This is where it all began. I built my first NBA betting model back in 2013 as a hobby to see if I could crack the NBA betting code. I had already dipped my toe into sports betting, but the idea of manually handicapping each game seemed incredibly time-consuming and left myself vulnerable to my own biases and assumptions – which as a passionate Chicago sports fan, I had plenty (read: fuck the Packers).

I am a hyper-analytical person though and have always loved reading through stats in all sports, so creating a model made a lot of sense to me. Without much of a statistics background, my initial model was incredibly basic, but I could see the potential. I had played and watched basketball all my life, so I understood which stats and information could predict the outcome of a game. I had also read a few books in my day, so I was a bit more open to analytics than Charles Barkley.

Since then, I've added a few more ingredients each season to my model. It is still relatively simple compared to other predictive models, so I won’t be working at FiveThirtyEight anytime soon, but it’s been successful nonetheless. Now, I unfortunately cannot give away every little detail. I want this site to be as transparent as possible, but at the same time I value my model and the work I’ve put into it. I will, however, go over the basics so you know what I consider for my projections and can feel more comfortable subscribing to my picks if you decide to do so.

First, the base of my model utilizes advanced statistics known as efficiency metrics. Since teams are limited in time by both the game clock and the shot clock, basketball is all about efficiency. Maximizing points scored and minimizing points allowed on each possession is the base of success on the hardwood.

Before I go deeper, it's important to understand how efficiency stats work. Since game totals can vary widely based on the teams playing (the slow, methodical Grizzlies and Jazz would have a much lower score than a game with the high-flying Rockets and Warriors), these stats adjust for pace (or possessions per game). Two of the basic efficiency stats, offensive and defensive rating, measure how many points a team scores and allows per 100 possessions. 100 is used because this is near the league average of possessions per game (101.13 currently for 2017) and it makes the stats look similar to average total scores.

The algorithms I have created with my model produce a predicted score based on how efficient each team is on offense and defense. I use more than just team offensive and defensive efficiency, however, as there are more variables that contribute to game outcomes and cause a team's efficiency to vary based on their opponent. The main variables include pace, effective field goal percentage, turnover rate, and rebounding rate. These all help give us a general idea of how a game will go, and I have weighted each stat in my model to reflect how much they affect a game.

If you follow the NBA closely though, you know that rest is also a huge factor in a team's performance. If Team A is on a long road trip and playing their second game in two nights, while Team B is home after two days off, you can confidently assume that Team B will perform better. It's not a guarantee, but the fatigue of playing and traveling is very likely to negatively affect Team A. Therefore, I factor a team's schedule into my model. In this example, Team A is downgraded and Team B is upgraded due to the imbalance of rest.

ESPN has put out a great article the past two seasons discussing this and what head coaches call schedule losses. They are situations where a team is at a large disadvantage due to their schedule (and their opponent's). ESPN highlights these games in this article and is a great tool for handicapping. It is updated each month for the next games to watch for. My model does something similar by measuring how teams do in certain schedule situations over the season. Since the general public doesn't consider this, it's a great way to find an edge.

Similarly, there is another variable to watch for with these scheduling issues. For those that don't regularly watch the NBA, it has become a common practice to rest healthy players. Due to the grind of the NBA season, especially with back-to-backs (two games in two nights) three-in-fours (three games in four nights), and long road trips, coaches may choose to give top players a night off to recover. The NBA has improved this issue the past few seasons and has eliminated the grueling four games in five nights, but this is still an important factor to consider.

Because of this, there can be a problem with merely using team statistics. Like my baseball model, I need to account for who is actually playing each day. That is why I incorporate their own individual efficiency stats along with their usage rate and average minutes per game to estimate their impact if a main starter is out of the line-up.

For example, the Cavaliers have a 109.0 offensive efficiency rating while LeBron James has an individual offensive rating of 111.5 (which means the Cavs score 111.5 points per 100 possessions with LeBron on the court). With a usage rate of 30.4% (12th most in the NBA) and averaging 38 minutes per game (2nd most), the Cavs offensive rating will be significantly less if LeBron sits out a game. My model attempts to estimate this drop-off.

By blending all this information together after years of testing and tinkering, my model spits out an impressively accurate projected score for each game and first half. If the projected score varies enough from the current odds, I make it a play and send that pick to my subscribers. I created a first half version of my model for this season, and it is off to an incredibly hot start.

So that's the main idea behind my model and all the metrics I consider. If your head hasn’t exploded yet from all the geek talk, then thank you for reading! Like I said, I can’t give away the exact formulas, but I want you to know the basic concept if you are considering subscribing to my picks or want to build your own model. If you are interested in building your own, feel free to ask me a question about where I get certain stats, how to get started, or anything else that comes to mind.

And lastly, if you would like access to my model’s picks and all my sports betting plays, then subscribe to Fast Break MVP now! You can find more details about my picks service here, and if you have any questions, don’t hesitate to email me or send me a direct message on Twitter.

Thanks for reading!

- Stephen