#TCblessed... kind of
Posted: Thu Aug 30, 2018 9:01 pm
After the feedback on my last article, I decided to redo the analysis and article but using a different definition of a TC boost. Last time I looked just at players whose potential color rating increased from season to season. As Wig pointed out, this could have happened because of paid player training and as I pointed out, this also won't capture players that get their potentials boosted but do not flip rating colors.
The new definition is also imperfect but I think it is better. This time I defined a boost as a player that had more than one potential rating improve to a better letter. This is better because it definitely only captures TC boosts because you can only pay to train one potential rating per year. So even if a player received paid training in a given year, they will only be counted as receiving a boost if they also had an increase in potential from TC. It is also better because it captures boosts that may not change a player's color rating. One weakness it has is that it will miss boosts that only improve one category. So I think this one should be better.
The New Pictures
Under the new definition, the Hawks are both more and less blessed. More because no they have the undisputed crown for most boosts and less because there are quite a number of other teams in the neighborhood (e.g. Wizards, Bucks, Spurs, and Hornets). I mean, sure they still have 15 more boosts than say, I don't know, the Blazers, but they are certainly not in a league of their own.
I also wanted to take a more sophisticated shot at the question of whether winning is correlated with more TC boosts. To do this I included the pat season's winning percentage of a player's current team in the regression. Using the past season should help with endogeneity (i.e. they win because of the boost not get a boost because they win). Looking at the regression results this time I noticed that I made a number of mistakes last time so please disregard the past results. This time age is negative and significant. This is good. It shows that the analysis is better this time. We all know that players older than 30 or even 29 or 28 don't get boosts. Being older definitely means you're less likely to get a boost. The team's winning percentage from the previous year is not correlated with boosts. So that crazy GM's conspiracy theory is just not true. Looking at the R squared, this regression explains about 3% of the variation. Read as: boosts are random.
So yes, the Hawks, and other teams, have significantly more boosts than SOME other teams but not all. Their recent surge of 5 boosts over the past two seasons is most likely a combination of identifying good young players, paid training, and, of course, luck.
The new definition is also imperfect but I think it is better. This time I defined a boost as a player that had more than one potential rating improve to a better letter. This is better because it definitely only captures TC boosts because you can only pay to train one potential rating per year. So even if a player received paid training in a given year, they will only be counted as receiving a boost if they also had an increase in potential from TC. It is also better because it captures boosts that may not change a player's color rating. One weakness it has is that it will miss boosts that only improve one category. So I think this one should be better.
The New Pictures
- boosts1.PNG (81.8 KiB) Viewed 837 times
- boosts2.PNG (78.29 KiB) Viewed 837 times
I also wanted to take a more sophisticated shot at the question of whether winning is correlated with more TC boosts. To do this I included the pat season's winning percentage of a player's current team in the regression. Using the past season should help with endogeneity (i.e. they win because of the boost not get a boost because they win). Looking at the regression results this time I noticed that I made a number of mistakes last time so please disregard the past results. This time age is negative and significant. This is good. It shows that the analysis is better this time. We all know that players older than 30 or even 29 or 28 don't get boosts. Being older definitely means you're less likely to get a boost. The team's winning percentage from the previous year is not correlated with boosts. So that crazy GM's conspiracy theory is just not true. Looking at the R squared, this regression explains about 3% of the variation. Read as: boosts are random.
- boosts_reg.PNG (140.23 KiB) Viewed 837 times