MedGraph
A dependency cascade model of the Eastern Mediterranean, Red Sea and South Caucasus, built so that the difference between evidence and judgement stays visible in the output.
Executive summary
On a Suez Canal disruption the ranking is Egypt, Bab el-Mandeb, the United States, NATO, then Turkey and Greece level. It holds under all three damping functions and under three of the five stated assumptions.
It rests on two: NATO cohesion in this theatre, and Egypt's tolerance for a Suez revenue shock. The second is graded low confidence, and the evidence that would settle it is recorded as an intelligence gap rather than estimated.
The finding is not the ranking. It is that measuring three edges did not confirm the earlier conclusion. It promoted the weakest assumption in the register to load-bearing, and the evidence needed to test it was already logged as missing.
The tool
Sensitivity
Each assumption is pushed to its plausible bounds in both directions and the model re-run. The figure is how many places the baseline top five move. Zero means the conclusion does not depend on that assumption for that scenario.
| Assumption | Confidence | Suez | Red Sea | Straits | Cyprus | Gas |
|---|---|---|---|---|---|---|
| A001 France acts on the 2021 clause | Low | 0 | 0 | 0 | 2 | 0 |
| A002 NATO cohesion constrains escalation | Moderate | 3 | 0 | 0 | 0 | 0 |
| A003 Egyptian neutrality outweighs revenue | Low | 2 | 0 | 0 | 0 | 0 |
| A004 Ansar Allah aligned, not directed | Moderate | 0 | 0 | 0 | 0 | 0 |
| A005 Escalation needs armed escort | Moderate | 0 | 0 | 0 | 2 | 0 |
Only A004 is inert everywhere. On the Suez scenario alone, A001 and A005 both look irrelevant; across five scenarios they are not. The tool prints a warning to that effect rather than letting a single-scenario reading stand.
Method
Provenance
Every weight is labelled measured, from a citable dataset with a retrieval date, or assigned, an analyst judgement. Political friction and alliance strength cannot be measured, so those are declared rather than disguised, each referencing an assumption with a recorded falsifier: the evidence that would overturn it.
Currently four edges measured, twenty assigned, five unextracted and reported as such on every run.
Propagation
Impact along a path is the product of the edge weights multiplied by a distance damping term, kept as separate factors so the closed form can be stated. Three damping modes are selectable, and the Suez ranking is identical under all three.
Edges encode vulnerability flow: an edge from A to B means a shock at A transmits to B, so B depends on A. The database stores dependencies the other way, because that is the natural way to record ownership, and the loader reverses them.
What it does not do
It is not a forecast and cannot say what will happen. Five edges remain unmeasured because terminal throughput by source field is commercially sensitive; those are logged among eight recorded intelligence gaps rather than estimated. Reference figures such as defence budgets are unverified placeholders that feed no output and should not be cited.