Our 2018 NFL preseason projections are out and they are free. Subscriptions for our season-long and DFS weekly projections are available now (previous year subscribers will get a discount code in July). They will be updated throughout the preseason. See below for the links. You can also access via the Tools/Projections section in the main menu.
|All | QB | RB | WR | TE | K | FLEX | DEF(Team) | DEF(IDP) | VS FANTASYPROS|
Here is an explanation of how the projection sausage is made. This summary is heavily cribbed from our FAQ document which also goes over our tools:
We have invested significant time in building a data model that strives for excellence in each component of a great NFL projection. The below is a top-level explanation. Please feel free to ask in-depth questions in the comments. The projection models for each component have all been created through backtesting against games from previous seasons.
Our preseason projections sum up the results of our weekly model for all 17 weeks of the NFL season (the rest of season model works the same way). This is the same weekly model that finished 1st in four of the last 10 weeks last year on FantasyPros. There are some small playing time adjustments made for games 2+ weeks in advance to account for the possibility of injuries and role changes. Injured/suspended players are removed until they are expected to play again.
It starts with leveraging multiple data sources to determine player roles and estimate snap counts for both the offense and defense that reflect today’s NFL. Yes, there are still a few ‘bell cow’ every-down RBs but there are also RB job shares, 1st/2nd down backs, 3rd down backs, etc. There are every-down WRs, WR3s who play a majority of downs, WRs who come out in 4 WR sets. Our model aims to distinguish across all those types and making sure snap counts reflect a team’s offensive philosophy (e.g., some teams are heavy 3-4 WR, some are heavy 2 TE, etc.). Official injury status (particularly ‘Doubtful) is used to discount snap projections with backups typically receiving the allotted snaps. This same snap allotment process is used when ‘starters’ are Out. This allows us to dynamically handle last minute injury news.
Team total plays and rush/pass split are projected based on team and gameday variables (spread, home/away, is the starting QB playing, etc) while points are based on Vegas over/under and spread.
Player carries and targets are distributed based on player role, skill, projected snaps, and previous game rush/target distribution.
Player rate stats (completion rate, yards per carry, yards per target etc) are based on previous game stats (including aDOT for 2018), player scouting grades, and gameday variables (surface, weather, opposing defense, QB (for WRs)). Defensive player stats look at similar stats/grades/data and mirror the offensive stats (e.g., tackles are a percentage of the offense’s runs + completed passes).
Player TDs are based on past game red zone usage, projected rush/target volume, and height/weight and are adjusted to match the Vegas team point projection (e.g., projected TDs + FGs = Vegas line).
Hit me up in the comments or on Twitter (@rudygamble) with any questions/feedback on the projections. We have a few awesome tools/reports in the works so stay tuned!