NBA model projections for Friday, October 27th
This is part of a recurring series on VSiN.com in which VSiN host Jonathan Von Tobel tracks the progress of a player-based model he created for the NBA season
Welcome to the first entry in our latest journey in predictive modeling! For those new to the process, let’s fill you in on what we are doing here.
Over the summer I decided that I wanted to take a deep dive into the process of building a model. There are many in the sports betting community who find their edges by building models based on parameters they deem to be significant. It’s a process that has always fascinated me, so I took a crack at it with an NFL model I lovingly call Jon Von Model.
The process I used to build the NFL model – which is based on regression testing and team-based statistics – was somewhat simple. So I tried to use those same parameters to build out an NBA model, but the results were inconsistent and wild.
I found that it was impossible to build an accurate model for the NBA based on team statistics. There are only five players on a court at once for a team. One single player’s production can shift the balance for a team, unlike the NFL where there are 11 players on the field at one, different personnel for offense and defense and a third aspect of the game in special teams.
So, I attempted to build and understand the concept of player-based models. Let me tell you that this is the most challenging thing that I have done in my career here at VSiN.
How do you project minutes for players? How do you boil down the value of a player to a single number? How do you calculate pace? What in the world is the command I’m looking for in Excel?!
These are just some of the questions I shouted out while sitting in my office while my family worried for my mental well-being.
After all that we have arrived at this day; the first day in which we will pump out game projections for the day of NBA action. Before we begin there are two things we need to establish:
First, those who are looking to get started on building any sort of model should look into the work of Andrew Mack. Mack’s books, ‘Statistical Sports Models In Excel’ helped me immensely with both the NFL and NBA models. This player model is based heavily on his work in the second volume.
Second, this model is more experimental and will be much more volatile than the NFL model. As I have stated in the journal with Jon Von Model 1.0, I do not believe that model is worth actively following yet and that is the same for this one. This journal is about the journey of building and maintaining a model for the first time. I expect these projections to be wrong much more often than not, but the hope is that by the end of the process we have a decent functioning model that is close to the market.
With that out of the way, let’s take a look at the projections for Friday’s NBA card!
Friday, October 27th Projections
The following numbers are the projected margin of victory – or spread – for the game listed.
Los Angeles Clippers (-1.658) at Utah Jazz
Detroit Pistons at Charlotte Hornets (-2.833)
Denver Nuggets (-2.403) at Memphis Grizzlies
New York Knicks (-0.286) at Atlanta Hawks
Miami Heat at Boston Celtics (-9.406)
Oklahoma City Thunder at Cleveland Cavaliers (-5.109)
Toronto Raptors at Chicago Bulls (-5.13)
Brooklyn Nets at Dallas Mavericks (-7.277)
Orlando Magic at Portland Trail Blazers (-2.423)
Golden State Warriors (-1.157) at Sacramento Kings
For example, the Clippers are favored by 1.658 points according to the model.
The most encouraging sign is that four games are within two points of the current consensus lines. According to the projections we also have four games that have three or more points of value, so for the sake of this experiment we will use these four games as our plays to track for the model this week:
Grizzlies (+5.5)
Bulls (-2)
Trail Blazers (+3)
Warriors (+2.5)
There are only two of these I vehemently disagree with: Grizzlies and Trail Blazers. Denver just doesn’t seem to be getting enough respect from the model, and part of it is due to a low rating of Jamal Murray. Murray did not have an incredible BPM last season, so his value is diminished by the model. As far as the Trail Blazers are concerned, I believe the unknown of Scoot Henderson and the massive BPM numbers for players like Malcolm Brogdon are altering the number.
Outside of those I do not have any real argument. Three points is essentially a possession so being that far off market – especially at the beginning of the season when the market is not as tight – does not give me much pause.
For now, this is how we will track this model. I will also start to include model projections for each game when I do the daily best bet write ups as well. Hopefully this is the start of another fun experiment that becomes a labor of love, as opposed to something I dread doing each week.
Good luck!
