How forecasts are made
A forecast in the radar is not an opinion but a verifiable statement with a probability and a deadline – sourced down to the reference and measured against reality afterwards.
Analyseteam vor einer Weltkarte (symbolic image, AI-generated)
It starts with a verifiable question with a deadline. The editorial team then gathers checked statements from situation, global situation, legislation and key figures, starts from the base rate of comparable cases and adjusts it to the specifics. A reviewer checks and releases it; after the deadline the forecast is evaluated.
As of 21 October 2026 · SAFETEE editorial team
Anatomy of a forecast
Every forecast in the radar has the same structure. This makes it comparable, verifiable – and scoreable after the deadline.
Step 1 · The right question
A good forecast answers a question that really matters to operators and companies – and that can later be answered clearly with “happened” or “did not happen”. Questions arise from new developments in the situation and from customer enquiries.
Step 2 · Gathering the basis
For each question, the editorial team compiles what is sourced: checked statements from the knowledge base, situation entries, the effect of the global situation on Germany, the state of legislation, key figures from energy and grid data and historical comparisons.
Step 3 · Base rate
We never start from gut feeling but from the base rate: how often has a comparable event occurred in comparable cases? Only then do we adjust – up or down, depending on the specifics of the case. Every adjustment is justified.
Step 4 · Probability
At the end there is a percentage. It is not a certainty but a statement about frequency: at 65 % we would expect about six to seven out of ten comparable cases to happen. If the situation changes, the number changes – visible as a trend since the last issue.
65 % likely
Step 5 · Review and release
Before a forecast appears, a named reviewer checks question, basis, base rate and number. He or she can switch sources on and off and immediately sees how this shifts the probability. Only with the release does the forecast become part of the issue.
Step 6 · Evaluation and calibration
After the deadline every forecast is scored. What counts is the direction: above 50 % it should happen, below 50 % it should not. Results are published in a table – including the misses. The radar learns from all evaluated forecasts: if we were wrong too often at 70 %, later forecasts become more cautious.
Questions and answers
That we consider the event rather likely: out of ten comparable cases we would expect about six to seven to happen. A 65 % forecast can therefore also not happen without having been poorly reasoned. That is why we evaluate across many forecasts.
The frequency with which a comparable event occurred in comparable cases – for example how many grid connections in a grid area were completed on time. It is the starting value of every forecast; specifics of the case shift it up or down with reasons.
What counts is the direction: a forecast above 50 % was correct if the event occurred by the deadline; one below 50 % was correct if it did not occur. We do not assign exactly 50 %.
Yes, if the situation changes. Each new issue shows the current value and the trend since the last issue in percentage points. The value valid at the time of publication is the one that is scored.
Yes. With the packages for companies, SAFETEE sets up research for your sector, project type and region and creates forecasts on the questions that affect your project – with reports and a fixed contact. Enquiries to radar@safetee.eu.
Next
Questions about the radar: radar@safetee.eu
