Every fatality figure here is an estimate, and UCDP makes that explicit by publishing low, best and high values for each conflict. The “best” estimate is the researchers’ most defensible figure given the available reports; the low and high bound the range of reasonable uncertainty. Treating the best estimate as a single hard number discards the very information the range is meant to convey.
Recorded figures almost always undercount. Deaths that are never reported, reports that cannot be corroborated, and access restrictions during active events all push recorded totals below the true toll. The uncertainty is largest precisely when interest is highest — while a situation is ongoing — and it narrows only as documentation accumulates afterward. A recorded figure is best understood as a floor.
It matters which deaths a figure counts. Battle-related deaths — the basis for UCDP’s conflict thresholds — are those resulting directly from fighting between parties. They do not include the indirect toll: deaths from disease, hunger, lack of medical care and displacement that conflict can cause. Those indirect, or “excess,” deaths are often estimated separately and can far exceed the battle-related count.
Because of all this, small differences between figures rarely carry the weight people assign to them, and comparisons across places or years can mislead if reporting conditions differed. The responsible way to read these numbers is to lead with the range and the uncertainty, name what the figure does and does not include, and defer to humanitarian sources for the human picture.