Your Data

YUNO Sports runs on public data, most of it openly licensed. Here's where it comes from, what we do with it, and which numbers we compute ourselves.

How It Works

Automated pipelines pull play-by-play records, snap counts, injury reports, rosters, and other reference data from their source projects. We process and compress that data, then publish it to our servers as versioned artifacts the rest of the platform reads from.

Most pages — team pages, game pages, fantasy tools, leaderboards — run their numbers on the server and send you the finished result. The Lab is the exception: it streams the raw play-by-play data to your browser, so every combination of filters, metrics, and groupings recomputes on the spot without a round trip. Historical seasons stream in as you expand the range.

The Metrics

Two kinds of metrics live on this site, and you should know which is which.

From the source. Foundational advanced metrics — Expected Points Added (EPA), win probability, completion percentage over expected (CPOE), success rate — are computed upstream by the data providers using open-source, publicly documented statistical models. We present them as-is and don't alter the methodology.

Built by YUNO. Some numbers are ours: Elo ratings, the MVP Meter, Team Health scores, market-based power rankings, and others. We compute these ourselves and explain exactly how on each tool's methodology page — plain language, no black boxes. Where a metric involves a judgment call, we say so.

AI Features

A few features — the CHAD assistant, AI-generated summaries and chart titles, natural-language query parsing — send your prompt and the relevant aggregated data to a third-party language-model provider to produce a response. Every feature that does this is clearly labeled in the UI. If a feature isn't labeled as AI, it doesn't use a model provider.

Data Sources & Attribution

We're grateful to the open-source community that makes this kind of analysis available to everyone. YUNO Sports relies on the following projects:

• nflverse-data — The primary source of play-by-play data and game-level metadata. Published as open datasets on GitHub by the nflverse community, under the CC BY 4.0 license described below.
• nflverse/nfldata — A different repository from the one above, and worth separating. Schedules, final scores, and every betting-line field we hold historically — the closing point spreads the market power rankings are solved from, and the over/under totals — come from here, maintained by Lee Sharpe. It carries no license file, and its own dataset documentation names a source for draft picks, rosters, trades, logos and colors while giving none for the line fields. We name it wherever those numbers are shown, as a statement of where they came from rather than a grant we can pass on.
• nflfastR — The statistical engine behind the play-by-play data. EPA, win probability, expected pass rate, and other advanced metrics come from nflfastR's models, developed by Ben Baldwin and Sebastian Carl.
• nfl-data-py — A Python interface to nflverse data, and how snap counts, injury reports, player ID crosswalks, and weekly rosters reach us. Maintained by Saiem Gilani and contributors. Note that it is the route rather than the origin for some of these — snap counts in particular are compiled by Pro Football Reference, below. Injury reports are read from the nflverse-data release files, which begin in 2009.
• Pro Football Reference — The origin of our snap count data. Every per-player snap figure on the platform — the participation tables on team and game pages, the Snap% column and snap-share charts on player pages, the “per snap” basis in the Lab, and the starter-retention component of Team Health — is compiled by Pro Football Reference and reaches us through nflverse. Their player IDs also appear in our records, linking players across datasets.
• ffopportunity — The source of our expected fantasy points (xFP), by Joe Sydlowski and Tan Ho of the ffverse project. Expected points measure the opportunity a player was already given — how many points an average player would have scored from that usage. They describe games that have already been played, and they are not a forecast of what anyone will do next week. Published as weekly release files covering 2006 onward; the models need target-depth inputs that play-by-play data does not carry before then, so earlier seasons have no expected points at all. We pin one model version and record which one in every file we build.

The live betting lines are not an open-source project

Everything above is published openly. The pre-game spread, total and moneyline on a game page are not, and they do not belong in the same list. They come from The Odds API, a commercial service we hold an account with, and the prices we display are the ones posted by DraftKings, which the card names alongside the time they were last refreshed. Neither is a partner or a sponsor of YUNO Sports, and neither has endorsed anything here. We show those lines to say what the market thought, not to redistribute anybody's feed. Nothing on this site is betting advice; see the Terms.

Where injury reports come from

Injury data is worth spelling out, because the chain has a link we cannot see the end of. We read weekly injury reports from the nflverse-data release files described above, and that hop we can point at directly. nflverse in turn collects them from an API, and its own documentation describes the source only as “an API for weekly injury report data” without naming the operator. That endpoint is not publicly documented, and we have not independently verified who runs it, so we do not claim to know. What reaches you is a designation and a body part as they were reported that week, by way of a project whose own upstream is unnamed.

The nflverse projects above — with the exception of nflverse/nfldata, which states no license at all — are published under the Creative Commons Attribution 4.0 International (CC BY 4.0) license. Attribution is required for any use or redistribution of this data, and any adaptation of it must also say that it was changed. Charts you export from YUNO carry both — “Data: nflverse (CC BY 4.0), modified” — and there is no setting that removes it — but the credit travels with the chart, not with you: if you crop it out, rebuild the figure, or quote the numbers somewhere else, the attribution obligation is still yours to meet.

That CC BY 4.0 license covers the nflverse projects. It does not extend to material those projects compile from elsewhere: snap counts originate with Pro Football Reference, whose own terms govern their data. Where a chart draws on snap counts we name them alongside nflverse, as a statement of source rather than a license grant — if you plan to redistribute snap-derived figures yourself, check their terms rather than relying on ours. The same limit applies to injury reports, with the added difficulty that we cannot tell you whose terms those would be: the grant is nflverse's own, and it does not speak for an upstream nobody has named.

Expected points carry a different license

ffopportunity is not covered by the license above, and the difference is not a technicality. Its models and expected-points data are published under the Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0) license — attribution, as with nflverse, plus a share-alike condition that CC BY 4.0 does not impose. The R code of that project is separately licensed under the GPL v3; we do not use the code, only the published data files. Charts and downloads that carry expected points name ffopportunity and its license alongside nflverse rather than inside the nflverse credit, because the two grants are not the same and one line claiming both would misstate at least one of them.

We show expected points as ffopportunity publishes them. We do not recompute, rescale, or blend them into our own numbers, and the files we build record that no values were changed. If you intend to redistribute expected-points figures or build something of your own on top of them, read the share-alike condition rather than relying on ours — it can reach work derived from the data, and that is a question about your use, not ours.

Legal Notices

YUNO Sports is an independent analytics platform. It is not affiliated with, endorsed by, or sponsored by the National Football League (NFL), any NFL team, Pro Football Reference, Sports Reference LLC, or any other entity mentioned on this page.

NFL-related trademarks, including team names and logos, are the property of the NFL and its member clubs. Team logos are displayed solely for identification purposes, and references on this platform are solely for informational and editorial purposes; neither implies any affiliation or endorsement.

The underlying play-by-play data is sourced from open-source projects that compile publicly available game information. Snap counts are compiled by Pro Football Reference and reach us through one of those projects. Injury reports reach us the same way, but the origin of that compilation is not publicly documented; see “Where injury reports come from” above. YUNO Sports does not claim ownership of any underlying data. Our contribution is the curation, processing, and presentation of that data through original software and analytical tools.

Metrics presented on this platform are produced either by third-party open-source models (presented without modification) or by YUNO's own documented methods. YUNO Sports makes no guarantees regarding the accuracy or completeness of any data or derived metric displayed on this platform.

This platform is provided “as is” without warranty of any kind, express or implied. In no event shall YUNO Sports or its operators be liable for any claim, damages, or other liability arising from the use of this platform or the data presented herein.

If you believe any content on this platform infringes on your rights, please contact us and we will address it promptly.

See also: Terms of Use · Privacy Policy