Outline — WTI Went Negative. Brent Didn't.

The skeleton. Prose is a rendering of this; if the prose drifts from the argument here, the prose is wrong. See #10.

Story #3, written the #10 way: outline and chart plan committed before any prose. Draws on oil-prices, enriched in the previous session — the negative-price observation and the "it disappears once averaged" point both came out of SUMMARY.md's "What stands out".

What this story is about

On 20 April 2020, the U.S. WTI spot price fell to -$36.98 a barrel** — the only negative value anywhere in eight EIA price series covering Brent and WTI, daily through annual, back to the 1980s. Brent, priced on the same day, stayed at **$17.36. This is a story about why the same commodity, priced two different ways on the same day, produced two completely different numbers — a landlocked storage hub versus a seaborne cargo. Not a story about oil-price history in general, and not about the wrangling; that gets one short paragraph at the end.

Argument, in order

1. The chart, first. Brent and WTI daily spot prices, March–May 2020. WTI dips below the zero line on 20 April; Brent, plotted alongside, does not. Show this before any prose beyond a one-line caption.
2. What these numbers are. Both series are EIA's daily "spot price FOB": Brent for waterborne North Sea cargoes, WTI for oil delivered at Cushing, Oklahoma — a landlocked pipeline and storage hub, not a port. Caveat, stated plainly: "spot price" is doing real work here. Around contract expiry, the near-month NYMEX futures price and the physical spot price converge; the collapse below zero was a futures-and-storage event (the expiring May contract, with Cushing tanks close to full during the COVID-19 demand collapse) that fed straight into the spot number EIA reports for that day. This is a spot-price series, not a futures series, but the two are not independent at expiry — worth one sentence, not a detour into derivatives.
3. What it says.
   • WTI: $18.31** (17 Apr) → **-$36.98 (20 Apr) → $8.91 (21 Apr). One trading day, a round trip of more than $55.
   • Brent, the same day: $17.36**. Brent's own all-time daily low across the full 1987–2026 record is **$9.10 (10 Dec 1998) — low, but never negative.
   • The dip barely survives even one level of averaging. WTI's weekly figure for that week (ending 24 Apr) is $3.32** — positive, but the closest any of the coarser series comes to zero. Monthly (April 2020, dated to the 15th): **$16.55. The year to 30 June 2020: $39.16. Both comfortably positive.
   • Of the 25,415 daily/weekly/monthly/annual observations across all eight series in this dataset, exactly one is negative.
4. The wrinkle: two ways to store the same commodity. Brent trades as waterborne cargoes — a tanker that cannot unload can, in the worst case, sail somewhere else. WTI settles at Cushing, a landlocked tank farm connected by pipeline; when its storage neared capacity in April 2020, there was nowhere else for a barrel to go. The same demand collapse hit both benchmarks; only the one with a fixed physical delivery point could be pushed negative. Nominal-dollars caveat: every figure here, including the 1998 and 2008 comparisons, is nominal USD as EIA published it — not adjusted for inflation.
5. How this was made. Brief. Both series come from oil-prices in this repo — EIA's own .xls workbooks, rebuilt by build.ts; the consolidated statistics table that surfaced the "exactly one negative value" claim is enrich.ts's work, local to this repo (not part of any published dataset). The community datasets/oil-prices package is the same underlying EIA source, independently re-wrangled; see this dataset's README.md for the row-for-row comparison.
6. Friction notes (keep — for the skill work):
   • The story's argument is a comparison between two resources at one moment (like story #2's scoreboard), not a single time series (like story #1) — but unlike story #2 it still resolves to a plain line chart, because both resources share units and a time axis. The "multi-resource but still a line chart" case wasn't covered by either existing outline.
   • enrich's consolidated stats table (this session's other task, datapressr-q96) is what actually surfaced this finding — scanning the min column across all eight rows is what turned up the single negative cell. Without a table that put every resource's min in one place, this argument would have needed a manual per-file scan to find. Worth noting in enrich's skill draft as a reason the consolidated table earns its keep, beyond tidiness.
   • The "spot vs futures" caveat in beat 2 is deliberately one sentence, not a derivatives explainer — matching the voice guide's "one clear caveat > three confident claims." A longer treatment would turn this into an article about futures markets, which is not the argument.

Chart plan

Both rendered by oil-prices-make-charts.mjs with Observable Plot per docs/charting.md — annotate the 20 Apr point directly on chart 1 rather than relying on a legend.

Voice

Plain and factual, per docs/voice-guide.md. Let the two numbers on the same day — $17.36 and -$36.98 — carry the contrast. No "unprecedented", no "the day the market broke". The futures/storage explanation is one sentence of mechanism, not a lecture. A "sounds like me" pass is a separate step the author runs.