Method
This site publishes empirical work on electricity markets: what the current rules actually produce, measured rather than asserted. The method matters more than any individual result, so it is stated here in full and applied without exception.
Every figure carries an evidence class
Numbers differ in what stands behind them, and blurring that distinction is how analysis becomes advocacy. Each figure is labelled:
| Class | Meaning | Depends on a model? |
|---|---|---|
| [M] | Measured — computed directly from a public data series | No |
| [C] | Computed — arithmetic on measured quantities, under stated assumptions | No |
| [P] | Public accounts — a published figure, quoted and cited | No |
| [S] | Simulated — output of a dispatch model | Yes |
The arguments here rest on [M], [C] and [P]. Where an [S] exhibit appears, it is there because it shows a mechanism reproducing — never because the argument needs it. Delete it and the conclusion should be unchanged. If it isn't, the essay says so.
Bands, not point estimates
Where a quantity depends on assumptions that are not directly observable, the assumptions are given lo/central/hi values before the calculation runs, and the result is reported as the band those corners produce. A central figure is the central corner, not a best fit. Known biases are stated in the direction they push, and not silently corrected for.
Predictions are registered before results
A model that is adjusted until it agrees with the answer cannot be surprised, and a analysis that cannot be surprised has not tested anything. So predictions — including the specific numbers expected and the thresholds that would count as failure — are written down and committed before the calculation is run. Contradicted predictions are reported as findings, not quietly absorbed.
Corrections are made in public
Both essays on this site carry retractions. Figures that were quoted and later found to rest on a bad comparison have been withdrawn in place, with the reasoning shown and the corrected number given. This is deliberate and it is the point: work that never visibly revises itself is not being checked, including by its author.
Check the numbers yourself
Every figure quoted in an essay is recomputed, in front of you, by the reproductions repository. Clone it and run one script: it reads the hourly public data that ships inside it, redoes the arithmetic, and prints pass or fail for every figure the essay states. It needs no API key, no registration and no dependencies — the Python standard library, about ten seconds. The stored results are outputs of those scripts, never inputs to them, so nothing grades itself against its own answer. Change a number in the data and the run fails.
That leaves one step to take on trust — that the shipped hourly files really are the public
series they claim to be. A free ENTSO-E token closes it: --refetch re-pulls the
series from the source and diffs them against what ships here.
This is a standing invitation. If a number here is wrong, I would rather hear it from you than not hear it.
Who
Paul Schalm. Background in philosophy and economics. The simulated exhibits come from a European electricity market model I build and maintain — a dispatch engine over a real plant fleet, validated against published prices and carbon intensities. That model is not public; the evidence drawn from it is labelled [S] and is never load-bearing.
Corrections, disagreements and requests: hello@ordo-markets.com.