Energy · Forecast users

Does the forecast band you act on cover as often as it says?

A trading, scheduling or balancing team reserves against the range a forecast states. If the range is too narrow, the team carries exposure it believes it has covered.

EvidenceAround the German day-ahead load forecast, a 90% band built the usual way covered 84.8% of quarter-hours [82.3%, 87.1%], over 32,252 quarter-hours on 336 scored days. It missed 19.2% of the most expensive 20% of intervals against 14.2% of the rest.

EvidencePriced at reBAP, that is EUR 90,042 of unreserved adverse exposure per 100 MW of mean load over the 336 scored days. An adaptive band on the same forecast covered 89.6% and carried EUR 44,495.

Evidence

Offer

Vendor Band Audit

Input
One CSV: timestamp, forecast, actual, and the stated band if there is one. 12 months or more.
You get
Coverage against the stated level by regime and by season, the misses priced at your market’s prices, a findings register and a 45-minute readout.
Timing
2 weeks from data delivery
Reference price
EUR 3,000, fixed

Full scope and terms

Forecast Calibration Audit

Input
One CSV per series, 12 months or more.
You get
The same audit across every forecast you rely on, with an adaptive recalibration demonstration on your own data.
Timing
2 weeks from data delivery
Reference price
Fixed, set by the number of series. About EUR 9,500 for 4 series.

Full scope and terms

If you own a battery that a third-party optimiser trades, the page for battery owners is the one written for you.

Try it

The Band Checker runs the first 2 sections of the audit in your browser, on your own CSV or on the bundled DE-LU sample, and your file never leaves your computer.

First step

Send one paragraph describing the series or the records to matthew@mtanti.com, and you will have a quote within one working day. A standard mutual NDA comes before any data moves, and your data is deleted on delivery and never used elsewhere.

What this does not support

One public series, German system load, over 336 scored days (2025-07 to 2026-07). It says nothing about your forecast, your vendor or another market until the same audit is run on them. The exposure figure assumes a desk reserves flexibility equal to the stated band and prices only the imbalance spread against day-ahead; a desk that reserves more carries less. It does not show that the point forecast can be improved, and it does not show that the adaptive band would hold through a future regime change.