Accountability

Forecast track record

How Warconomy's published forecasts have performed. Honest from launch: no live forecast has resolved yet, so no accuracy rate is claimed. Includes strict-cutoff hindcast demonstrations scored against naive baselines.

No live forecast has resolved yetWarconomy issued its first forecast portfolio on 18 July 2026. None of those forecasts has reached its resolution date, so Warconomy has no track record yet and claims no accuracy rate. Any site that shows you a precise hit-rate on a brand-new forecasting product is showing you something it cannot have earned. The hindcast demonstrations below show how scoring will work once live forecasts start resolving — they are not evidence of predictive skill.

Warconomy has issued 8 live forecasts, of which 7 are active and 0 have resolved. Because the first portfolio was issued in July 2026 and the earliest resolution date falls after that, no live forecast has yet been scored. Rather than publish an invented accuracy figure, this page reports the raw lifecycle counts, explains exactly how scoring will work, and publishes 2 hindcast demonstrations — historical forecasts reconstructed under strict data cutoffs — each compared against a naive baseline. Hindcasts are excluded from live performance because they were built knowing the outcome.

  • No invented accuracy percentage is published anywhere on this site.
  • Hindcasts are labeled as demonstrations and never counted as live predictive skill.
  • Every model must beat a simple naive baseline to be worth running.
  • Forecasts that cannot be measured are marked unresolvable and excluded from scoring, not quietly dropped.

Lifecycle counts

Every live forecast Warconomy has issued, by lifecycle status. Hindcasts are counted separately.
StatusCount
Issued (live forecasts)8
Active7
Nearing resolution0
Resolved0
Invalidated0
Unresolvable0
Superseded1
Expired0
Withdrawn0
Hindcast demonstrations (not live)2

How forecasts will be scored

Each forecast names, at issue time, the exact criteria and published source that decide it. When that source publishes, the outcome is attached to the original record without altering the prediction. Scoring uses:

  • Directional accuracy — did the predicted direction happen?
  • Interval coverage — for range forecasts, did the outcome fall inside the published range?
  • Brier score — for probability forecasts, squared error against the outcome (0 is perfect, 0.25 is a coin flip).
  • Baseline comparison — every forecast is measured against a naive baseline (no-change, no-revision, or persistence). Beating the baseline is the minimum bar for a model to be worth running.

Where an outcome cannot be measured, the forecast is marked unresolvable with a stated reason and excluded from scoring rather than judged against a substitute measure. Full rules: the forecasting policy.

Hindcast demonstrations

These are not live predictions. Each was constructed after the fact using only information available at its stated data cutoff, to show how scoring works. Building a forecast when you already know the answer is easy — that is precisely why these are excluded from the counts above.

Hindcast: would a naive 'no revision' rule have predicted the EIA's July 2026 Brent cut?

Cutoff · resolved · 18-day horizon

Forecast saidvery likely (roughly 80-95%)
What happenedThe July 2026 STEO (released 7 July 2026) cut the 2026 annual Brent average to $81.91/b from $95.39/b — a downward revision of about $13.48/b, explicitly attributed by the EIA to increased oil flows through the reopened strait.
Naive baselineThe July STEO repeats the June figure of $95.39/b (no revision).
ResultForecast correct; baseline wrong. The model beat its baseline. Brier score 0.0225.

Hindcast: would trend continuation have beaten 'no change' on Sudan's IDP total?

Cutoff · resolved · 425-day horizon

Forecast said8,500,000–9,500,000
What happenedDTM's 29 March 2026 round recorded an estimated 8,936,175 IDPs — a 23% decline from the peak, inside the forecast's 8.5-9.5 million plausible range. The no-change baseline was wrong by roughly 2.65 million.
Naive baselineThe March 2026 total remains at the January 2025 peak of 11,585,384.
ResultForecast correct; baseline wrong. The model beat its baseline. The published range covered the outcome.

2 of 2 hindcasts beat their naive baseline. With a sample this small, that is a demonstration of method, not evidence of skill.

Forecasts awaiting resolution

Active forecasts and when each becomes measurable.
ForecastPredictionIssuedResolves by
EU adopts a 21st Russia sanctions package by 30 September 2026likely (roughly 60-80%)July 18, 2026September 30, 2026
Suez transit volumes remain far below the pre-2024 baseline through 2026very likely (roughly 80-95%)July 18, 2026January 31, 2027
EIA revises its 4Q2026 Brent forecast upward in the August 2026 STEOlikely (roughly 60-80%)July 18, 2026August 31, 2026
Sudan IDP count stays above 8 million in IOM DTM's next published roundvery likely (roughly 80-95%)July 18, 2026March 31, 2027
OHCHR reports at least one month in H2 2026 above the H1 2026 monthly averagevery likely (roughly 80-95%)July 18, 2026February 15, 2027
The next Ukraine RDNA raises the ten-year reconstruction estimate above $588 billionvery likely (roughly 80-95%)July 18, 2026December 31, 2027
Strait of Hormuz daily transits stay below pre-crisis levels through October 2026likely (roughly 60-80%)July 18, 2026November 30, 2026

Machine-readable

All forecasts · Active only · Resolved only · Model registry

Explore next