EP 01 — The Receipts
Every chart, replay, and clip we made for the airport episode, in one place. Take them, share them, check our math. If you post one somewhere fun, tell us.
One ordinary day, replayed
Not a storm day, not a holiday — a random June Thursday. Every dot is a real flight from the federal on-time record, replayed by a discrete-event simulation. Mornings run clean; by 8 p.m. half of departures are late, because delay compounds down each aircraft's rotation all day. Passengers ate 1.01 years of accumulated delay before midnight.
Winter Storm Elliott — the meltdown
Same simulation, worst day in the picker: the storm that broke an airline. Watch a regional weather event become a national common-cause failure as airport queues pass the delay across airlines. This day (plus the CrowdStrike outage and the first nationwide ground stop since 9/11) is loadable in the interactive model.
The cheat sheets
The headline chart: it isn't the weather. In the federal cause data for that day, late-arriving aircraft — the delay your plane inherited from wherever it flew before you — accounts for 40.3% of delay minutes. Weather gets 7.9%. Five to one.
The traveler's version, built from all 7,001,619 US domestic flights of 2025: book before 9 a.m., fly Tuesday or Wednesday, pad afternoon departures 60–90 minutes, never book the last flight out.
Deep cuts
Who runs late (2025)
Airports with 60,000+ departures and the major airlines, ranked by share of flights 15+ minutes late, with the p90 delay alongside. DFW 28.6% late at the bottom, SLC 15.8% at the top.
Newark, month by month (2025)
Made for a Reddit thread that asked whether EWR is really that bad. (Of the ten big evening hubs, it came out least bad in 2025.)
Clips
The day replay, narrated
Cascading delays, argued
Thomas blames the weather. The data has other ideas.
The numbers behind all of it
Every US domestic flight since 2003 — 150+ million of them, one parquet file per year — is public on the EP 01 data page, queryable straight off the URL with DuckDB. The simulation methodology (and its known simplifications) is on the table in the model itself.
Reuse
The underlying data is the U.S. DOT/BTS Reporting Carrier On-Time Performance table (public domain). Our charts, renders, and clips on this page: share them anywhere, no permission needed — a link back to mathvsvibes.com is appreciated. Found an error in any of it? Bring receipts to the comments.