Guides · Updated 2026-10-28
Conflict data sources compared: ACLED, UCDP and GDELT
Researchers, journalists and aid agencies rely on a few large datasets to track political violence. They often give different answers to simple questions such as “how many wars are there?” because they are built for different purposes. Knowing how each one works helps you read the charts and maps based on them, including ours.
ACLED
The Armed Conflict Location & Event Data project records individual events of political violence and protest, grouped into types such as battles, explosions and remote violence, violence against civilians, riots and protests. Each event has a date, a location and the actors involved, coded by researchers from news, local partners and other reports. It is widely used for maps and weekly trend analysis. Access and licence terms depend on the user, so check its website before reusing data.
UCDP
The Uppsala Conflict Data Program at Uppsala University in Sweden is one of the oldest academic sources. It defines an armed conflict as a contested incompatibility over government or territory in which the use of armed force between two parties, at least one of which is the government of a state, results in at least 25 battle-related deaths in a calendar year. It also keeps separate datasets for fighting between non-state groups and for one-sided violence against civilians. Its reviewed annual figures are published after the year ends, alongside faster preliminary data.
GDELT
The GDELT Project automatically processes news from around the world and extracts events, places and themes by machine, updating many times a day. It is very fast and very broad, but because it is automated it can misread headlines, count the same event many times or place it in the wrong location. It is useful for spotting where coverage is rising, less so for counting events precisely.
Why the totals differ
- Definitions. A threshold of 25 deaths per year, every recorded clash, or every protest give very different counts.
- Sources. Datasets that rely on international news miss events in places with little coverage; those with local partners record more.
- Speed versus review. Real-time data is revised later; annual data is slower but more stable.
- Units. Some count conflicts, some count events, some count pairs of fighting groups.
- Coding choices. Whether a riot that turns deadly is counted as a riot or as violence against civilians changes the totals for each category.
Tips for using any dataset
Read the codebook or methodology page before drawing conclusions, compare trends rather than single numbers, and be careful with places where reporting is restricted: a fall in recorded events may mean less reporting, not less violence. When two datasets disagree, the reasons are usually in their definitions rather than in an error.
How this site differs
JellyDock Conflict Watch is not a research dataset. We collect headlines from publisher feeds and news data services, match them to cities and group them into events. An event is shown in full only when at least two independent outlets report it at a known city; single-source reports appear faded and labelled “Unverified”. We do not code actors or attribute attacks. See how the map works and our data sources.