What this data cannot tell you
This page matters more than the two before it. The tool influences how people judge a road or a junction, so the limits have to be stated plainly. Every claim below was measured against the data.
It cannot rank the most dangerous junctions
This is the important one. Taichung records the most accidents in the country, yet of the top 200 hotspots nationally, New Taipei holds 121 and Taichung only 19.
The reason is not that New Taipei is more dangerous. It is that geocoding behaves differently between counties. The rate at which a single coordinate is reused is 1.38 in New Taipei and 1.01 in Taipei City, where almost every accident has a unique coordinate and therefore no hotspot can ever accumulate.
Any "most dangerous junction" ranking built on this data is really a ranking of geocoding behaviour by county. That is why this tool has no leaderboard and only reports areas. Over a radius of 300 to 500 metres, a coordinate error of a few tens of metres averages out.
More than a quarter of addresses are not exact positions
- 26.2% are written as "0.0 metres in front of X Road", which is a reference point for a stretch of road, not the point of collision
- 41.5% use the "A Road / B Road" junction format, but in 19.5% of those the two road names are identical, meaning a segment of one road rather than a junction
So a cell at maximum zoom can be tens of metres off. Street-level reading is reasonable; junction-level conclusions are not.
Only two years and eight months
Open data with coordinates starts in 2024. The earlier records, 2013 to 2021, carry addresses only. Two years and eight months is enough to show a stable spatial pattern, but it is thin for trend questions such as whether a place has recently become more dangerous, especially for fatalities: about four thousand across the whole country over 33 months, which for a single stretch of road is often a single-digit number with large random swings.
More accidents does not mean more dangerous
Places with more traffic have more accidents. This data contains no traffic volume, so there is no way to compute a rate per ten thousand vehicle passes. A main road downtown will always be hotter than a mountain forestry road; that reflects how many vehicles use it, not how dangerous it is. The denominator that "dangerous" needs is not in the open data.
Only reported accidents
A1 and A2 both mean an accident was reported and attended. Minor collisions settled privately never enter the data, so the true number of light accidents is higher, and reporting rates may differ between areas.
What the tool is good for
Seeing the overall pattern in an area, comparing time ranges, and finding stretches of road where fatal accidents cluster. It is not suitable for claiming one junction is more dangerous than another, and it should not be used as a basis for property or insurance decisions.
Other tools on this site
- Old Photo Motion — Turn an old photo into a 6-second clip that blinks and smiles
- age-me — See the same person older or younger
- Campsite Legality Check — Which of 1,807 campgrounds comply, violate, or are not listed
- Flood Potential Map — WRA flood potential under 10 rainfall scenarios, by township
- Find My Pet — Lost & found pets across Taiwan on a map; list without a microchip, report sightings without login
All run on one machine at home; the data tools use Taiwan government open data and state their limits.