What this page leaves out, and how much to trust it
Militias and irregular forces are excluded from the rankings on
purpose. The sources that exist for them, actor lists coded from
recent conflicts, skew so hard toward the post 1989 world, and toward
Africa and the Middle East, that drawing them would paint irregular
force as a recent southern phenomenon. It is neither. Rather than
reproduce that distortion, the page keeps to states and marks the
places where the uncounted bent history anyway. Absence of a militia
in the data was never absence of a militia.
The spine of the counted era is the Correlates of War National
Material Capabilities series, which records personnel and
expenditure for every state annually from 1816, with SIPRI carrying
spending from 1949 and the IISS Military Balance the recent decades.
One thing that series shows is worth flagging: it carries Qing China
at a flat million men every year from 1860, a registered figure that
would put China first on personnel for four straight decades. This
page leaves the Qing out of those decades because the register and
the field had come apart. The deep past leans on the narrative tradition,
Dupuy and Dupuy's encyclopedia above all, and polity level work like
Seshat. These layers do not share a unit of analysis, which is why a
seam is drawn across the page at 1816 rather than papered over.
The map draws every power as the modern countries its territory
overlapped, because modern borders are the only ones the atlas
knows. No empire on this page respected them. Read the fills as
gesture, not cartography: the deepest ink is the age's first power,
the wash fades down the ranking, and the faint dashed wash marks the
furthest lands a power could credibly strike.
The relic photographs in the "what remains" strips come from
Wikimedia Commons and museum open access collections, under public
domain or Creative Commons licenses; each is credited beside the
image and links to its source page.
What the ranking actually measures changes as you scroll, so it is
worth saying plainly. Down to the 1940s the order roughly follows
mass. From the 1950s it follows capability and money, while the
personnel figures beside each power keep counting heads, so in the
later slices those numbers are not what the ranking is sorted by. In
the 1960s they run in almost exactly the reverse order. The figures
themselves also mix peacetime strength, wartime mobilisation and
present day active duty, which are three different things.
Every slice was drafted from general knowledge and then checked
against published scholarship in July 2026, the counted era against
the Correlates of War personnel series itself. That pass corrected
real errors: a chariot count that belonged to both sides of Kadesh
rather than to Egypt, a Chinese dynasty two centuries out of place,
Carthage paired with the wrong city in 146 BCE, an emperor confused
with his father, and a missing Mongol empire. Every personnel figure
is still a rounded estimate and the ancient ones are contested by
profession. The rankings remain what Claude thinks conventional
military history says, and that tradition skews toward states that
kept records.
Part of the melting pot.