Rule 17 in the Wild: 72,113 Handoffs, 75% Extremis, and What That Says About Who Is Actually Avoiding Collisions

If you write marine insurance, train pilots, or set port traffic policy, the COLREGS handoff rate is your blast-radius indicator. Here is what 30 days of close-quarter encounters look like when you measure it.

The Setup

Across the last 30 days of vessel-tracking data, 125,559 close-quarter encounters were extracted globally. In 57.4% of them (72,113 encounters), the vessel that should have held course and speed under COLREGS Rule 17 — the stand-on vessel — eventually had to maneuver itself. Of those handoffs, 75.2% were "extremis" triggers, meaning the stand-on vessel maneuvered as a last resort because the give-way vessel had not. The remaining 24.8% were "giveway_inaction" — the give-way vessel never moved at all, and the stand-on vessel took action without waiting for a near-miss.

The headline number is the one that should land first: 3 of every 4 Rule 17(b) handoffs in our data were last-resort maneuvers, not preemptive ones.

The Chain

The pipeline is straightforward to trace. AIS positions stream in at ~92,000 rows/hour. An hourly extractor groups them into pairwise tracks where two vessels were within a closest-point-of-approach (CPA) threshold and a kinematic risk window. For each encounter we compute who was the stand-on vessel under Rule 15 (crossing) or Rule 13 (overtaking), watch the next several minutes of course/speed, and log a Rule 17(b) handoff if the stand-on vessel changed course by more than the deviation threshold. The trigger code records why: extremis if the change happened with the CPA already inside an extremis bound, or giveway_inaction if the give-way vessel had failed to alter course early enough.

The geography breakdown is where the signal sharpens. Open-sea encounters had the highest deviation rate (50.2%) and the highest handoff rate (61.5%), with vessels averaging 0.45 nautical miles minimum range and just 0.17 nm minimum DCPA. Approaches and anchorages followed at 51.5% and 53.0% deviation rates respectively, with slightly looser CPA distances. The average maximum course change across all flagged encounters was 53–66 degrees — these are not subtle adjustments. They are hard-over rudder events.

The Implication

The conventional reading of Rule 17 is that handoffs are rare exception cases. The data says they are modal. Insurers underwriting hull and P&I premium on the assumption that COLREGS handoffs cluster at extremis are pricing without the right base rate. Pilot organizations training on Rule 15 expecting that the give-way vessel almost always acts first have the prior backwards, at least in the open-sea cohort. And classification societies running near-miss audits on samples are likely understating the rate at which avoidance burdens default onto whichever vessel has the better watch.

The actionable thesis: encounter context determines handoff origin. At anchorage, give-way inaction is a higher fraction of triggers (the other vessel is often static or slow). At sea, extremis dominates because both vessels are moving and the give-way vessel sometimes maneuvers, just too late. Reading those distributions per-waterway gives you a better risk model than a flat global rate.

What to Watch

• Rule 17 deviation rate per waterway over the next 30 days. A rising rate in a specific corridor is a leading indicator for hull-loss frequency, especially in chokepoints with VTS coverage gaps.
• Extremis-vs-inaction ratio by flag state of the give-way vessel. A flag with disproportionate giveway_inaction triggers warrants port-state-control attention.
• Average max course change distribution. If hard-over maneuvers (>90°) become a larger share of the deviation count, encounters are getting closer to incidents, not further from them.

Limitations

This rests on AIS-derived pairwise encounters. Vessels that go AIS-dark in the middle of an encounter drop out of the dataset for the duration of the gap, which biases the count downward in the busiest corridors (Singapore Strait, Bab-el-Mandeb). The Rule 17(b) handoff classifier triggers on observed kinematic deviation, so vessels that maneuver below the threshold are not counted. Encounter context (open_sea / approach / anchorage) is heuristic-derived from port-zone proximity and depth; mis-classifications are possible at the boundaries. Finally, the 30-day window includes encounters from one quarter of the global fleet, not all of it — coverage skews toward terrestrial AIS-receiver regions and away from satellite-only zones.

Data current as of 2026-05-01. Sources: pairwise_encounter (AXO-117 pipeline) over the last 30 days; AIS positions ingested via AISStream.io and AISHub. Methodology: COLREGS Rules 13–17 applied to extracted encounter geometries, Rule 17(b) handoff classifier per AXO-134.