Performance study · baseline edition · as of June 2026

How our calls have held up

A rating is only worth what its record proves. The credit-rating agencies earn trust by publishing, every year, whether their ratings actually predicted defaults, and by admitting where they were wrong. This is our version of that study, for power-market ratings. It is a young record, so this baseline edition does two honest things: it scores every dated call we have made so far, and it states plainly what each future edition will measure, so you can hold us to it. We publish it quarterly, we never revise a past edition, and from now on we measure our calls against the point-in-time archive we began keeping in July 2026.

The scoreboard so far

0
Validated
5
Holding
7
Pending
0
Reversed

We have logged 12 material, landmark calls and 33 dated review actions beyond the initial ratings (23 affirmed, 9 under review, 1 downgrade). A call is Holding when it stands and nothing since has changed it, Pending when it is an open call awaiting an event that will resolve it, Validated when a later event confirmed it, and Reversed when a later event moved the rating the other way. Every one is dated on the track record. We publish reversals in full when they happen; there have been 0 so far, on a deliberately short record.

What each future edition will measure

These are the falsifiable tests we commit to, and will report against as the record lengthens:

What we cannot show yet, honestly

A genuine performance study needs realized outcomes measured over time, and this record is only months old. So this edition does not yet claim that our high tiers delivered cheaper or more reliable power over a multi-year window, because that window has not elapsed. What we can say is that the method is fixed and published, every call is dated, and the point-in-time archive we started in July 2026 now captures each day's full ratings, so future editions can compare exactly what we said on a past date against what actually happened. The value of this study compounds with time, and the clock has started.

How this study is governed

Published quarterly. Append-only: a past edition is never revised, only followed by a new dated one. Built from our own dated track record and the point-in-time ratings archive, not from numbers assembled after the fact. Next edition: 1 October 2026.