About DraftAxis
A scouting-analytics site for the NFL Draft. Dashboards covering executive tendencies, pre-draft visits, and team-by-team comparisons.
How It Started
DraftAxis started as a solo R Shiny project. The question was which measurable traits specific NFL front offices actually prioritize. Mock drafts and consensus boards were everywhere. Systematic views of the people making the picks were not.
The first version tracked pre-draft visits team-by-team. Later versions added positional tendencies for each GM, then league-wide comparisons on the same axes.
This Next.js site replaces the Shiny app. The dashboards here use the same scouting dataset.
The Dashboards
Exec Profile
How each GM's picks cluster by athleticism and position, against the league median.
Draft Visits
Team-by-team pre-draft visit usage. Which players each club brings in, how often those visits become picks, and how the visit types (Top-30, Combine, Pro Day, Private) compare.
League Comparison
All 32 teams on the same grid, ranked and plotted by visit volume, pick conversion, and positional preference. Filter by round, visit type, and availability at the pick.
How We Approach It
Team Behavior
The focus is on what front offices do: who they visit, who they draft. Not prospect predictions.
Teams First
Every view is organized by team or exec, not by prospect.
Public Data
Data from the NFL Combine, transaction wires, and beat-writer visit reports.
Readable Charts
Every chart includes a short note explaining how to read it.
How It's Built
The analysis pipeline is R, inherited from the original Shiny prototype. This Next.js site is a separate project that reads the pipeline's output.
- R for data preparation and the underlying analytics
- Next.js + React + TypeScript for the web application
- Observable Plot for the charts and hover tooltips
- Tailwind CSS for styling
How the Mock Draft Works
The cross-board mock draft comes from a model trained on roughly a decade of past first rounds, combining consensus rankings and expert grades with measurables, pre-draft visit signals, and exec tendencies. Instead of building the board pick by pick, the model solves all 32 slots together as a single assignment problem, so positional runs and fits are considered globally rather than greedily.
Three board variants — SAFE, BLENDED, and CHAOS — reflect different mixes of consensus signal versus team-specific behavior. Results are best read as a guided prior, not a prediction.
Data Sources
DraftAxis is built on publicly available data. Credit to the people and projects that compile and maintain it:
- Pre-draft visits — compiled from CBS, Pro Football Network, Walter Football, and beat-writer reports on X.
- Combine and Pro Day measurables — via array-carpenter/nfl-draft-data.
- Mock Draft Data — via nflmockdraftdatabase.com.
- RAS (Relative Athletic Score) — Kent Platte (@mathbomb).
- Historical draft picks — nflreadr, by Tan Ho and the nflverse contributors.