Daily guide · BE · 2026-09-11
The number your Belgium price model has to beat this week
All numbers computed from the week's published auction results at publication and frozen — this article is an honest snapshot, not a page that rewrites itself. Reproduce via the API.
The number to beat this week is 38,18 EUR/MWh, the persistence MAE across the six days in the block. That is the naive benchmark: yesterday's shape carried forward. Any model worth running should sit comfortably below it, and our own recent record suggests the gap is real, with a 25,65 EUR/MWh MAE against a 30,40 baseline over the trailing sample.
What stands out is how uneven the week is. Persistence errors swing from 14,92 EUR/MWh on 8 September to 57,24 on 7 September, so the weekly average hides two very different regimes. The clean days sit alongside a rising price level, from a 89,43 mean on 5 September to 202,56 on 10 September. Days where the level jumps but persistence stays tolerable, like 8 and 9 September, are where a naive carry-forward quietly does most of its work for you.
The concrete takeaway: do not judge your model on the 38,18 weekly figure alone. Beating it on 7 and 10 September, where persistence MAE exceeds 50 EUR/MWh, is where the week is won or lost.
Before you train anything on BE prices, you need the number your model has to beat. In day-ahead forecasting that number is naive persistence: predict that tomorrow's curve equals today's. It costs nothing, needs no features, and it is embarrassingly hard to beat — because power prices are dominated by daily and weekly rhythms that persistence gets for free. We computed it on this exact week:
38.18 €/MWh
persistence MAE on this week's BE auctions — the bar. A model that can't get under this adds negative value.
25.65 vs 30.4
our live production model vs its persistence baseline, trailing 443 scored days — the same scoring, in public, on /accuracy
Why the bar moves: this week, day by day
| Target day | Day mean €/MWh | Persistence MAE |
|---|---|---|
| 2026-09-05 | 89.43 | 49.87 |
| 2026-09-06 | 115.03 | 32.95 |
| 2026-09-07 | 165.48 | 57.24 |
| 2026-09-08 | 173.79 | 14.92 |
| 2026-09-09 | 151.93 | 23.49 |
| 2026-09-10 | 202.56 | 50.62 |
Calm days make persistence look unbeatable; transition days (weather swings, weekend boundaries) are where a real model earns its keep. Averages hide this — always look at the daily distribution.
The part everyone gets wrong: the features
The model is the easy half. The silent killer is lookahead: training on weather reanalysis instead of the forecast that existed at issue time, on revised load data, on same-day flow averages that hadn't finished flowing at the auction. Backtests built that way look brilliant and die in production. Our Quant feature matrices serve one row per (issue day, 15-minute slot) where every column is documented with when it became knowable — the same rows our own models train on:
curl -H "Authorization: Bearer YOUR_KEY" \
"https://voltcast.com/api/v1/features/BE?issue_from=2026-07-01&horizons=1"
# columns include: price lags · previous-run weather (as forecast AT issue)
# · grid drivers (lagged to last COMPLETE day) · day-ahead load/RES forecasts
# · ensemble spread · target_price label (null until matured, never imputed)
Start with Home: the persistence bar above needs nothing but GET /v1/prices/BE
(Home covers one zone of your choice). When your walk-forward backtest beats it honestly, prove it in public on the
Forecast Bench — forward-blind, open scoring code, and our
own model is auto-entered under identical rules. If you'd rather buy the plumbing than build it,
contact us about ML-grade data access.
Method & citation. Prices are hourly means of published day-ahead auction results (native 15-minute periods averaged; ENTSO-E/SMARD, attributed). Wholesale-price component only — grid fees and taxes come on top and vary by supplier. Cite as "Voltcast Research, voltcast.com/guides/quant-features-be-2026-09-11".
Next issue 2 Oct
The European power roundup. Once a month. Real numbers.
Negative-price records, the month's widest spreads, the zones we failed to forecast — computed from the same production recorders as the API. No filler. Unsubscribe in one click.
- ✓ Negative hours and the cheapest zones
- ✓ The month's widest day-ahead spreads
- ✓ Forecastability ranking — where we missed
Get the next issue in your inbox.
No account. Double opt-in. Nothing else, ever. Privacy.