This time of year, the NBA becomes a bit of a mess. With changing player lineups, rest days and mysterious "injuries" to star players, you can pretty much throw models out the window. Right?
Well, yes and no. While all of the above is true, it doesn't apply to every team. Many teams that are still jockeying for playoff spots and seeding. That said, it pays to be selective and cut down on your exposure. I only want to handicap and bet on games involving two teams that still appear to be making an effort, which naturally limits your action. And since models are based on past performance, teams with constantly changing lineups (think San Antonio Spurs) have made themselves more or less model-proof.
This season (so far) our record on NBA totals is 181-143. That's a 55.9% winning clip for a net of about 22.7 units in the black. We have also made some bets on sides. The record on point spread selections is 126-118. It's 51.6% winners, which in the standard 11/10 world of the bookmaker gives you a net of about -6 units in the red.
What do these numbers say? First of all, the model is not great at choosing sides for point spread bets, while it is pretty good at overs and unders. It's also true that totals are inherently easier than sides. This makes sense, since totals have as much to do with how a team plays, as it does with how efficient they are on any given night. So the relative success of totals vs sides is part model, part market.
We will probably have a pick or two on a busy NBA card, and if so, will post on them later.