Anchorages Have the Highest Rule 17 Deviation Rate of Any Encounter Context — 52.8%, Ahead of Open Sea's 48.2%.

The Setup

Over the trailing 90 days, the pairwise-encounter graph logged 3,233,317 vessel encounters across three context tags: open sea (1,558,213), approach channels (1,101,495), and anchorages (573,609). Each encounter is checked for a Rule 17 deviation — whether the give-way vessel failed to give way, forcing the stand-on vessel into an evasive course change. Anchorage encounters deviate at 52.8% (302,931 of 573,609). Open-sea encounters deviate at 48.2% (751,690 of 1,558,213). Approach-channel encounters deviate at 46.5% (512,503 of 1,101,495). Anchorage is the smallest of the three populations by volume and the highest by deviation rate.

The Chain

The obvious read on a give-way failure is pressure: tight quarters, high traffic density, little room to maneuver, so the give-way vessel either can't or doesn't act in time. That read doesn't survive contact with the geometry here. Anchorage encounters carry the largest average minimum range of the three contexts — 0.605 nautical miles, versus 0.544nm in approach channels and 0.484nm in open sea — and the largest average minimum DCPA, 0.222nm versus 0.208nm and 0.181nm respectively. Anchorage encounters have more room than open-sea encounters, not less, and still deviate more often. Whatever is driving the anchorage rate, it isn't spatial pressure forcing bad calls under constraint — there's more slack in these encounters than in the other two contexts, and the give-way vessel still fails to give way at the highest rate of the three.

Open sea, by contrast, produces the tightest average minimum DCPA (0.181nm) of the three contexts and still deviates less often than anchorage (48.2% vs 52.8%). That's the inversion worth sitting with: the context with the least room to work with isn't the context with the worst compliance rate.

The Implication

Encounter-risk models that treat anchorage-adjacent traffic as inherently lower-complexity — vessels slow, often near-stationary, presumed easier to predict — are working from an assumption the deviation data doesn't support. A plausible explanation is congestion pattern rather than physics: anchorages concentrate transiting vessels weaving among anchored ones, a mix that open-sea traffic, moving in roughly the same direction at roughly the same speed, doesn't reproduce. Another plausible explanation is operator behavior — complacency or fatigue near the end of a transit, just before anchoring, when a crew may be mentally shifting off open-water vigilance. Both are hypotheses, not conclusions; nothing in this dataset distinguishes between them.

What to Watch

Whether the 52.8% anchorage rate holds as a stable baseline over the next quarter or moves with traffic season. Whether it concentrates in a handful of high-traffic anchorage clusters — prior coverage here has flagged Gulf anchorage rings as disproportionate contributors to dark-event volume, and the same geographic concentration may be driving this deviation rate rather than anchorages behaving uniformly worldwide.

Limitations

rule17_deviation is derived from course-change thresholds (max_course_change_a_deg, max_course_change_b_deg) against modeled give-way/stand-on roles, not from a human review of each encounter — a threshold-based flag can misclassify ordinary maneuvering as deviation, particularly in anchorage traffic where course changes for berthing or swinging on anchor are routine and not COLREGS-relevant. context_tag itself is a model assignment, not a certified zone boundary, so some misclassification between anchorage and approach-channel encounters near port limits is possible. This aggregates globally across vessel classes, flag states, and individual anchorages with no breakdown by any of the three — the 52.8% figure could be driven by a small number of high-volume anchorages rather than reflecting anchorage behavior broadly.

Data as of 2026-08-12, trailing 90-day window (2026-05-14 to 2026-08-12). Source: pairwise_encounter (context_tag, rule17_deviation_a/b, min_dcpa_nm, min_range_nm), Axiom Overwatch encounter graph.