Risk & Rescue
What Accident Databases Cannot Tell You
Incident records count events without knowing how many people were exposed, so they describe what goes wrong rather than how likely it is, which changes how they should be read.
Published accident reports are the most valuable learning material adventure sports have. They are also systematically incomplete, and reading them without knowing how invites wrong conclusions.
The missing denominator
A database records incidents. To turn that into a rate you need to know how many people did the activity without incident, and almost nobody counts that.
Without participation figures, a rise in recorded accidents cannot be distinguished from a rise in participation, better reporting or a change in what gets reported at all.
This is why headline comparisons between activities are usually meaningless. The activity with more recorded incidents may simply be the one more people do.
Selection decides what appears
Records generally capture events that triggered a rescue callout or a fatality. Injuries where somebody walked out and drove to a hospital often never enter any dataset.
That biases the record toward remote locations, serious outcomes and situations where a party could not self-recover, which is not the same as the full picture of harm.
Near misses are almost entirely absent, despite being the most instructive category, because nothing happened to generate a record.
Cause is assigned after the fact
Investigators reconstruct events from survivor accounts, scene evidence and inference. Where nobody survived, the sequence is a reasoned hypothesis rather than an observation.
Hindsight makes the eventual outcome look more predictable than it was at the time, which pushes reports toward describing decisions as obvious errors.
Reports written by practitioners for practitioners tend to be more careful about this, and the better ones explicitly state what is uncertain.
What they are genuinely good for
The value of the record is qualitative. It shows which failure modes recur, which is stable information even when the rates are unknowable.
Patterns such as rappel errors, communication breakdowns, late starts and equipment used outside its design purpose appear consistently across regions and decades.
That recurrence is the useful signal, because it identifies the small number of things that account for a large share of serious outcomes.
Reading them productively
The instructive question is not what the person did wrong, but what made the wrong option look reasonable at the time and what would have interrupted it earlier.
Annual reports from national rescue organisations, avalanche centres and climbing federations are the primary sources, and most are published openly.
They are best read alongside training rather than instead of it, because recognising a pattern in a report is a different skill from recognising it in the field.
