When two ancient accounts of the same event disagree, the disagreement is usually about a number. That is so common that it needs explaining rather than scoring.
Three kinds of numerical disagreement
They are not equivalent, and treating them as one is how a real discrepancy gets buried under a false one.
- Between the Bible and an external record. Sennacherib’s prism gives eight hundred talents of silver; 2 Kings 18:14 gives three hundred. Both give thirty talents of gold.
- Between two biblical passages. 2 Kings 24:14 gives ten thousand deported; two verses later, seven thousand fighting men and a thousand craftsmen; Jeremiah 52:28 gives 3,023.
- Inside the external record itself. Assyrian accounts of the same campaign differ between exemplars, and Assyrian and Babylonian records disagree about who took Samaria.
The second kind is the one most often left out of popular treatments, and it is frequently the largest gap on the page.
Why ancient numbers behave this way
Units were not standard. A talent in one system is not a talent in another, and tribute could be assessed in one standard and recorded in a second.
Payment was rarely a single event. Tribute delivered in instalments, or continued over years, can be totalled differently by two records without either being wrong.
Inflation was a convention. Royal inscriptions round upward, and the larger the figure the more likely it is doing rhetorical work. Army sizes are the least reliable numbers in the entire corpus.
And transmission damaged numerals. Numbers in Hebrew manuscripts are among the readings most vulnerable to copying error, which is one reason the same event carries different figures in Kings and in Jeremiah.
Which disagreements actually matter
A figure matters when it is specific, when it appears independently in two traditions, and when it could have been otherwise. Thirty talents of gold in two records written in different languages by opposite sides of a war is a strong agreement precisely because it is arbitrary — there is no reason for both to land on thirty unless something real is behind it.
A figure matters much less when it is round, large, and inside a passage doing rhetorical work. Forty-six cities, two hundred thousand captives, thousands of chariots: these are the parts of the record where the genre is speaking.
The useful question is therefore not whether the numbers agree. It is whether the specific ones do.
What this site does with them
A dossier that reports agreement on a number and omits a disagreement in the same passage has selected its evidence, and the verdict scale is designed to make that visible: a grade attaches to a stated claim, not to a chapter, so a page can document the framework and still record that the counts conflict.
Where an internal biblical discrepancy is larger than the external one, that is stated in those words. It is the opposite of what most readers expect, and hiding it would be the easier and worse choice.
Related: the dossiers on Sennacherib in 701 and the fall of Jerusalem in 597, and the method note on the five verdicts.