COLREG collision-risk assessment: a framework for learning

1. Establish which rules and which context

COLREG entered into force on 15 July 1977. IMO’s convention overview maps its main subjects and annexes; use the full rules for actual requirements.

The USCG compilation distinguishes international and US Inland text. Do not generalize Inland-only provisions. Check local special rules separately.

Before working a training scenario, state visibility, whether the vessels are visually in sight, waterway context, vessel status and the observation basis. If those facts are missing, the exercise may not support the neat encounter label suggested by a sketch. A disciplined learner records the missing facts rather than supplying them unconsciously.

2. Look-out is an information function

Rule 5 addresses continuous, proper look-out using sight, hearing and all other available means appropriate to the circumstances. Rule 6 links safe speed to effective avoiding action and stopping within a suitable distance. Its considerations include visibility, traffic, manoeuvrability, environment, hazards and radar limitations. See the international rules in Canada's official Collision Regulations, distinguishing the expressly labelled Canadian modifications.

For study, separate detection from understanding. “A target appears” answers whether something was detected. “Its motion is understood well enough” is a different claim. An object may be visible while its aspect, movement or relationship to other traffic remains uncertain. A display can therefore contain many symbols without supporting a complete appraisal.

A classroom observation record can distinguish what was seen, what an instrument reported, what was inferred, and what remains unknown. This is an authored learning device, not an additional rule. It helps expose a common reasoning error: an inference later being remembered as a direct observation.

3. Safe speed concerns the available response margin

A fixed number is a poor teaching substitute for the factors in Rule 6. For an original thought experiment, compare the same fictional speed in two environments: a clear, uncongested training scene and a scene with reduced visibility, uncertain targets and less manoeuvring room. The number has not changed, but the reasoning supporting it has.

Consider the information-to-action chain. Time can be consumed by detection, interpretation, discussion, execution and the vessel's physical response. In an educational analysis, showing those stages makes it easier to see why an apparently generous distance may offer less useful time than expected. No universal allowance is assigned to any stage here.

Do not turn this explanation into a stopping-distance formula or a recommended speed. Actual response depends on the vessel and conditions, and the applicable rule remains controlling. The learning objective is to understand why speed must be considered together with the ability to observe, decide and act, rather than as an isolated setting.

4. Risk assessment uses evidence, including uncertainty

Under Rule 7, risk must be assumed when doubt remains or an approaching vessel’s compass bearing does not appreciably change; bearing change does not always remove risk. Use all appropriate available means and avoid scant-information assumptions. Read international Rule 7(a)–(d).

This is not the same as proving that contact is inevitable. The question concerns a situation requiring attention before certainty exists. An analyst who waits for a perfectly resolved track may be answering the wrong question too late. Conversely, a single noisy sample does not support a confident explanation of another vessel's motion.

In a training debrief, ask which observation changed the assessment and why. Was it additional range information, a corrected reference, a newly detected object, or a changed environmental assumption? This makes uncertainty discussable. Simply labelling the final outcome “obvious” hides the information that was actually available earlier.

5. A small numerical example, with explicit limits

Suppose an invented target is recorded at 3.0 nautical miles and then 2.5 nautical miles two minutes later. The average decrease in range over that interval is 0.5 nautical mile in two minutes, equivalent to 15 knots of average range closure. The calculation is 0.5 × 60 ÷ 2 = 15.

This number is not the target's speed through the water, own ship's speed, or a recommended operating speed. It is a rate of change of separation in this fabricated observation pair. Two ranges alone do not determine a complete two-dimensional relative-motion track.

Now add an approximately unchanged measured compass bearing. The combined observations deserve a different assessment from the ranges alone, but measurement uncertainty, target extent and changing motion still matter. The calculation has not supplied vessel responsibilities, a safe passing criterion or a manoeuvre.

A useful classroom extension is to ask what cannot be concluded. Can the target's intentions be known? Can the available sea room be inferred? Can other vessels be ignored? The answer is no. A correct arithmetic result can coexist with an incomplete understanding of the encounter.

6. CPA and TCPA describe a model, not permission

Closest point of approach, CPA, is a predicted minimum separation within a chosen motion model. Time to CPA, TCPA, describes when that minimum occurs relative to the calculation time. Under a constant-relative-velocity model, a change in either vessel's motion changes the prediction. These are mathematical interpretations, not statutory definitions of safety.

The UK's MGN 379 on electronic navigational aids discusses limitations of electronic aids and cautions against relying on AIS alone for collision avoidance. It is UK guidance, not an additional international COLREG rule.

For an authored training comparison, imagine two identical displayed CPA values. One follows a stable, well-observed track; the other follows a recent manoeuvre and uncertain inputs. The displayed digits are equal, but the evidence supporting them is not. Similarly, a point-target model does not automatically describe the clearance between the physical extremities of large vessels or a tow.

The educational lesson is to inspect assumptions, reference frames and data quality along with the result. More decimal places do not repair a poor model. No CPA alarm threshold is recommended here.

7. Action and checking belong together

Rule 8 addresses timely, positive action where circumstances permit, changes that are readily apparent, avoiding a succession of small alterations, and checking effectiveness until the other vessel is past and clear. It also preserves obligations associated with not impeding passage. The international Rule 8 text should be read in full.

A teaching diagram should therefore include a return path from action to observation. Without that path, it encourages the false impression that selecting an action completes the task. The intended result and the observed result are separate pieces of information.

In a fictional multi-target exercise, a change that improves one predicted separation could worsen another. This is a reason to review the complete scene, not an instruction to choose a particular course or speed. The same principle applies to an action whose apparent success depends on the other vessel continuing unchanged. The assumption remains something to test, not something established by the initial prediction.

8. Responsibilities are conditional, not an unlimited right

Rules 11–18 concern vessels in sight of one another and include overtaking, head-on, crossing and responsibilities between vessels. Rules 9 and 10 add channel and traffic-separation context. Rule 10 does not remove other rule obligations. IMO's COLREG explanation maps these subjects.

Rule 17 qualifies the stand-on vessel’s course-and-speed duty: specified circumstances allow action, and others require it. Read the complete conditions. Stand-on does not confer an unlimited right.

For learning, avoid a league table of vessel types or an isolated arrow labelled “priority.” Such graphics can conceal the very conditions that decide which provisions matter. A sailing vessel, for example, is not adequately described merely by a sail icon; the actual definitions, operation and encounter context need examination.

The purpose of this section is to prevent overgeneralization. It does not reproduce every encounter rule, signal or exception and should not be used as an operational decision tree.

9. Restricted visibility changes the reading task

Rule 19 addresses vessels not in sight of one another when navigating in or near restricted visibility. The in-sight stand-on/give-way encounter framework cannot simply be imported into that situation. Rule 19 includes its own conditions for safe speed, radar-based assessment and avoiding action. MCA's MGN 369 Amendment 1 explains this context.

An important classroom distinction is between seeing a symbol representing a vessel and seeing the vessel visually. A target identity displayed on an instrument does not itself answer the visibility question. Another useful distinction is between a radar track's apparent precision and the observer's confidence that all relevant objects have been detected.

This guide deliberately gives no radar-only manoeuvring instruction. A simplified fog diagram without the complete situation can create false confidence. The appropriate learning task is to identify the applicable section, read all its conditions and explain why the in-sight assumptions do or do not hold.

10. AIS and VHF add information, not exemptions

MCA's MGN 324 Amendment 2 highlights incomplete or incorrect AIS information and the dangers of VHF misidentification, misunderstanding and delay. The guidance emphasizes compliance with COLREG rather than using a radio discussion to bypass it.

An original example illustrates the information problem: a named AIS symbol, a radar echo and a voice on the radio are three observations that may need correlation. Treating them as one confirmed target without evidence collapses three uncertainties into one apparently reassuring label. More communication can be helpful, but the number of messages alone says nothing about whether the right parties understood the same situation.

For a debrief, distinguish “message transmitted,” “message received,” “meaning understood” and “action observed.” These are not interchangeable milestones. This reasoning is educational; it is not a recommendation to negotiate passing arrangements.

11. A learning framework and debrief

Use the cycle to review a completed fictional case. Identify the first relevant observation, the assumptions about visibility and waterway, the evidence used to assess risk, and the reason the selected rule context was appropriate. Then ask what later information would require reassessment.

Keep arithmetic correctness separate from decision quality. A correctly computed prediction based on the wrong target is not a sound result. Keep outcome separate from method too: a scenario ending without collision does not, by itself, show that every earlier judgement was well supported.

A useful debrief ends with specific learning points. “Understand equipment better” is vague. “Explain the reference used for this displayed motion value” or “distinguish visual observation from identification data” is reviewable. No real-world clearance decision follows from completing the exercise.

12. Frequently asked questions

Is there one universally safe CPA? This guide supplies none. Context, uncertainty, vessel characteristics and applicable procedures matter; a numerical threshold alone does not establish compliance.

Can the numerical example reveal intentions? No. The invented observation pair and its arithmetic do not establish what another vessel intends to do.

Does stand-on mean do nothing? No. Rule 17 has conditional stages and obligations that must be read together.

Does this article cover everything? No. Lights, shapes, sound signals, definitions, special circumstances and every detailed exception require fuller study. The purpose here is to connect evidence and context with the rules without pretending that a short diagram can navigate a ship.