How the sighting probability score works

Most birding tools tell you a species was seen somewhere. That's a fact, not a forecast. The more useful question — if I drive there tomorrow, what are my chances? — needs more than a pin on a map. Flyway answers it with a single 0–100 score per hotspot, and this page explains exactly what's behind it.

What goes into the score

Four signals, drawn from public eBird observations over roughly the last 30 days, combined into one number:

Reporting rate — the biggest factor. Of the recent checklists filed at that hotspot, what share included the species? “On 70% of lists” is a far stronger statement than “someone saw one”. This is the closest thing to a real probability in the data.

Checklist volume — how much to trust it. A spot with sixty recent visits gives a reliable percentage; a spot with two doesn't. Heavily-birded sites are weighted up because their numbers mean more.

Abundance — how many were there. A flock of forty is a different prospect from a lone individual. eBird's “X” (present, not counted) is treated as one, so a vague report never inflates the score.

Recency — is it still there? A sighting from this morning counts for much more than one from three weeks ago. Recency decays smoothly rather than falling off a cliff.

How to read it

66–100 (green) — a genuinely good bet. The bird is reported often at a well-watched site, recently.
40–65 (amber) — plausible. Worth a stop if it's near your route, but bring patience.
Below 40 (gray) — possible, not likely. Often a single old record, or a site almost nobody birds.

Compare scores against each other within a result set rather than treating them as calibrated odds. The score's job is to rank your options, and it does that well.

What it can't tell you

Being straight about the limits matters more than a clean-sounding number:

It measures reports, not birds. Species are “found” where birders look. A quiet marsh nobody visits will score low even if the bird is standing in it.
It's a 30-day window. It reflects current conditions, not seasonal patterns. A migrant scoring zero in July may be reliable in May.
It doesn't know time of day, weather or tide. Those often matter more than any statistic — an owl scoring 80 is still an owl at 2pm.
Skill and effort aren't in the model. Reporting rates come from birders of every experience level, some of whom walked further than you will.

eBird's public API doesn't expose the full historical bar-chart frequencies you see on the eBird website, so the score is computed from recent observations and checklist data rather than long-run seasonal averages. It's an honest estimate from what's available — not a black box, which is exactly why the underlying numbers are shown on every result card.

Where you'll see it

The score appears on every result in Find a Bird, where hotspots are ranked by it. A related figure — the recent-sighting probability for a species in your quiz area — appears when a quiz answer is revealed, which is why some birds are worth more points than others: rarer birds score higher when you get them right.

See it in action →