Bellcow GM / Data Coverage
Data coverage
Last updated 25 August 2026 · figures describe the 24 August 2026 intelligence bundle
What Bellcow knows, how often it finds out, and — the part most products leave off — what it doesn't know. If a feed is late or switched off, the app names it on the Data Health page and the confidence on your plan drops with it. Nothing here is aspirational: if it isn't live, it says so.
What's in the bundle right now
1,071 players valued
Every rosterable offensive player plus team defences, with a weekly projection and a rest-of-season value.
396 with market values
Live trade-market values from Stats Guy Fantasy cover the players who actually trade. The rest are priced by the engine alone.
Counts come from the published bundle and move with each refresh. The app always shows the live figure on its Data Health page.
Sources, and what each one is allowed to do
| Source | Provides | Licence | Status |
|---|---|---|---|
| nflverse | Play-by-play, snap counts, participation, rosters, depth charts, official injury reports, schedule, game lines, combine & draft, historical archives | CC BY 4.0 — commercial use permitted | In production |
| Stats Guy Fantasy | Live trade-market values from real trades across thousands of leagues | Free & open — credit required | In production |
| National Weather Service | Public forecast data for outdoor stadiums | US government — public domain | In production |
| Sleeper | Your league only: rosters, settings, transactions, plus an add/drop velocity signal | League connector | League data only |
| ESPN | Your league only, by league ID (private leagues via your own browser cookies) | League connector — legal review open | League data only |
| Yahoo | Your league, by draft file or roster paste today | Direct sign-in pending Yahoo's API approval | Import only |
| Team sites & beat reporting | Breaking news and practice-report discovery | Pending per-site terms review | Built, not switched on |
| Commercial injury / inactives feeds | Real-time inactives, 90 minutes to kickoff | Would require a paid licence | Not purchased |
| FantasyPros, PFF, RotoWire, 4for4, ETR, Fantasy Points, RotoViz, Footballguys | Nothing. Used to benchmark our output when evaluating a change. | Licence required to ingest | Never ingested |
Sleeper is never a football-intelligence source. It supplies league state — who owns whom — and an injury-status overlay, and that is the limit of what it is allowed to influence. The projections, values and rankings are computed by Bellcow's own engine from the licensed sources above. The full legal data-sources statement →
Refresh cadence
A guard sits in front of every bundle upload: a refresh that produces fewer than 500 players is rejected rather than published, so a partial scrape can never quietly replace a good bundle with a thin one.
What “Last Updated” means
It is the timestamp of the last successful refresh of that specific category — stats, rosters, depth charts, injuries, schedule, market values — not the time the page was loaded and not the time the bundle was published.
This distinction caused a real bug, which is why it is spelled out here. Per-source status used to be stamped at publish time, so a refresh that had stalled kept reporting “healthy” next to a timestamp that was days old. Status is now recomputed against the clock every time you look at it. Each category has its own tolerance:
| Category | Considered stale after |
|---|---|
| Injuries | 6 hours |
| Depth charts | 12 hours |
| Stats & play-by-play | 24 hours |
| Rosters | 24 hours |
| Market values | 48 hours |
| Schedule | 7 days |
When a source is late
Bellcow degrades in the open rather than failing quietly. In order:
- The category is marked stale on the app's Data Health page, by name, with its real age.
- Confidence on your plan drops automatically — it can never read better than the freshness of the data behind it.
- The rest of the product keeps working from the last good bundle. You get a plan, and you get told exactly which part of it is standing on older ground.
- If the delay is long enough that a recommendation would be unsafe — an injury feed that has gone quiet, for example — the affected advice is labelled rather than dressed up.
Data Health also flags league-level limitations that are not about our feeds at all: a league imported without scoring settings, a FAAB budget inferred from team balances, a schedule that had to be modelled because the platform doesn't expose one, or a league that hasn't synced in a week.
What confidence actually measures
Every player carries a completeness score out of 100, built from how much real evidence exists behind him right now: stats coverage, injury-report coverage and depth-chart coverage. Your plan confidence is the average of those scores across the players in your three moves — the ones being added, dropped or traded — and it is then capped by data age.
Average player completeness 85+ and the bundle refreshed inside 24 hours.
Completeness 65–84, or a bundle between 24 and 72 hours old, or a league sync over 12 hours old.
Completeness under 65, or anything older than 72 hours. Treat the plan as a starting point, not an instruction.
Confidence is not a probability that the move works. It is a statement about how much the engine had to work with. A high-confidence recommendation can still be wrong — fantasy football is played by people, not by projections.
What is not live
Breaking news and practice-report discovery. The event engine that reads team sites and beat reporting is built and switched off, pending a per-site terms review. We are not going to scrape publishers into a paid product before we have established that we are allowed to. Until that review finishes, the news you see in Bellcow comes from official injury designations via nflverse plus your own league's transactions, filtered through your roster.
Gameday inactives. There is no commercial inactives feed behind Bellcow, because we have not bought one. That means Bellcow cannot see a scratch announced 90 minutes before kickoff. Re-check your lineup before kickoff on Sunday; the app tells you so on the Data Health page rather than letting you assume otherwise.
Direct Yahoo sign-in. Pending Yahoo's API approval. Yahoo leagues come in today through a draft file or a roster paste, and the connect screen says exactly that instead of pretending.
What Bellcow does not cover at all
- IDP leagues — individual defensive players are not valued. Offence plus team D/ST only.
- Best ball — no lineups to set and no waivers to work, so there is nothing for a GM to do.
- Dynasty — Bellcow optimises for winning this season, not multi-year asset value or rookie picks.
- Guillotine — the elimination format needs its own strategy model.
- DFS, betting and season-long prop markets — out of scope. Bellcow is a league-management tool.
- College and rookie scouting — combine and draft data are used as priors for NFL players, not to rank prospects.
If you are in a format Bellcow doesn't handle well, it says so on the connect screen rather than quietly giving you bad advice.
Your league data
Bellcow reads your league; it never writes to it. There is no lineup submission, no waiver claim and no trade sent on your behalf, on any platform. Your league data is used to produce your plan and your alerts, is never sold or shared, and is never used to train anyone's AI models. Privacy Policy → · Security →
Related
Methodology — how Bellcow computes what it computes
Live system status
A full sample GM plan
Legal data-sources statement