The Setup
The Houston Ship Channel approach zone logged 13,000 pairwise vessel encounters in the last 90 days — every one of them a give-way interaction close enough to the port to matter for anything built along the channel. 35.4% of those encounters show a Rule 17 deviation, and the average encounter confidence sits at 0.335, with 47.4% of encounters below the 0.3 confidence floor. That confidence profile is unremarkable by port standards: Rotterdam runs 0.337, Antwerp 0.379, Singapore 0.411. It only looks unusual next to one outlier — Persian Gulf Interior, where the same approach-zone methodology produces 0.778 average confidence and a 13.8% deviation rate, less than half of Houston's.
Houston also carries 130 scored H3 cells in the composite cell-scoring system. Those cells average a Safety & Environment component of 73.7 and an overall composite of 60.4, with 90% data coverage. The scoring-level confidence on those same cells averages 0.180 — lower than the confidence on the raw vessel-encounter data feeding the channel next to them.
The Chain
Composite cell scores near Houston's channel are built from eight component groups: demographics, accessibility, population momentum, development pipeline, economic strength, safety & environment, amenity demand, business vitality. None of them ingests a vessel encounter, a Rule 17 deviation, or an approach-zone confidence figure. The safety score attached to waterfront and near-channel cells is a land-use safety read — crime, environmental hazard, code enforcement — with no line item for the 13,000 close-quarters ship movements per quarter running past the same parcels.
That absence would be a minor gap if the underlying navigational signal were uniform across ports and could be assumed away. It isn't. The same table that produces Houston's 35.4% deviation rate at 0.335 confidence produces Persian Gulf Interior's 13.8% deviation rate at 0.778 confidence — a nearly threefold difference in deviation and better than double the confidence, on the same methodology, in the same 90-day window. A navigational-risk overlay built on approach-zone data cannot be a single national or global constant; it would have to be scoped port by port, and right now no cell-scoring input is scoped to it at all, constant or otherwise.
The Implication
Port-adjacent CRE risk in Houston is currently priced on land signals with a confidence score (0.180) three times lower than the shipping-channel data sitting a few hundred meters away (0.335, and climbing to 0.778 at other ports where the corridor runs cleaner). The channel-side safety read of 73.7 has no mechanism to move if Houston's approach-zone deviation rate spikes, and no mechanism to reflect that it's already running higher than a comparable bulk corridor elsewhere in the world.
What to Watch
Whether Houston's 35.4% approach deviation rate moves against its own 90-day baseline in the next quarter, and whether that movement shows up anywhere in the 130 scored cells' safety or development-pipeline components — the honest expectation, given the current architecture, is that it will not, because there's no column for it to move.
Limitations
Confidence in this context measures two different things across the two datasets — encounter confidence reflects AIS track quality, cell-score confidence reflects composite-model certainty across all eight component groups — so the 0.180 vs. 0.335 comparison is directional, not a like-for-like ratio. The 130 Houston cells span the full metro, not only channel-adjacent parcels; a tighter geofence around the ship channel itself would likely show a different — probably lower — safety-confidence pairing than the metro average used here. This is a structural gap, not a claim that the current Safety & Environment score is wrong on its own terms.
--- Data as of 2026-09-08. Source: `pairwise_encounter`, `port_zones`, `ports` (Axiom Overwatch); `cell_scores` (Axiom Locus), metro_slug='houston'.