What's the one-line summary?

US GDP misses send EURUSD up (7 of 8 big-miss prints) and beats send it down (6 of 8 big-beat prints) in the first 15 minutes, with both median AND pct-up walking monotonically across all 5 buckets. The miss-side response HOLDS through 4 hours (big-miss median +15.3p 87.5% up at 4h, essentially unchanged from 1h). The beat-side response REVERSES: big-beat median goes from -10.0p at 15m to +0.5p at 1h to +14.0p at 4h — the initial USD-strength on a GDP beat fully unwinds and flips. First-ever US GDP × EURUSD anchor in the ledger.

n=8 in each tail is small — how much can we lean on this?

Small enough that both tails are drawn faint on the chart (below the ledger's MIN_N=30 threshold). Two things prop up the reading anyway. First, the whole walk is monotonic — every consecutive bucket points the right way on both median and pct-up, which is unlikely to be pure noise across 5 buckets. Second, the direction agreement is tight: 7 of 8 big-miss prints up, 6 of 8 big-beat prints down at 15m. The lone big-miss down print is 2026-02-20 (-11.9p, actual 1.4 vs consensus 2.8 — Q4-2025 print released during a February 2026 EUR selloff week). The two big-beat up prints are 2019-04-26 (+11.0p on actual 3.2 vs consensus 2.2) and 2020-07-30 (+5.4p on actual -32.9 vs consensus -34.5 — the COVID Q2 rebound where dollar was flowing in both directions). Everything else in each tail points the textbook direction. Small n means wide CIs, not wrong sign.

Why does the beat side reverse by 4 hours but the miss side hold?

The most likely reading is asymmetric information persistence. A US GDP beat is a backward-looking one-off print (Q1/Q2/Q3/Q4 quarterly revisions in April/July/October/January) — the market reacts to it in the first 15 minutes as 'stronger economy, hawkish Fed, USD up' but the print doesn't change the current-quarter path meaningfully, so the intraday rally fades and can reverse by 4h. A US GDP miss carries the opposite signal but with more weight: 'the growth story is breaking, Fed will need to ease' is a harder narrative to unwind in the same session because it lines up with what the pre-print rate-cut positioning has been probing for. So the miss-side signal compounds through the session while the beat-side signal decays. The exact channel here is speculative — the data just shows the pattern.

How does this compare to the same event on USDJPY?

The USDJPY sibling (us-gdp-usdjpy-directional, 2026-08-20) has 7 of 8 big-miss prints sending USDJPY DOWN at 15m (87.5% down on USDJPY = same USD-weakness as EURUSD's 87.5% up here). The miss-side response amplifies 25% by 4h on USDJPY — the same direction as EURUSD's miss-side amplification here (though the size differs). The USDJPY sibling didn't have the same beat-side-reversal story: USDJPY's beat side is muddier (2 big-beat prints at 15m don't share a clear direction and the 4h number isn't a reversal). So the reversal-by-4h beat-side story is EURUSD-specific here. Both posts agree on the miss-side persistence, though.

Can I reproduce this in the tool?

Yes. Event = US GDP q/q, instrument = EURUSD, window = 15m for the main table. Toggle through 30m, 1h, 4h to see the beat side fade toward zero then reverse. The releases table lists all 71 individual prints — including the 2026-02-20 lone big-miss-down print (surprise_z = -2.03), the 2019-04-26 and 2020-07-30 big-beat-up prints, and the full history from Q4-2012 forward.

← All insights

Analysis24 August 2026· 5 min read

US GDP q/q on EURUSD: 71 releases with monotonic median AND pct-up walks at 15m — but the beat side REVERSES sign by 4 hours while the miss side holds

71 non-contaminated US GDP q/q releases on EURUSD, 15-minute window. Both the median walk (+10.65 / +8.75 / +3.40 / -2.40 / -10.00p) and the pct-up walk (87.5 / 80.0 / 64.5 / 42.9 / 25.0%) are perfectly monotonic across all 5 buckets. 7 of 8 big-miss prints sent EURUSD up; 6 of 8 big-beat prints sent it down. The story flips by 4 hours — miss side holds at +15.3p 87.5% up, big-beat median REVERSES from -10.0p to +14.0p 62.5% up.

First-ever US GDP × EURUSD News Impact post in the ledger. Prior US GDP anchors covered USDJPY (2026-08-20) and gold; EURUSD is the natural third leg.

Six line panels stacked, one per window (1m / 5m / 15m / 30m / 1h / 4h). Each panel plots the median move in pips for the five surprise buckets (big_miss, small_miss, in_line, small_beat, big_beat). At 15m the walk is monotonic: +10.65, +8.75, +3.40, -2.40, -10.00 pips. At 1h big-miss stays at +15.2p while big-beat has collapsed to +0.45p. At 4h big-miss is still at +15.3p (essentially the same as 1h) while big-beat has reversed sign to +14.0p. Miss side persists across the entire window set; beat side amplifies through 15m then fades and reverses. Big-miss and big-beat points are drawn faint because n=8 is below the display threshold of n=30. — US GDP q/q × EURUSD across all 6 windows — miss side holds; beat side fades and flips by 4h.

The full 5-bucket table at 15 minutes

Bucket	n	Median	Pct-up	Reading
big_miss	8	+10.65p	87.5%	7 of 8 up; USD weak
small_miss	10	+8.75p	80.0%	textbook: US GDP miss → USD down
in_line	31	+3.40p	64.5%	biased mildly up
small_beat	14	-2.40p	42.9%	textbook: US GDP beat → USD up
big_beat	8	-10.00p	25.0%	6 of 8 down; reverses by 4h to +14.0p 62.5% up

Big-miss prints listed individually

n=8 is thin enough that every print earns its own row. The surprise-z threshold is |z| ≥ 1.5.

Date	z	actual / cons	15m move	Direction
2013-01-30	−2.84	−0.1 / +1.1	+9.90p	up
2013-04-26	−1.73	+2.5 / +3.1	+20.10p	up
2014-04-30	−2.09	+0.1 / +1.2	+11.40p	up
2016-07-29	−2.29	+1.2 / +2.6	+35.30p	up
2020-04-29	−1.94	−4.8 / −4.0	+2.90p	up (COVID Q1)
2021-07-29	−2.87	+6.5 / +8.5	+9.00p	up
2022-04-28	−2.35	−1.4 / +1.1	+16.50p	up
2026-02-20	−2.03	+1.4 / +2.8	−11.90p	down (the one exception)

7 of 8 up. The lone 2026-02-20 down print landed during a broader EUR-selloff week (EURUSD had fallen 90 pips over the two prior sessions on ECB dovish repricing); the GDP miss came in the same direction as the market’s pre-existing USD-strength flow.

Big-beat prints listed individually

Date	z	actual / cons	15m move	Direction
2013-07-31	+1.65	+1.7 / +1.1	−28.90p	down
2013-11-07	+1.82	+2.8 / +2.0	−36.80p	down
2014-07-30	+1.51	+4.0 / +3.1	−21.50p	down
2017-08-30	+1.94	+3.0 / +2.0	−8.00p	down
2018-03-28	+2.07	+2.9 / +2.0	−4.70p	down
2019-04-26	+2.76	+3.2 / +2.2	+11.00p	up (outlier)
2020-07-30	+3.56	−32.9 / −34.5	+5.40p	up (COVID Q2)
2022-01-27	+1.66	+6.9 / +5.3	−12.00p	down

6 of 8 down. The 2020-07-30 up print is the COVID Q2 rebound (actual -32.9% was a smaller contraction than the -34.5% consensus; dollar was already selling off broadly on Fed’s ultra-dovish stance in mid-2020). The 2019-04-26 up print sat inside a broader EUR-strength week; the specific mechanism is harder to explain from the print alone.

The miss-side persistence and beat-side reversal

The most notable thing about this pair-event isn’t the 15m table (though the double-monotonic walk is clean) — it’s the asymmetric multi-hour arc.

Window	big_miss med	big_miss pct_up	big_beat med	big_beat pct_up
1m	+9.90p	87.5%	-7.40p	0.0%
5m	+13.25p	75.0%	-11.00p	11.1%
15m	+10.65p	87.5%	-10.00p	25.0%
30m	+13.20p	62.5%	-7.85p	37.5%
1h	+15.20p	87.5%	+0.45p	50.0%
4h	+15.30p	87.5%	+14.00p	62.5%

Miss side stays at +15p 87.5% up through the whole window set. Beat side amplifies through 5-15m, holds around -10p at 30m, halves to zero at 1h, then flips positive to +14.0p 62.5% up by 4h. The initial USD-strength on the GDP beat unwinds and reverses; the initial USD-weakness on the GDP miss holds and even mildly deepens.

Verification note

All numbers verified against live /api/v1/news-impact/stats and /api/v1/news-impact/releases on 2026-08-24. The 71-release sample spans Q4-2012 through the 2026-07-30 print. Chart generated via the standard pnpm insights:chart CLI for all 6 windows; big_miss and big_beat markers are drawn faint because n=8 is below the MIN_N=30 display threshold. First-ever US GDP × EURUSD anchor; the USDJPY and gold siblings are 4 and 20+ days old.

Keep reading