Euros per US dollar · USD/EUR

Forecast outcome — 31 Aug 2026

Settled

Forecast issued on 31 Jul 2026, 06:12 UTC

31 days before the target. This is the scored version, chosen by the documented evaluation rule. The original estimates below are unchanged.

Result

Original reference: 0.8711; Observed ECB fixing, Thursday, 30 Jul 2026
Target observation: 0.8613; ECB fixing, Monday, 31 Aug 2026 · Derived from ECB daily reference rate
Realized return: −1.12%; Price at the target versus the reference
Outcome at ±1%: Down 1% or more; 0.8624 or lower
Outcome at ±5%: Between −5% and +5%; 0.8275 to 0.9146
Outcome at ±10%: Between −10% and +10%; 0.7840 to 0.9582
Outcome at ±50%: Between −50% and +50%; 0.4355 to 1.3066

Movement threshold

1% 5% 10% 50%

1% — see every threshold 5% — see every threshold 10% — see every threshold 50% — see every threshold

One outcome is scored at each threshold. The page keeps all four.

Original forecast · AI average — uncalibrated

As published on 31 Jul 2026. Nothing here has been rewritten after the outcome.

Down 1% or more

42%

0.8624 or lower

Happened

Between −1% and +1%

27%

0.8624 to 0.8798

Did not happen

Up 1% or more

31%

0.8798 or higher

Did not happen

The forecast put 42% on the outcome that happened.

Down 5% or more

0.8275 or lower

Did not happen

Between −5% and +5%

88%

0.8275 to 0.9146

Happened

Up 5% or more

0.9146 or higher

Did not happen

The forecast put 88% on the outcome that happened.

Down 10% or more

0.5%

0.7840 or lower

Did not happen

Between −10% and +10%

98.7%

0.7840 to 0.9582

Happened

Up 10% or more

0.8%

0.9582 or higher

Did not happen

The forecast put 98.7% on the outcome that happened.

Down 50% or more

0.0%

0.4355 or lower

Did not happen

Between −50% and +50%

>99.9%

0.4355 to 1.3066

Happened

Up 50% or more

<0.1%

1.3066 or higher

Did not happen

The forecast put >99.9% on the outcome that happened.

Score for this result

Brier score at ±1%: 0.497

Brier score, three outcomes. Lower scores mean better probability estimates; 0 is perfect for an individual result and 2 is the worst.

Down, Within, Up: (0.43 − 1)² + (0.27 − 0)² + (0.31 − 0)² = 0.497

Brier score at ±5%: 0.022

Brier score, three outcomes. Lower scores mean better probability estimates; 0 is perfect for an individual result and 2 is the worst.

Down, Within, Up: (0.07 − 0)² + (0.88 − 1)² + (0.05 − 0)² = 0.022

Brier score at ±10%: 0.000

Brier score, three outcomes. Lower scores mean better probability estimates; 0 is perfect for an individual result and 2 is the worst.

Down, Within, Up: (0.00 − 0)² + (0.99 − 1)² + (0.01 − 0)² = 0.000

Brier score at ±50%: 0.000

Brier score, three outcomes. Lower scores mean better probability estimates; 0 is perfect for an individual result and 2 is the worst.

Down, Within, Up: (0.00 − 0)² + (1.00 − 1)² + (0.00 − 0)² = 0.000

Each term is (forecast − actual)², where actual is 1 for the outcome that happened and 0 for the others. One result says little about skill. See the track record.

Outcome at every threshold

The realized return of −1.12% falls into a different category depending on the threshold.

Realized outcome and score for each threshold
Threshold	Outcome	Original Down / Within / Up	On outcome	Brier (0–2)
±1%	Down 1% or more 0.8624 or lower	42% / 27% / 31%	42%	0.497
±5%	Between −5% and +5% 0.8275 to 0.9146	7% / 88% / 5%	88%	0.022
±10%	Between −10% and +10% 0.7840 to 0.9582	0.5% / 98.7% / 0.8%	98.7%	0.000
±50%	Between −50% and +50% 0.4355 to 1.3066	0.0% / >99.9% / <0.1%	>99.9%	0.000

Values shown as <0.1% or >99.9% are rounded for display; the underlying values sum to 100%.

Scores by source

At ±1%, same scored version and the same outcome. Differences on a single result are not evidence that one source is better.

Brier score by source at ±1%
Source	Original Down / Within / Up	Brier (0–2)
AI average	42% / 27% / 31%	0.497
OpenAI	45% / 26% / 29%	0.452
Claude	40% / 28% / 32%	0.543
Statistical reference Synthetic fixture	33% / 33% / 34%	0.663
Prediction markets	No matching market	Not scored

At ±5%, same scored version and the same outcome. Differences on a single result are not evidence that one source is better.

Brier score by source at ±5%
Source	Original Down / Within / Up	Brier (0–2)
AI average	7% / 88% / 5%	0.022
OpenAI	8% / 87% / 5%	0.028
Claude	6% / 89% / 5%	0.017
Statistical reference Synthetic fixture	2% / 95% / 3%	0.003
Prediction markets	No matching market	Not scored

At ±10%, same scored version and the same outcome. Differences on a single result are not evidence that one source is better.

Brier score by source at ±10%
Source	Original Down / Within / Up	Brier (0–2)
AI average	0.5% / 98.7% / 0.8%	0.000
OpenAI	1% / 98% / 1%	0.000
Claude	0.3% / 98.9% / 0.8%	0.000
Statistical reference Synthetic fixture	<0.1% / 99.7% / 0.3%	0.000
Prediction markets	No matching market	Not scored

At ±50%, same scored version and the same outcome. Differences on a single result are not evidence that one source is better.

Brier score by source at ±50%
Source	Original Down / Within / Up	Brier (0–2)
AI average	0.0% / >99.9% / <0.1%	0.000
OpenAI	0% / 100% / 0%	0.000
Claude	0.0% / >99.9% / <0.1%	0.000
Statistical reference Synthetic fixture	0.0% / >99.9% / <0.1%	0.000
Prediction markets	No matching market	Not scored

Which version is scored

One version was published for this date. The evaluation rule picks the scored version per target in advance, so the score never depends on which version happened to look best. Evaluation rules

Published versions for this target date
Issued	Lead time	Reference	±1% Down / Within / Up	Evaluation
31 Jul 2026, 06:12 UTC	31 days	0.8711 30 Jul fixing	42% / 27% / 31%	Scored version

Each stored version keeps its own reference price and timestamps.

Later USD/EUR forecasts

This page stays at this address. Later dates have their own pages; you are not redirected.

30 Nov 2026 forecast 31 Dec 2026 forecast 31 Jan 2027 forecast

Before settlement

Between the target time and the arrival of the settlement price, this page shows a pending state instead of a result.

History

USD/EUR forecast history Forecast track record