Where Football Analytics Is Discussed and Published Today

Football analytics writing used to live almost entirely on social media, in threads that were useful the day they were posted and close to unsearchable a year later. Most of the field’s actual working infrastructure — the datasets, the packages, the competitions, the peer-reviewed research — now lives somewhere more durable, and knowing where to look is most of the battle for anyone trying to get past the surface-level version of the field.

That shift matters for a practical reason, not just an archival one. A social feed is organized around what’s being posted right now, which is a poor fit for a question like “what’s the standard way to compute expected points added” or “has anyone already built this metric” — questions that need a durable, searchable answer, not a real-time conversation. The venues below are organized the opposite way: around a stable body of code, data, or published research that’s still there, and still correct, months or years after it was first posted — the same durability that matters when building a public portfolio out of work posted to one of them.

The Open-Source Data Layer: nflverse

Nearly everything downstream in current public football analytics sits on top of nflverse, a community-maintained ecosystem of R and Python packages built around NFL play-by-play data going back to 1999. Its component packages — nflfastR for play-by-play, nflreadr for fast data access, nfl4th for fourth-down decision modeling, among others — are released on GitHub under a CC-BY-4.0 license, free and without an API key. Because so much public analysis is built directly on this dataset, understanding what nflverse actually contains is close to a prerequisite for following most of the writing described below.

The College Football Equivalent: cfbfastR

The NFL side of the ecosystem has a direct counterpart for college football. cfbfastR, maintained as part of the SportsDataverse project on GitHub, is an open-source R package that pulls clean, tidy college football play-by-play data — sourced from CollegeFootballData.com and ESPN — and includes tools for benchmarking open-source expected-points and win-probability metrics against that data. It plays roughly the same structural role for college football research that nflverse plays for the NFL: a shared, versioned, publicly maintained dataset that individual researchers build analysis on top of instead of each scraping and cleaning their own copy from scratch.

Where the Writing Lives: Open Source Football

Individual analyses posted only to a social feed tend to disappear from view within days, which is the specific problem Open Source Football was built to solve. Created by Sebastian Carl with Ben Baldwin — co-developer of nflfastR — as a key contributor, the site centralizes community-written analyses built on the nflverse ecosystem into a single, searchable, GitHub-backed archive, not leaving them scattered across timelines with no durable home.

Peer-Reviewed Research: JQAS

For work held to a formal academic standard, the Journal of Quantitative Analysis in Sports (JQAS) — published by De Gruyter and affiliated with the American Statistical Association — is the closest thing the field has to a flagship peer-reviewed outlet, publishing quantitative research across football alongside other sports. It’s a slower and more rigorously reviewed channel than a blog post or a community write-up, and it exists specifically for methodology that needs to hold up to formal peer review before it’s treated as established.

Public Competitions With Real Data

Two currently active competitions function as both a publishing venue and, for many participants, a hiring pipeline. The NFL’s own Big Data Bowl, hosted annually on Kaggle and powered by AWS since 2019, releases a season of player-tracking data around a themed problem and has the league’s own operations department citing more than 75 hires into sports analytics roles among past participants. The MIT Sloan Sports Analytics Conference, held annually since 2007, runs a Research Paper Competition with a dedicated football track alongside basketball, baseball, soccer and business-of-sports research, plus a separate case competition — both genuinely open to outside submissions, not limited to conference speakers.

GitHub Itself, as a Discussion Venue

A meaningful share of current football-analytics discussion doesn’t happen in an article or a forum post at all — it happens directly inside the GitHub repositories that host the open-source packages, in issue threads and pull request discussions. A question about why a specific expected-points value looks off for a given play type, or a proposal to add a new column to a dataset, tends to get resolved as a public, permanently archived conversation attached to the actual code in question, with the package’s maintainers participating directly — the same minimal, reproducible framing that gets a question answered anywhere else applies just as directly here. It’s a narrower and more technical venue than a community blog or a conference, but for questions about exactly how a specific number in the open data was computed, it’s often the most direct and most durable source of an answer.

That durability is the throughline across every venue described here. A social-media thread, a GitHub issue, a conference paper and a peer-reviewed journal article all differ enormously in how formal and how fast they are, but each one leaves something searchable behind — a fixed artifact someone can point back to months later, not a real-time exchange that’s effectively gone once the moment passes. That’s a fairly specific and fairly recent shift in how the field organizes itself, and it’s a large part of why current football analytics is easier to actually learn from than it was when most of it lived only in a fast-scrolling feed.

Outlets That Publish Their Own Models

A handful of publishers maintain and publish their own proprietary or semi-proprietary metrics alongside more general coverage. Pro Football Focus builds player grades from its own play-level charting across the league. FTN Fantasy has continued publishing Aaron Schatz’s DVOA (Defense-adjusted Value Over Average) ratings and methodology, a body of work with roots in the earlier Football Outsiders site, which closed in 2023. SumerSports publishes public articles and player-facing data alongside the team-facing analytics products it builds for NFL and college programs. None of these fully replace the open, reproducible ecosystem built around nflverse, but they represent a different kind of publishing — proprietary methodology paired with public output — that sits alongside the open-source community, not inside it.

A Direct Line to the Tooling: the nflverse Discord

For questions specifically about using the open-source packages themselves — a broken function, an unexpected column, a modeling question that depends on knowing exactly what a given field represents — the nflverse project maintains a public Discord server aimed at package support and broader NFL-analytics discussion. It’s a smaller and more technical space than the writing and competition venues above, but it’s the most direct way to reach the people who actually maintain the datasets everything else on this list is built from.