Quote Originally Posted by johnmillions:
Hey si1l,
I try not to bug ya but keep up the excellent work.
If you don't mind me asking, you had provided some details on how you're moving your focus in NBA in your model from the "four factors" to shooting accuracy and some new data. Can't find the post now. Care to elaborate any more on that? Does anyone have the link to that post?
I'm curious what kind of numbers you're going to use most for NBA and how you'll compile them..
Cheers !!
I'll go into greater detail when I start publishing the sheets, but I can elaborate on my research findings to give you an idea on how it influenced the redesigning of my models.
If you're a traditionalist in terms of advanced NBA statistics, you'd swear by the 'four factors'. Those are: offensive rebounding differential, free throw rate differential, turnover rate differential & effective field goal percentage differential. Each of these four factors has a weight attached to it to describe how much of an influence it has on the outcome of a game. Shooting (eFG%) was said to influence 45% of the game. Turnovers 25%, Offensive Rebounding 20% and Free Throw's 10%. These are just general guidelines and can fluxuate a few percentage points depending on the expected style of the game. I believed this was true because people told me to. Well, after losing 13 units in the regular season last year and hitting only 48%, I wanted to know why.. so I dug into the data.
I did a series of logistic regression studies using the last 12 years of data to extrapolate the relationship between the four factors and margin of victory and season winning percentage. The results were surprising. It turns out that offensive rebounding, turnovers and trips to the free throw line have almost no correlation to MOV and W%. The only variable that does - and it does with an R^2 value of .857 - is effective field goal percentage. The rest of the equation is superficial. So I used simple logic to find out why.
Free throws are highly dependent upon home court advantage and how the refs are calling the game that night. This is unfortunately quite inconsistent from game to game. Even if you did have data to suggest an advantage on one team or another - you just couldn't be confident that it would play out that way in reality.
Offensive rebounding rate and turnover rate are essentially measures of possessions - and consequently field goal attempts. The more rebounds and fewer turnovers, the more shot attempts. It's really that simple. In 2012, teams averaged 81 shots per game and made 36 of them. Let's say Team A out-rebounds their opponent by +6 and forces a +4 turnover differential. This would be considered a substantial advantage in both departments. On those 10 additional possessions, that team attempts 8 additional shots. So let's say that translates to 85 attempts for Team A and 77 attempts for Team B.
Team A - 85 ATT
Team B - 77 ATT
If both teams shoot the same eFG% (2012 league average 48.6%):
Team A - 85 ATT * 48.6% * 2 PTS = 82 points + free throws
Team B - 77 ATT * 48.6% * 2 PTS = 75 points + free throws
If Team B outshoots Team A by 3.0%:
Team A - 85 ATT * 47.1% * 2 PTS = 80 points + free throws
Team B - 77 ATT * 50.1% * 2 PTS = 77 points + free throws
If Team B outshoots Team A by 5.0%:
Team A - 85 ATT * 46.1% * 2 PTS = 78 points + free throws
Team B - 77 ATT * 51.1% * 2 PTS = 78 points + free throws
So even if Team A finishes the game with pretty significant advantages in rebounds and turnovers... Team B can overtake a 7 point deficit simply by shooting 5% better than Team A. This should illustrate the power of effective field goal percentage.
In a game with an equal number of shot attempts for both teams +1 eFG% is worth about 2.6 points. In a game where one team attempts 8 fewer shot attempts +1 eFG% is worth about 1.4 points.
After doing countless studies like this dealing with the relationship between the four factors and MOV and W%, I determined that the most accurate way to project an NBA game is to determine two things: 1) which team will get more shot attempts and how many and 2) which team will shoot better and by how much. I reworked my models to reflect these findings. And quite frankly, I can't wait to get my money back from my bookie this season.