One year of day-ahead forecasts for Germany, every day
Every quarter-hour of 366 delivery days (7 Nov 2025 – 7 Nov 2026), compared with the price that cleared. No hand-picked days.
The benchmark is the simplest plan anyone can make: take yesterday’s price at the same local time. A forecast that cannot beat it is worth nothing. Ours was closer on 300 of 366 days and on average 42 % closer over the whole year. On 66 days yesterday’s prices were closer — they are in the data below, not left out.
Error by month
Our forecastYesterday’s prices
| Month | Days | Our MAE (€/MWh) | Yesterday’s prices MAE (€/MWh) | Lower error | Days we were closer |
|---|---|---|---|---|---|
| Nov 2025 (7–30 Nov, partial) | 24 | 16.2 | 27.9 | 42 % | 23 of 24 |
| Dec 2025 | 31 | 10.4 | 20.5 | 50 % | 27 of 31 |
| Jan 2026 | 31 | 15.8 | 25.0 | 37 % | 22 of 31 |
| Feb 2026 | 28 | 10.0 | 17.4 | 43 % | 24 of 28 |
| Mar 2026 | 31 | 18.1 | 40.2 | 55 % | 29 of 31 |
| Apr 2026 | 30 | 18.9 | 34.3 | 45 % | 20 of 30 |
| May 2026 | 31 | 18.5 | 26.8 | 31 % | 26 of 31 |
| Jun 2026 | 30 | 23.3 | 37.9 | 39 % | 25 of 30 |
| Jul 2026 | 31 | 16.5 | 25.8 | 36 % | 26 of 31 |
| Aug 2026 | 31 | 17.4 | 26.6 | 35 % | 24 of 31 |
| Sep 2026 | 30 | 28.5 | 49.7 | 43 % | 24 of 30 |
| Oct 2026 | 31 | 23.5 | 47.1 | 50 % | 26 of 31 |
| Nov 2026 (1–7 Nov, partial) | 7 | 15.1 | 16.8 | 10 % | 4 of 7 |
Download every day as CSV — date, our MAE, MAE of yesterday’s prices.
How these numbers were made
- Same model as the live forecast. The model that produces the daily forecast was run day by day over the past year (walk-forward backtest).
- Only what was known before the auction. For each delivery day the model saw prices up to midnight before that day — they come from the previous auction — and weather forecasts issued before 12:00 CET on the day before delivery.
- 15-minute resolution. MAE is the average absolute difference between forecast and cleared price over all quarter-hours of a day (96; 92 or 100 on clock-change days), then averaged over days.
- Prices. Day-ahead prices of the SDAC auction for the Germany–Luxembourg bidding zone from the ENTSO-E Transparency Platform.
What this does not show: intraday or balancing prices, and how the error behaves in other zones. Methodology · Case study: a 10 MW battery in Germany
See the same comparison for your zone.
Request a Demo