Where ACLED and UCDP Disagree — and Why It Matters
UCDP registered 65 state-based conflicts in 2025, the most since 1946. ACLED publishes no such number at all — by design. Neither is wrong. Understanding why they diverge is the difference between citing conflict data and misquoting it.
Open the live conflict map →The short answer
The two most-cited conflict datasets are built to answer different questions, so they cannot agree on "how many wars there are" — and were never supposed to. The Uppsala Conflict Data Program (UCDP) counts conflicts: in its June 2026 annual release, it registered 65 conflicts involving states in 2025, the highest number since its records begin in 1946 (Davies, Pettersson & Öberg, Journal of Peace Research, 2026). ACLED counts events: more than 200,000 political-violence events and over 240,000 reported deaths across a recent 12-month period, without ever publishing a single "number of wars." Media "major war" lists count 14–20. This site tracks 29 theaters. Same planet, four defensible answers.
Three design choices create the whole gap
| Design choice | UCDP | ACLED |
|---|---|---|
| 1 | Unit: the conflict — a named incompatibility between two organized parties, tracked across years. | Unit: the event — one battle, strike, riot or attack, geolocated and dated. |
| 2 | Threshold: 25 battle-related deaths in a calendar year before a conflict enters the count. | Threshold: none. A protest with zero deaths and a battle with hundreds are both records. |
| 3 | Purpose: comparability across decades — the same definitions applied since 1946 are exactly what makes "highest ever" a meaningful sentence. | Purpose: monitoring — near-real-time coverage of who did what, where, this week. |
Every headline disagreement between the two follows from this table. For the fuller three-way comparison including the CFR Global Conflict Tracker, see ACLED vs UCDP vs CFR.
Five places you can watch them disagree
1. Mexico: a war to one dataset, a category puzzle to the other
By event and fatality volume, Mexico's cartel violence ranks among the most violent situations ACLED records anywhere. UCDP files most of it as non-state conflict — organized violence between cartels rather than war against the state — so it barely moves the "wars" number that makes headlines. Neither coding is wrong; "is Mexico at war?" simply has no dataset-independent answer.
2. Haiti: enormous violence, zero "wars"
Haiti's gang crisis produces mass casualties and displacement, yet involves no rebel movement fighting a government over an incompatibility — the thing state-based conflict definitions exist to capture. A count of wars can be flat while organized violence against civilians surges. Event data sees it; conflict counts mostly don't.
3. The Taiwan Strait: the risk no battle-death dataset can see
The inverse case. Taiwan Strait tensions and the South China Sea are among the most strategically dangerous situations on earth and generate almost no battle deaths — so they are invisible in both datasets' totals. Qualitative trackers like CFR's exist precisely for this blind spot, and it is why this site labels such theaters "tensions" rather than forcing them into a war count.
4. The Sahel: one insurgency or six datasets' worth?
This site tracks the Sahel insurgency as one transnational theater — JNIM and Islamic State affiliates operating across Mali, Burkina Faso and Niger. Country-coded event data splits it into three national series; dyad-based conflict data splits it again by armed pair. One editorial war can be half a dozen statistical ones, and cross-dataset "war counts" quietly mix those units.
5. 2025's interstate spike: when a four-day clash counts like a four-year war
UCDP's 2026 release reports that interstate conflicts doubled for the second year running — from two in 2023 to eight in 2025 — including Russia–Ukraine, Iran–Israel, India–Pakistan and Israel–Syria. A binary count weighs the days-long May 2025 India–Pakistan exchange the same as the fourth year of full-scale war in Ukraine. Event data shows the thousandfold difference in intensity that the count, by design, flattens. Both facts are true; they are different facts.
What both approaches miss
Battle-related death counts exclude most indirect deaths — famine and disease kill quietly and outside the definitions. In Sudan, direct-kill estimates and researcher estimates that account for indirect deaths diverge by large margins, and access restrictions mean even the direct counts are floors, not totals. Whatever number you cite for Sudan, it is an undercount; the datasets differ mainly in how honest their footnotes are about it.
How to cite a conflict number without misquoting it
- Name the dataset and the year. "65 state-based conflicts (UCDP, 2025)" is a fact. "There are 65 wars" is not.
- Never mix units. An ACLED event total and a UCDP conflict count cannot be compared, added, or trended against each other.
- Match the tool to the question. Trends across decades: UCDP. What happened this week and where: ACLED. Strategic stakes of a quiet flashpoint: a qualitative tracker.
- Date everything. Every control claim on this site carries its source and an as-of date for exactly this reason — see our methodology.
This site's 29 theaters are an editorial middle path, not a fifth dataset: broad enough to include what qualitative trackers flag, narrow enough to brief. We hold the sources side by side and attribute rather than reconcile — where they disagree, the disagreement is the information.
Sources & further reading
- Davies, Pettersson & Öberg, "Organized violence 1989–2025, and violent political protests", Journal of Peace Research (June 2026) — the UCDP annual release.
- Uppsala University, "UCDP: record number of conflicts between states" (2026-06-09).
- ACLED — methodology and event data; UCDP — definitions and datasets; CFR Global Conflict Tracker.
- On this site: ACLED vs UCDP vs CFR and How many wars are happening right now?
Related reading
Frequently asked questions
Why do ACLED and UCDP give different numbers of conflicts?
Because they count different things: UCDP counts conflicts that pass 25 battle-related deaths in a year, applying the same definitions since 1946; ACLED counts individual political-violence events with no fatality threshold and publishes no single number of wars. The difference is design, not error.
How many armed conflicts did UCDP record in 2025?
UCDP's June 2026 annual release registered 65 conflicts involving states in 2025, the highest number since its records begin in 1946, with interstate conflicts doubling for the second year running to eight.
Is Mexico's cartel violence a war?
It depends on the definition. ACLED records it as one of the largest concentrations of political violence it tracks; UCDP classifies most cartel violence as non-state conflict rather than war against the state; many major-war lists exclude it entirely.
Which conflict dataset should I use?
Match the tool to the question: UCDP for trends comparable across decades, ACLED for near-real-time event monitoring, and qualitative trackers such as CFR's for flashpoints that generate risk but few battle deaths. Always name the dataset and year when citing a figure.