Key Inputs in an NBA Projections System
The old adage, “you get out what you put into it,” applies to most things in life and that includes a daily fantasy basketball projections model. Your projections are only going to be as good as the inputs that create them. This lesson will not walk you through how to create a projections model (would take too long to go through all of the intricacies in Excel), but rather which inputs I deem most important in my model. If you have a decent amount of knowledge in Excel, you can create your own set of projections that you can customize to your liking.
First, we will go through each of my inputs and how much weight they have in my NBA model. The best part is that all of these inputs have been talked about in the lessons in this Blueprint Volume 2 course. After that, I will give my thoughts on using projection models as a whole and where they have their shortcomings.
Fantasy Points Per Game (FPPG) – Luckily, each site’s .CSV salary export has this information included. I use FanDuel’s FPPG for my FD projections, DraftKings’ FPPG for my DK projections, and so on and so forth. You need to account for the differences in each site’s scoring system. (Weight: 50-55% of total projection)
• Season Average FPPG – As with any projections model, it’s important to look at season long production, as well as recent production. I use both, with a slightly higher weight on season averages (65% of FPPG weight).
• Recent FPPG – The sample size for recent FPPG needs to be big enough that it carries value, but not too large that it starts to look identical to your season average FPPG. I’ve found that the last five games is a nice sweet spot in the middle. (35% of FPPG weight).
Minutes Per Game (MPG) – Minutes are king in daily fantasy basketball and we should treat them like gold. I get my minutes per game straight from the bird’s mouth: NBA.com. The previous lesson explained this in more detail. (Weight: 20-25% of total projection).
• Season Average MPG – Once again, we need to look at how many minutes a player averages on the season, as well as their recent playing time. I use the same weights for MPG as I do for FPPG. (65% of MPG weight)
• Recent Minutes MPG – Once again, I use five games as my sample size for recent playing time. (35% of MPG weight).
Turnover-Adjusted Usage Rate – Do you remember why we prefer this to the standard usage rate? Of course you do! We don’t want to reward players that are turnover prone, as turnovers are worth negative points in DFS. (Weight: 15-20%)
Team Totals – In our previous lesson, we mentioned the importance of team totals. We like to let Vegas do the work for us, because they must put out accurate lines. I compare team totals to the other teams on the schedule that night, as well as to each team’s average points per game. Both get a 5-7% weight. (Weight: 10-14% of total projection).
Defense vs. Position – Once you find a Defense vs. Position ranking system that you can trust (I personally use the statistics found here on RotoGrinders), you can input them into your NBA model. You will notice that the weight isn’t as large as the team total weighting is. (Weight: 5-10% of total projection).
Projected Pace – Pace is important, because the more possessions in a game, the more opportunities there are to score fantasy points. I look at projected pace and the difference between each team’s average pace and their projected pace. Both get a weight of 3-5%. (Weight: 5-10% of total projection).
Vegas Line – This is where the blowout potential comes into play. You can find the formula that I use for projecting blowouts in one of the previous lessons. Take that number and multiply it by the entire projection for a player.
The main variable in the projection is minutes. As I mentioned in that lesson, I create a minute projection for each player and then adjust each based on the eye test. All of the inputs listed above help me come up with my first projection.
The next projection that I come up with is based strictly on fantasy points per minute. I take the FPPG equation from above and divide by the MPG equation from above. Then, after I have made my adjustments and have come up with an accurate minutes projection for each player, I multiply it by the FPPG/MPG to come up with my second projection. I then take the average of the two projections, which I then compare to needed fantasy production based on each player’s salary on each site (I have an entire course on how to create a value-based expectation system).
Once you have all of your inputs, the only real variable each day is the minutes. Of course, you do have to update all of your statistics, but that part is easy. I have found that using an NBA model has really helped me avoid biases against certain players – whether I take someone too often or not enough.
I do not use a lineup optimizer with my projections. This is where my old school mentality comes into play. I feel that one of the reasons I have been so successful is my ability to create solid lineups. Lineup optimizers are based strictly on a points-per-dollar basis and can end up ignoring obvious value plays. When there is an obvious value play, you should target them regardless of what their projection is.
My final note is that while projections can be a very useful tool, you have to constantly adjust your inputs and weights in order to perfect your projections model. You always have to be looking to improve, even when you are winning.
If you haven’t read the Blueprint Volume 1, check it out by clicking here:
https://rotogrinders.com/market-place/notorious-s-blueprint-to-daily-fantasy-basketball-178
Additionally, here are the other courses that I offer:
Injuries in Daily Fantasy Basketball
What is Value and How to Find it in DFS
Vegas Lines and Opportunity