The first history we measured is our own.
Two years of trading, 1,084 rows, 33 assets. The full report is below, as a PDF, no account — with its limits and the three measures it could not produce. Nothing is removed from it, nothing is rounded.
The Mirror runs in your browser: nothing is sent. How to read a report →
A statement shows a balance and 582 orders. The other 502 rows are conversions from one crypto to another: each one closes a position and opens another. A calculation that sets them aside is not about what you did.
What the measure establishes
The four main measures, from 08/11/2023 to 18/11/2025, valued as of 18/11/2025. Each figure carries its convention, published on the method page, and reads on an example you can redo by hand. To the cent, as the report writes them.
Two years of back-and-forth produced €106.30 less than the basket left untouched. Bitcoin reference, for guidance only: +€16,671.36 — the same money, locked on a single asset, with hindsight. The report calls it a control, not a comparison. Concentration at 164.3%: the three best trades return more than the total, everything else, together, is at a loss.
What it does not establish
Three of the eight measures are not produced. The report says why; these are its words, translated from the French report:
- 06 · Maximum drawdown and 07 · Abstention counterfactual — “your history carries deposits and withdrawals, and the convention that removes them from the value curve is not yet published. Without it, a withdrawal would be counted as a drop: on our own history, the measure announced a 99.8% loss when the money had simply been withdrawn. We would rather say nothing.”
- 08 · Reactions within 48 hours after a drop — “the threshold and window that define a drop are not yet published on /en/schema; they will be measured on real histories before they are written there.”
And two scope limits, named: 63 deposits or withdrawals with no counterpart, 13 rows with no price at the date of the trade. The full list is on our limits; we do not write it case by case.
What this measurement cost our method
A rule wrong on 47% of conversions, corrected in public.
We assumed a conversion matched on an identical timestamp. Across 251 conversions, that rule matched only 192 — with no error, no warning, and a normal-looking figure at the end. We established the real key on the actual file, corrected the rule on the page that publishes it, then the code. In that order, on 18 September 2026.
Does your history say the same thing?
The method is published. So are the conventions. If a figure in this report looks wrong to you, write to us: we will publish the challenge and the answer, even if it proves us wrong.
The Mirror tells you what your file allows to be measured. Nothing is sent.