paper from Dave Berri.

Other alternative to evaluate objectively the coaches’ performance is to compare the actual results with those expected from betting odds. The probability of the two possible outcomes in a game (i.e., local victory and away victory) is embedded in betting odds so that it is enabled to calculate the probability of getting each of the possible number of victories.


Next graph shows the density function of victories for Miami Heat and Charlotte Bobcats within this season till their matchday 74. It can be seen that the probability of obtaining less than 40 victories for the Miami Heat is almost zero. The opposite is true for Charlotte Bobcats. The reason is straightforward: Miami Heat has a much better rooster than Charlotte Bobcats.
In order to obtain an objective coach performance measure it can be calculated the probability of obtain more victories than the actual. Next table shows the probability of obtain each of the possible number of victories greater than 36 for the Charlotte Bobcats.
As it can be seen, the probability of obtaining more victories than the actual (36 out of 74) is 2.71%. Then, to calculate an objective measure of the performance of the team it has to be calculated the inverse of 2.71%, that is 97.3% or 9.73 out of 10.

Next table shows the coaches’ performance using the methodology described above in the 74 first matches for each team such as the ratings provided by the ESPN.
As it can be easily observed several differences arise between both ratings. Particularly, the coefficient of linear correlation is 63 whereas the coefficient of Spearman rank correlation is 68%.

The best five coaches according to ESPN are Gregg Popovich, Tom Thibodeau, Doc Rivers, Erik Spoelstra y Rick Carlisle. They won five out of the last seven NBA Finals and it is straightforward that they were important in doing so. However, the performance of their teams is not as brilliant as expected, especially Indiana Pacers and Miami Heat.


On the other hand, Steve Clifford (Charlotte Bobcats) and Jeff Hornacek (Phoenix Suns) are situated in the top of the ranking from our methodology but both are not situated so high in the ESPN rank. Both teams have been two of the positive surprises in the NBA. Charlotte Bobcats was the second worst team in the NBA last season (2012-2013) while Phoenix Suns was the second worst team in the Western Conference. Subsequently, their expectations were rather low. However, both teams are in play-off positions 8 matches to go. Therefore, it is sensible that both coaches are leading the ranking.

The ESPN ratings seem to show some favoritism to big teams such as Miami Heat and Indiana Pacers. Both teams have very good roosters and it is expected a very good performance from them. For instance 9 and 10 matches for Miami Heat and Indiana Pacers respectively have a probability of victory higher than 87% while playing at home. Thus, even though both teams are leading the East Conference the number of victories from them is lower than a number to be considered as a great performance.

Boston Celtics is other example of sharp difference. Brad Stevens was ranked 11th by the ESPN but 27th from ours, but 23 victories seems very few for them. The opposite is true for Mike D’Antoni. Even though Lakers has a very good rooster injuries have decreased a lot the Laker potential. As an example in the first game versus LA Clippers the embedded probability of Laker victory was only 20% playing as local. Hence, even though the Laker season is being awful it is not due to a bad coach performance but to bad luck for injuries.

In summary, so far the best coaches have been Jeff Hornacek (Phoenix Suns) and Steve Clifford (Charlotte Bobcats) with a very remarkable 9,73 out of 10 rating followed by Gregg Popovich (San Antonio Spurs). On the other hand the worst coaches have been Larry Drew (Milwaukee Bucks) and John Loyer (Detroit Pistons).
Hence, we should carry out this analysis for more seasons in order to get an accurate ranking of coaches. We are working on it!

Follow me on twitter: @jdelcorraltm

No hay comentarios:

Publicar un comentario