2027 prediction · Frozen September 2026
See you in
August 2027
We’ve frozen our guesses. Next summer, we’ll find out how wrong they were.
These are speculative guesses, not official 2027 boundaries. Future papers, student performance and grading decisions are unknown. Use the history for context, and leave yourself a margin.
Pearson Edexcel · 1GA0 · GCSE
Geography A · Grade 9
No headline guess.Three comparable June series from 2024 to 2026 are needed.
The numbers we can actually check
Bars share a 0–100% scale. Historical rows link to official sources.
Open the Geography A calculator →Retrospective test · 2026 held out of method selection
The boring guess won
Our selected method: Last year again.
We tested four simple methods on 2025, chose the smallest average error, then checked that choice against 2026. The fancier guesses didn’t earn the headline.
Show methods, errors and the important limitations
| Method | 2025 selection | 2026 evaluation |
|---|---|---|
| Last year again · selected | 1.92 pp | 1.57 pp |
| Recent median | 2.12 pp | 2.24 pp |
| Recent weighted mean | 2.05 pp | 1.96 pp |
| Half-strength recent trend | 2.22 pp | 1.84 pp |
How each method works. “Last year again” carries forward the previous boundary. The median uses up to three recent annual values. The weighted mean gives those years weights 1, 2 and 3, from oldest to newest. The half-strength trend adds half the annual change from the first to the last available observation to the last value. Each result is rounded to a whole mark, limited to the qualification maximum and checked for grade order.
A hindsight exercise. We ran these tests in September 2026 using the current data snapshot. The algorithms use only years preceding each target, but the research design was developed after seeing those years. This is not an independently preregistered trial. All four methods use matching eligible cases within each evaluation year. No method was selected on its 2026 score.
The 2026 evaluation has 558 grade boundaries from 77 qualification-entry options. Grades within one entry are related; these are not 558 independent experiments. The coverage also reflects this site’s available data.
The historical-error band. The band uses the 90th percentile of absolute 2026 testing errors in the same qualification group, converted to marks and rounded up. It is a description of those past errors, not a calibrated 90% forecast interval. Future changes can be larger.
Headline 2027 guesses require matching qualification code, entry option and maximum in 2024, 2025 and 2026. Announced changes flagged in our data suppress a guess. We have not independently certified that every syllabus or assessment circumstance is unchanged. The forecast must never be treated as a safe target.
2020 and 2021 had no normal summer exam series. 2022 transitional grading is excluded. 2023 is used only where needed for the earlier retrospective tests, with its post-pandemic grading context retained as a limitation.
Version 2027-v1. The original guesses and input snapshot stay in a separate file. Future versions must keep this one available.
Verify the input fingerprint
334917833ba9add22d59179efd9b99d773742d6f3e6407277f58b83daff3ddae
This SHA-256 fingerprint identifies the input snapshot. It is not an independent timestamp or a guarantee about future predictions.