Our methodology for estimating total scores for NBA games relies primarily on a dual-computer simulation routine. We generally bet game totals where the two, independent simulation procedures 1) agree on an outcome and 2) point to a greater than five point difference between the actual betting line and the predicted total.
More recently, we have taken to cross-checking the computer-based output against the historical scoring tendencies of the involved teams. For example, tonight the Kings face off against the Nets at the Barclays Center in Brooklyn. Sacramento has played 18 games on the road so far this season, in which they have posted a median score of 102.5. Similarly, the 15 games the Nets have played at home feature a median score of 91. Thus the adjusted median expected score, per the old approach, would be 102.5 + 91 = 193.5.
We are now modifying the formula for determining adjusting median scores. The change attempts to account for the strength of the defenses on the floor during any given game. As an arbitrary example, consider that this year's Minnesota Timberwolves allow their opponents, on average, to score more points per game than all but one team in the league. Also, the Wolves give up the highest field goal percentage of any NBA franchise, and collect the third fewest defensive rebounds. In other words, Minnesota owns one of the worst defenses in U.S. professional basketball. As a result, teams put up significantly more points against Minnesota than they do when competing against premier defenses -- like that of the Indiana Pacers (to be precise, teams put up an average of 12.5 more points against the Timberwolves than the Pacers).
The refinement we have incorporated balances a team's old adjusted median against their competition's defensive prowess, relative to average. This is accomplished, quite simply, by multiplying a team's old adjusted median by the following:
[Average PPG Allowed by Opponent ÷ Average PPG of all NBA Teams]
Incorporating this factor into our process inflates scoring expectations for teams facing poor defenses, and reduces predicted values for teams contending with above-average defenses -- exactly achieving the desired effect. Going back to the Kings at the Nets matchup referenced previously, the new approach increases Brooklyn's expected productivity from 91 to 94 while reducing the forecasted level of scoring for Sacramento from 102.5 to 101. Thus the point estimate for the combined total increases from 193.5 to 195. In this case, the change in the total is not significant enough to matter since the line is set 10+ points higher, at 205.5. However, with the Kings team total currently fixed at 102 (per 5Dimes), this tweak does suggest changing a lean on Sac's total score from OVER to UNDER.
Happy betting!!