How Sky Event Radar Uses Evidence, Confidence and Missing Data
Why fit scores are not probabilities, why unknown evidence matters, and how independent data layers should be combined.
A fit score is comparative
The score measures how well one explanation fits the submitted behavior and available external signals. It is not a calibrated probability and the candidate scores do not need to sum to 100. Several explanations can remain plausible.
Independent evidence is more valuable
A blinking report and a nearby aircraft are not the same kind of evidence. The first comes from the witness description; the second comes from an external data feed. When independent clues agree, confidence strengthens more meaningfully than when several versions of the same clue agree.
Unknown is not negative
If a historical aircraft feed is unavailable, that source cannot be used for or against an aircraft explanation. Treating unavailable data as no aircraft would create false certainty. The site separates unknown, supporting and weakening signals for this reason.
Coverage is never perfect
Sensor networks, ADS-B reception, fireball catalogs, weather models and orbital elements all have limitations. A robust result shows the coverage boundary next to the data rather than hiding it in fine print.
The best result can still be inconclusive
Sometimes several categories genuinely fit. A responsible answer may be mixed evidence plus the next observation that would discriminate between them. That is more useful than inventing a precise percentage.
Save this before you go outside.
These are the details most likely to improve a later identification.
- ✓Check whether evidence is independent
- ✓Read source status
- ✓Distinguish unknown from weakens
- ✓Look at alternative candidates
- ✓Use the result to decide the next verification step
Turn the observation into an evidence report.
The identifier compares what you recorded with independent public data.