Certisyn operates this instrument at arm's length from the parties whose conduct it measures and from the body that publishes the findings. Every reading is drawn from a named public endpoint, every detector's state is published beside the number it produces, and the whole dataset seals to a digest anyone can recompute without trusting Certisyn.
Each truce window runs from seven days before the Olympic opening to seven days after the Paralympic closing. Alongside the violation count the Truce Foundation records for each Games, this table carries the measured state of the world in that Games' annual reference year. The two are produced by different methods over different populations and are set side by side, not merged.
| Gameshost, season | Truce windowUNGA resolution | Windowdays | Recorded violationscount, Truce Foundation | States with no eventcount of 193, UCDP measured | Events in windowcount, UCDP measured | Forced displacementmillion persons, UNHCR global total | Military expenditure% of world GDP | Military expenditurebillion USD, current | Appeal requirementbillion USD | Panel states at or above 500kcount of 47 |
|---|
Two axes are published and never merged: what the instruments observed, and how much of the picture those instruments can see. A state that cannot be observed is not a compliant state: it is an unobserved one, and it is rendered in its own visual register that no reader can mistake for a clean sheet.
Conduct on one axis, the structural difficulty of complying on the other. A peaceful state that did nothing and a pressured state that held are not the same achievement, and a single league table calls the first better. Confidence is never folded into the position: the colour carries root span, meaning how many genuinely independent origins stand behind the reading.
A statistic without a reference is not reportable. Each measure below carries its unit, the observed distribution across the population that has a reading, and the reference value that defines good. Where an external reference exists it is named and dated. Where none exists, the population median is used and labelled as such.
Coverage is published per detector, never as one global claim. Two detectors sharing a causal root raise confidence but do not raise root span. Three reports drawn from the same reporting chain are one witness, not three.
The resolution is decomposed into discrete stipulations and each is scored against the treaty text rather than against the peer mean. Where a stipulation has no detector, the instrument records not_run and says why. Where it cannot apply to a state, it records a written determination of inapplicability from a closed vocabulary. Neither is ever allowed to read as compliance.
Every state, every reading, every gap, with the percentile each reading occupies in its own distribution. Sort any column. The five chips are the stipulation states: solid means measured, hollow means not run, dashed means a written determination of inapplicability.
| State | CessationParis 2024 window | Evidence status | S1 to S5 | Displacement% change, 2023 to 2024 | Stockthousand persons, 2024 | Reliabilityindex points, 0 to 100 | Rootscount | Difficultyindex points, 0 to 100 | Military spend% of GDP | Forcesper 1,000 population | Appeal funded% of requirement |
|---|
Every statement below is produced by arithmetic over the dataset above and by nothing else. It carries no judgement about any state's conduct, no attribution of cause, and no forecast. Re-running the same ingest produces the same sentences.
The whole dataset above is canonicalised and digested. Anyone can recompute this digest from the published payload and get the same answer, or a different one, which is the only kind of accountability claim that survives contact with a state that disputes it. Certisyn holds no ability to alter a past determination without producing a different root.
Recompute it yourself. Copy the canonical payload, delete the two fields the seal cannot cover (summary.root and summary.canon_bytes), serialise with sorted keys and no whitespace, then SHA-256 the bytes:
The readings are not entered by hand. A scheduled Certisyn ingest re-runs every endpoint on the rack, recomputes the index, re-derives the root and republishes to this same address. The link does not change; what it shows does.
Each refresh produces a new root and the previous root stays resolvable, so a statement made on any past day can be checked against the data as it stood on that day. That is the difference between an instrument and a page that quietly changed its mind.
A naming instrument is attacked on process before it is attacked on data. This ladder is published in advance, so the answer already exists when the first foreign ministry asks. Certisyn operates the ladder; the publishing body retains the decision.
The Truce Foundation's record already establishes the thing that matters most, which is that the Olympic Truce is a commitment with named parties and a published history. What an operated verification layer adds is the machinery that lets that record be relied on by someone who was not in the room when it was written, and defended by someone who has to answer for it afterwards.
Everything on this page is qualified by the eight statements below. They are held in one place deliberately, so that a reader can audit the assumptions as a set rather than meeting them piecemeal at the foot of each view.