The Setup
Over the last 30 days, the approach corridor feeding Los Angeles/Long Beach logged 1,386 pairwise vessel encounters tagged with COLREGS context approach. Of those, 63.2% show at least one vessel deviating from Rule 17 stand-on obligations during the encounter window. Across every approach-context encounter tracked globally in the same 30 days — 325,747 of them — the deviation rate is 48.4%. Los Angeles/Long Beach runs 14.8 percentage points above that baseline, and it isn't a confidence artifact: encounter confidence at LA/Long Beach averages 0.347, essentially the same as the global approach-zone average of 0.337.
The Chain
Rule 17 of the International Regulations for Preventing Collisions at Sea requires the stand-on vessel in a crossing or overtaking encounter to hold course and speed until it becomes apparent the give-way vessel isn't taking appropriate action — and even then, maneuver only as a last resort. The deviation flag on each encounter fires when either vessel's course change during that stand-on window (max_course_change_a_deg / max_course_change_b_deg) crosses the detection threshold. A 63.2% rate means that in roughly two out of three approach encounters funneling traffic into San Pedro Bay this month, a vessel that COLREGS expects to hold its line instead altered course.
San Pedro Bay is the busiest container gateway in North America by throughput, and the approach geofence sits directly in the merge zone where inbound traffic queues toward the twin harbor's terminal complex. Compared to the other named approach zones pulled in this same window — Antwerp at 65.0%, Rotterdam at 59.0%, Baltimore at 58.4%, Norfolk at 53.8%, Singapore at 49.9%, New Orleans at 43.6% — LA/Long Beach's 63.2% is the second-highest of any geofence carrying meaningful volume, and the highest among U.S. ports in the pull.
The Implication
The directional claim: Los Angeles/Long Beach's approach corridor is running a materially elevated Rule 17 deviation rate relative to the global approach-context baseline this month, at comparable detection confidence. That doesn't establish that vessels here are behaving less safely than the aggregate — congestion, tug-assist maneuvering, and precautionary-area convention in a two-harbor merge zone can all produce legitimate course changes that this detector counts as deviation. What it does establish is that this specific corridor is a statistical outlier worth a closer look, not background noise.
What to Watch
Whether the 63.2% holds as more of the current 30-day window rolls off and gets replaced by fresh encounters, or whether it reverts toward the approach-zone baseline. Whether encounter confidence climbs meaningfully above 0.35 as AIS density in the corridor increases — right now the detection regime here is still low-confidence in absolute terms, even if it's on par with the global average. And whether the deviation concentrates in a specific sub-window of the approach geofence (closer to the breakwater vs. further out) once encounter-level geometry is broken down further.
Limitations
An encounter confidence of 0.347 is low in absolute terms — this is not a high-confidence detection regime, it's merely typical for approach-context encounters generally. The deviation flag is a course-change threshold, not a visually confirmed maneuver; it will register legitimate defensive seamanship, tug-assist positioning, and traffic-separation-scheme convention alongside genuine stand-on violations, and this data can't yet distinguish between them. The 1,386-encounter sample is real but modest next to Rotterdam's 72,921 in the same window — enough to flag an outlier, not enough to rule out a monthly fluctuation. No causal claim is made here about why the rate is elevated at this specific port; that requires encounter-level review this data doesn't yet support.
Data as of 2026-09-29. Source: Axiom Overwatch pairwise_encounter, context_tag='approach', Los Angeles/Long Beach approach geofence, trailing 30 days, compared against the global approach-context baseline over the same window.