246,754 open-ocean encounters in 90 days. 51% end in a Rule 17 deviation. The geometry is nearly identical to approach channels. The attention is not.
The Setup
The give-way obligation under COLREGS Rule 17 applies everywhere — not just at port approaches. When a give-way vessel fails to act, the stand-on vessel faces a choice: maintain course and speed, or deviate. If neither vessel moves, the encounter escalates to a Rule 17 deviation event.
The approach channel context gets most of the regulatory attention. Vessel traffic services watch it. Port state control inspectors know it. Maritime insurers model it. What's less watched is open water.
In the last 90 days — April 29 through July 14, 2026 — the open-sea context logged 246,754 give-way inaction encounters. The approach channel logged 50,066 over the same period. Open water is producing give-way inaction at 4.9× the rate of the approach context.
The Chain
51.1% of open-sea give-way inaction encounters — 126,201 of 246,754 — ended in a Rule 17 deviation. Among the deviating encounters:
- 73,390 had vessel A deviating only
- 52,811 had vessel B deviating only
- Zero had both vessels deviating simultaneously
That zero matters. In 246,754 encounters, both vessels never executed a maneuver at the same time. Every deviation was unilateral. The geometry when these encounters close: median minimum range of 0.250nm (450 meters), median DCPA of 0.114nm (211 meters). At the 5th percentile, encounters closed to 0.036nm — 67 meters.
The approach channel comparison is revealing. Average minimum DCPA in approach give-way inaction: 0.252nm — nearly identical to the open-sea figure of 0.250nm. Same geometry, 4.9× the volume, different scrutiny.
The Implication
The open-sea give-way inaction context produces a larger Rule 17 deviation queue than any other trigger-context combination in the current dataset. It is not a secondary signal. It is the dominant one.
What explains the mismatch? Approach encounters are spatially predictable — they happen at known waypoints, near known infrastructure, on routes that VTS systems watch. Open-sea encounters are diffuse. They distribute across millions of H3 cells. That spatial diffusion makes them structurally invisible to port-centric compliance frameworks even when their severity metrics are identical to the approach context.
The zero bilateral-deviation rate adds another dimension. These are not cases where both mariners sensed the risk and responded. They are cases where one vessel eventually moved while the other held. The resolution is always asymmetric.
What to Watch
- Whether the 51.1% deviation rate in open sea is trending up, flat, or down over the next 30 days. The current dataset covers a 90-day window; monthly resolution would reveal whether the compliance rate is deteriorating.
- Vessel class breakdown within open-sea give-way events. The dataset does not yet segment which vessel type is more often acting as stand-on versus give-way in unilateral deviations. If one class is systematically failing the give-way obligation, that's a vessel-type compliance gap.
- Whether the zero bilateral-deviation rate holds as sample size grows. It is theoretically possible both vessels deviate simultaneously and cancel each other in the classification logic — that would require ruling out.
Limitations
The 246,754 encounter count depends on AIS position coverage quality. Vessels with irregular ping rates, silent intervals, or degraded position accuracy may be under-represented or misclassified in encounter detection. The pairwise_encounter table captures encounters where both vessels have sufficient position history to compute CPA and DCPA — vessels with partial AIS records are excluded. The 5th-percentile figure of 0.036nm should be interpreted with caution: at that range, DCPA estimation is sensitive to AIS latency and clock skew. The 90-day window means short-duration trends are not resolved.
Data as of July 14, 2026. Source: Axiom Overwatch pairwise_encounter, April 29 – July 14, 2026. n=246,754 open-sea give-way inaction encounters.