What can I conclude from this? A theory of mine is developing for the NBA regular season. Take a look at the 1) expectations for a team for the season and 2) their effort level.
That seems to account for most of the variability in spread-covering percentages for each team.
Essentially, if a team is expected to be very good and they are very good, and they try at the same level of effort as before, they will go roughly fifty percent.
But if a team exceeds expectations, then when? Well they are likely to cover more than their share of spreads. If a team drops below expectations, they will do less.
The Cavs tried really hard and got a high level of effort from each player. Thus they covered more spreads. Orlando also (seemingly) tried hard this season. The Pistons? They didn't try hard at all. And they failed to cover the spreads because they had an established identity as a hard-working team.
Interesting that these patterns for season-long betting emerge. Other research points would include how long into the season do winning percentages and records solidify? I mean how long into the season, if a team has a record of say, .400, will they finish the season with roughly that record? Ten games? Twenty? Or forty?
And once that record is roughly locked in, do the betting lines reflect the point spread from the start of the seasons's expectations? Because what matters is how quickly and accurately the lines can reflect changes.
What I've read about NFL or NCAA football, and NBA basketball, indicates that once people know how good a team is they are very good at predicting point spreads and who will win. So what matters is when their is a change in a teams skill level whoever can identify it fast enough, and accurately, stands to make the most money. It doesn't matter if they get better or worse, what matters is accurately detecting change. And being able to disregard statistical noise.