Prophecies about nations, empires and Israel

The short version

1. Score every prediction, not a highlight reel.

   We took every oracle against a foreign nation in the Old Testament's own collections (62 propositions) and every famous case on both sides (32). Then, because the famous cases about Israel were picked from a much larger body of promises, we scored that whole body too: 696 promises of restoration to Israel and Judah, listed and coded by reviewers who were not told the purpose.

   The evidence ↓
2. Most hits were cheap, and some misses are real.

   In the 94: 49 hits, 4 clear misses, 10 partial, 26 ambiguous. Most hits are "this small kingdom will fall" in an age of empires. Egypt was never desolate for forty years, Damascus never stopped being a city, and Tyre was rebuilt. The foreign oracles alone show no sign of foreknowledge.

   The evidence ↓
3. The famous proofs turn on dating.

   An accurate prophecy is worth at most the odds that it was written before the event. On the dates most specialists accept, Daniel and Cyrus add little, and Daniel counts slightly against.

   The evidence ↓
4. The best hits come with failed promises.

   Israel's survival and return, the return from Babylon, Alexander's mole at Tyre and Jesus on the temple are real. But the oracles that promised the return also promised that Israel would never again be uprooted, that no foreign army would enter Jerusalem again, and lasting peace. Scored whole, the promises to Israel go from about 15 to 20 to 1 in favour to about 2 to 1 against, or far less.

   The evidence ↓
5. Net: no margin.

   Scored whole and blind, the record gives a likelihood ratio of about 0.3 (range 0.01 to 10) on mainstream dates: no evidence for foreknowledge, and on a plain reading a lean against. The same rows scored as a list of fulfilled prophecies give about 10,000,000. Traditional dates raise the result 1,000 to 100,000 times, and that is a judgment about dating, not new evidence.

   The evidence ↓

The question

Do the Bible's predictions about nations, cities, empires and Israel come true more often, and more specifically, than predictions written by well-informed people of the same period would?

Two hypotheses make different predictions about the record:

A variant of H0 adds canonical selection: under Deuteronomy 18:21-22, a prophet whose words failed was remembered as a false prophet (Jeremiah 28), so the books that survived may be pre-selected for success. We model it as a sharper human forecaster.

The measure throughout is the likelihood ratio (LR): how many times more probable the record is if D is true than if H0 is true. An LR of 1 means the evidence does not discriminate; above 1 favours foreknowledge; below 1 favours human authorship. This page covers the historical half of prophecy. Daniel 9's seventy weeks and Isaiah 53 are on the Messiah page.

Rules of evidence

These rules were fixed before any total was computed. Stratum C and the blind coding were added in a later pass; their rules and base rates were fixed before any of the new outcomes were seen.

1. The whole set, not a highlight reel. The fullest count of biblical predictions, J. Barton Payne's, is reported as 1,817 predictions on 737 subjects, in 8,352 verses (about 27% of the Bible).¹ Most of his items are messianic or about the end times. For the historical items we used the genre that biblical scholars themselves define, the oracles against the nations,² and included every one, whatever its outcome:
   • Stratum A (62 propositions): Isaiah 13-23, Jeremiah 46-51, Ezekiel 25-32, Amos 1-2, Zephaniah 2, Obadiah, Nahum, Zechariah 9:1-8 and Joel 3:4-8.
   • Stratum B (32 propositions): every case on both sides' lists. The apologists': Cyrus, the seventy years, Daniel, Josiah named, Israel's dispersion and return, the temple. The skeptics': Josiah "in peace", Jehoiakim, Zedekiah, Daniel 11:40-45, Mark 13:30, Amos 9:15, Jonah. Stratum B is selected, so it is reported separately.
   • Stratum C (696 propositions; 647 scored): every promise of restoration or blessing to Israel and Judah in Leviticus 26, Deuteronomy 28-30 and the prophets. This is the body of texts the famous Israel cases were taken from, so it is the fair test of them. Promises about an individual future king are scored on the Messiah page; promises whose fulfilment would run through Jesus or the church (the nations coming to Israel's God, the Spirit poured out) are set aside here so nothing counts twice.
   Not done yet: a random sample of Payne's 737 subjects, which would be a fully external check.
2. Blind where it matters. Stratum C was listed by one reviewer and its outcomes coded by others, none told what the test was for; strata A and B were coded a second time the same way, from the reference and a neutral wording of the prediction only. The reviewers were AI agents working mostly from memory, so the result can be reproduced from the published files but has not yet been replicated by human coders.³ The base rates for each kind of promise were fixed before any outcome was seen.
3. One proposition, one row. "X will fall to Y and never be rebuilt" is two predictions, each with its own score. Several oracles about the same event form a group that counts once, so five oracles against Philistia are not five independent hits.
4. Outcomes. Hit plainly fulfilled; Miss plainly not fulfilled within any horizon the text gives; Partial part fulfilled, part not; Ambiguous turns on interpretation, or the record is silent; Open about the "latter days", still future; Conditional cancelled in the text itself.
5. Dating weight w. For each row, w is the probability that the text was written before the event. It is set from non-prophetic evidence (language, historical setting, manuscripts, the text's own headings), never from "it is accurate, so it is late", which would assume H0. Weights follow the critical consensus; traditional dates are a sensitivity run.
6. Base rate p0. The probability that a well-informed contemporary would have got it right. Defaults: generic doom of a small state in 750-500 BC, 0.85; permanent abandonment of a threatened city, 0.42 (computed: 7.5 of the 18 target cities are abandoned today); disappearance of a named people, 0.75 (computed from 14 peoples); survival of a small Levantine people, 0.11 (computed, with a Laplace correction). Item-specific values are judgments with the reason in the table. Where readers of the text brought the fulfilment about, p0 is raised.
7. Nothing hidden. Every miss, partial, ambiguity and open item is in the table and the script. Bounding runs score every ambiguous and partial item as a miss (the skeptic's bound) or a hit (the believer's).

Hit:        LR = (Σ w_s·q + w_post) / (Σ w_s·p0_s + w_post)
Miss:       LR = (1 − q) / (1 − p0)     nobody writes a miss after the event
Ambiguous:  LR = 1                       a 50/50 mix of a hit reading and a miss reading
Partial:    LR = (hit) × (miss)
Open:       left out (neutral run); counted as a miss in the strict run
Conditional: left out, and listed
Group:      the mean of its members' log LR, counted once
Headline:   averaged over q ~ Uniform(0.6, 0.99), not tuned

The evidence

The whole set: 94 predictions

Here is how the 94 propositions came out. The full tables, row by row, are in the two drills below the chart.

A hit is not the same as evidence. Of the 34 hits whose texts are confidently dated before the event (w ≥ 0.8), only 4 had a base rate of 0.25 or less, that is, were more than "a small state will fall":

• Tyre's stones and dust thrown into the sea (Alexander's mole, 332 BC);
• Cyrus, but only on the early date of Isaiah 40-48;
• the return from Babylon;
• Israel's regathering, which readers of the texts partly brought about.

The misses, meanwhile, are not marginal. Ezekiel says of Egypt: "Her cities among the cities that are laid waste will be a desolation forty years. I will scatter the Egyptians among the nations" (Ezekiel 29:12, WEB).⁵ No such desolation happened. Isaiah says: "Behold, Damascus is taken away from being a city, and it will be a ruinous heap" (Isaiah 17:1, WEB). Damascus is among the longest continuously inhabited cities in the world.

The famous cases, both sides

Each case in one line, then the details in its own drill. Each LR is the event group's value in the neutral run at q = 0.9; numbers below 1 count against foreknowledge.

Stratum C: every promise of restoration to Israel, scored whole

The strongest item in the famous lists is Israel's survival and return. But it was taken from a large body of oracles that promise Israel a future: return, rebuilding, safety, peace, prosperity, and power over its neighbours. To test the item fairly we scored that whole body. One reviewer listed every checkable promise of restoration to Israel and Judah in Leviticus 26, Deuteronomy 28-30 and the prophets, 696 in all, checking the wording of each against the WEB text; three others coded what happened, without being told why.³

The hits are real: the return from Babylon, the rebuilt city and temple, the restored worship, the survival of the people, and the modern return. The misses sit where a hopeful human author would overreach: permanence and peace. "From now on the uncircumcised and the unclean will no more come into you" (Isaiah 52:1, WEB), yet foreign armies entered Jerusalem in 167 BC, 63 BC and AD 70. "The priests the Levites won’t lack a man before me to offer burnt offerings" (Jeremiah 33:18, WEB), yet sacrifice ceased in AD 70. "They will no more be plucked up out of their land" (Amos 9:15, WEB). Ezekiel's river from the temple that heals the Dead Sea (Ezekiel 47:1-12) never flowed. Promises the text itself sets "in the latter days" or at the day of the LORD are counted as open, not as misses.

The analysis

Multiplying the rows (each event group once) gives the likelihood ratio for the whole record. Every value averages over the reliability q from 0.6 to 0.99, so we never pick the q that suits us.^3,4

Cherry-picking, measured. On the same 94 rows, with the same base rates and dating discounts, the 49 hits alone give about 10⁷; all 94 give about 70. In the foreign oracles alone, hits give about 1,400 and the whole set 0.23. That gap is what a list of fulfilled prophecies hides, before any of Stoner's guessed odds are added.

Counting the promises once. The two ways of grouping stratum C bracket the truth. Counting each group as the lister drew them (as fixed in advance) treats the size, gates, district and river of Ezekiel's single temple vision as separate events, and gives 10⁻⁴ to 10⁻⁵. Counting each of the nine kinds once treats every "no more" promise in every book as one, and gives 0.13 to 4. Either way, the positive result is gone.

The interpretive rule is the biggest lever. If every ambiguous and partial item in the 94 is scored as a miss, the set gives about 10⁻⁵; if every one is a hit, about 10¹³. Historical prophecy cannot be scored without first fixing how literally to read it. On our plain-reading rule the 94 alone give a modest positive result; scored whole and blind, they do not.

Reliability cuts both ways. At a fixed q, the 94 give about 160 at q = 0.75, about 3 at 0.9 and about 0.01 at 0.95. A version of D that claims near-perfect literal accuracy is disfavoured by its own misses. Adding stratum C makes this sharper: with every promise to Israel included, the record favours D only if its prophets were often wrong in the plain sense, and then each hit is worth little.

The dating question

Every famous "proof" in this area depends on one question: was the text written before the event? The model makes the answer exact.

LR ≈ 1 + w / (1 − w)

So the case for Daniel or Cyrus as decisive evidence must first win the dating argument on independent grounds. Here is Daniel alone as the reader moves its early-dating weight (q = 0.95):⁴

Below w ≈ 0.45, Daniel counts against foreknowledge, because 11:40-45 is then a miss by an author writing in 164 BC. The same holds in reverse: if you are confident the book is from the 6th century, Daniel is strong evidence, and 11:40-45 must be read as a prophecy of a later figure, which can never be checked.

The headline (about 0.3) uses the critical-consensus dates. A reader who accepts the traditional dates of Daniel and Isaiah 40-66 should multiply it by about 1,000 to 100,000 (with everything scored, traditional dates move the result from 0.13-4 to about 340-230,000 when each kind of promise counts once). That multiplier is the dating question, not new evidence, and the Messiah page and the history page must use the same dating weights. The Dead Sea Scrolls bound the dates from above; they do not fix them.

Objections

What this shows, and what it doesn't

What we don't claim

• Stoner's odds: Tyre "1 in 7.5 × 10⁷", Babylon "1 in 5 × 10⁹", a chapter total of "1 in 5.76 × 10⁵⁹". Each factor is a guess, the product assumes independence, and it counts "never rebuilt" (Tyre is a city) and Babylon's "stones not reused" (its bricks were widely reused) as fulfilled. Stoner's grandson notes on the book's site that its probability numerators were usually taken as 1, and that this assumption is no longer satisfactory.¹¹
• Hugh Ross's "1 in 10²⁰" for Israel's scattering and regathering. We found no derivation.
• "Tyre is a bare rock to this day." It was rebuilt by 315 BC, and Sur exists today.
• "Nebuchadnezzar destroyed Tyre." He besieged it for thirteen years and Tyre submitted; Ezekiel 29:18 says he got no wages.
• "Babylon fell at its zenith, as Isaiah said, never inhabited again, its bricks never reused." The fall in 539 BC was peaceful, the decline took centuries, and the bricks were reused.
• "Isaiah named Cyrus 150 years in advance" as a settled fact. It needs an 8th-century date most scholars reject; the Great Isaiah Scroll (about 125 BC) does not establish it.
• "The Dead Sea Scrolls prove Daniel early" or "4Q114 makes a post-164 date highly improbable". Its range, 230-160 BC, overlaps the 160s, and the paper says so.
• "Jeremiah's seventy years were fulfilled to the year." That depends on the chosen endpoints, and seventy was a stock figure.
• "Daniel predicted Rome." An interpretation, contested within the text itself.
• "Petra's desolation fulfils the Edom prophecies." Petra was Nabataean and flourished after Edom fell.
• "Ezekiel's prophecies against Egypt were fulfilled." The forty-year desolation never happened.
• "Damascus will soon be destroyed, fulfilling Isaiah 17:1." A prediction unfulfilled after 2,700 years is not evidence.
• Day-exact calculations of 14 May 1948 from Ezekiel 4's 390 (or 430) days multiplied by 7 (Leviticus 26:18), or "Shall a nation be born at once?" (Isaiah 66:8, WEB). The multipliers and start dates are chosen after the fact.
• "Nineveh was destroyed by a flood exactly as Nahum said." The flood comes from Ctesias via Diodorus, a late and contested source.
• "No other sacred text has anything remotely comparable." It has never been tested on a common rubric.
• "Josiah was named 300 years in advance." The composition date of Kings removes the force.
• Any single probability for historical prophecy without its dating assumption stated. Ours is about 0.3 on critical dates; we always say so.
• "The Dead Sea Scrolls and the Septuagint prove the prophecies were written before the events." Manuscripts show a text existed by a date. That settles AD 70 and 1948, which every relevant text precedes. It does not settle Daniel's empires (the Daniel scrolls are from the 2nd century BC) or Cyrus (the Great Isaiah Scroll is from about 125 BC, four centuries after him).
• Any "1 in 10^N" built by multiplying a list of fulfilled prophecies. On this page's own rows, with its own base rates, the hits alone give about 10⁷ and the whole record about 0.3. The difference is the selection, not the evidence.
• "Israel's history fulfils the prophets, so the Bible is proven." The survival and return are real hits, but the same oracles promised that Israel would never again be uprooted and that foreigners would never again enter Jerusalem. Scored whole, the promises to Israel give no margin.

Sources

1. Payne, J. Barton, Encyclopedia of Biblical Prophecy (Harper & Row, 1973). The 1,817 predictions and 8,352 verses as reported by Christian Courier; the 737 subjects from the publisher's description; 27% is 8,352 of the Bible's roughly 31,100 verses. Cited from the scholarly literature; not read by us.
2. Hayes, J. H. (1968). "The Usage of Oracles against Foreign Nations in Ancient Israel." JBL 87: 81-92. Cited from the scholarly literature; not read by us.
3. Our whole-universe and blind checks: truth/research/calc/08-whole-universe-blind.py and its output, with the blind inputs and codes (08-blind-AB-input.tsv, 08-blind-AB-coding.tsv, 08-blind-C-universe.tsv, 08-blind-C-coding.tsv) and the group map fixed before coding (08-blind-C-map.tsv). The method follows the blind, pre-registered messianic-prophecy test (omega.logos.so, ceiling route c10: omega/research/ceiling/c10-salvation/RECORD.md). Promise wording checked against the World English Bible.
4. Our scoring script and its output: truth/research/calc/08-historical-prophecy-lr.py (all 94 rows, the base-rate lists, the model and every run) and 08-historical-prophecy-lr.out.txt.
5. World English Bible (public domain), ebible.org/eng-web. All Bible quotations on this page. Read by us.
6. Josephus, Against Apion 1.154-160 (Whiston), lexundria.com. Read by us.
7. Josephus, Antiquities 10.180-182 and 10.227-229 (Whiston), lexundria.com. Read by us.
8. Arrian, Anabasis of Alexander 2.15-24 (the siege of Tyre, 332 BC). Cited from the scholarly literature; not read by us.
9. Diodorus Siculus, Library of History 2.26-27 (Nineveh, from Ctesias) and 19.61 (Antigonus at Tyre). Cited from the scholarly literature; not read by us.
10. Till, F. (1999). "The Tyre Prophecy Again." The Skeptical Review, March-April, reposted at isitgodsword.blogspot.com. Read by us.
11. Stoner, P. W., Science Speaks, ch. 2 "Prophetic Accuracy", sciencespeaks.dstoner.net; and D. W. Stoner's note on the probability numerators, sciencespeaks.dstoner.net. Read by us.
12. The Babylonian Chronicles (ABC 3, fall of Nineveh; ABC 5, Nebuchadnezzar; ABC 7, the Nabonidus Chronicle), the Cyrus Cylinder and Sennacherib's prism. Grayson, A. K., Assyrian and Babylonian Chronicles (1975); translations at livius.org. Cited from the scholarly literature; not read by us.
13. Strabo, Geography 16.1.5 (Loeb), penelope.uchicago.edu. Read by us.
14. Cassius Dio, Roman History 68.30 (Loeb), penelope.uchicago.edu. Read by us.
15. BM 33041 (Nebuchadnezzar's 37th year, against Egypt), discussed at jhalsey.com. Cited from the scholarly literature; not read by us.
16. Kahn, D. (2018), on Nebuchadnezzar's failed campaigns against Egypt (title not verified). Cited from the scholarly literature; not read by us.
17. Herodotus, Histories 1.191; 2.177; 3.159. Cited from the scholarly literature; not read by us.
18. Josephus, Antiquities 13.62-73 (Onias and Leontopolis; Whiston), lexundria.com, Read by us. Antiquities 13.257 (Hyrcanus and Idumea): Cited from the scholarly literature; not read by us.
19. Dalley, S. and Goguel, A. (1997). "The Sela' Sculpture: A Neo-Babylonian Rock Relief in Southern Jordan." ADAJ 41. Cited from the scholarly literature; not read by us.
20. Polybius 31.9; 1 Maccabees 6; 2 Maccabees 9 (the death of Antiochus IV). Cited from the scholarly literature; not read by us.
21. Coşkun, A. (2019). "The Chronology of the Desecration of the Temple and the Prophecies of Daniel 7-12 Reconsidered." Historia 68: 436-462, hdl.handle.net/10012/18089. Abstract read by us.
22. Popović, M., et al. (2025). "Dating ancient manuscripts using radiocarbon and AI-based writing style analysis." PLOS ONE 20(6): e0323185, journals.plos.org; and the University of Groningen press release, "Dead Sea Scrolls older than previously thought" (June 2025). Read by us.
23. Associates for Biblical Research, "Dead Sea Scroll 4Q114 and the Date of Daniel", biblearchaeology.org. Read by us. (Apologetic; overstates the paper.)
24. Kitchen, K. A. (1965). "The Aramaic of Daniel." In Wiseman et al., Notes on Some Problems in the Book of Daniel (Tyndale). Cited from the scholarly literature; not read by us.
25. Collins, J. J., Daniel (Hermeneia; Fortress, 1993). Cited from the scholarly literature; not read by us.
26. Grayson, A. K., Babylonian Historical-Literary Texts (1975): the Dynastic Prophecy. Cited from the scholarly literature; not read by us.
27. Jerome, Commentary on Daniel, prologue (preserving Porphyry's argument). Cited from the scholarly literature; not read by us.
28. Josephus, Antiquities 11.1-18 (Cyrus and Isaiah; Whiston), lexundria.com. Read by us.
29. Leuchter, M. (2004). "Jeremiah's 70-Year Prophecy and the ATBASH Codes." Biblica 85: 503-522, bsw.org. Cited from the scholarly literature; not read by us.
30. Lewis, D. M., The Origins of Christian Zionism (Cambridge, 2010). Cited from the scholarly literature; not read by us.
31. Kloppenborg, J. S. (2005). "Evocatio Deorum and the Date of Mark." JBL 124: 419-450. Cited from the scholarly literature; not read by us.
32. Crossley, J. G., The Date of Mark's Gospel (T&T Clark, 2004). Cited from the scholarly literature; not read by us.
33. Sanders, E. P., Jesus and Judaism (Fortress, 1985), ch. 1. Cited from the scholarly literature; not read by us.
34. Josephus, Jewish War 6.300-305 (Whiston), lexundria.com. Read by us.
35. Schweitzer, A., The Quest of the Historical Jesus (1906; English 1910); Allison, D. C., Jesus of Nazareth: Millenarian Prophet (Fortress, 1998); Ehrman, B. D., Jesus: Apocalyptic Prophet of the New Millennium (OUP, 1999). Cited from the scholarly literature; not read by us.
36. Cross, F. M., Canaanite Myth and Hebrew Epic (Harvard, 1973), ch. 10. Cited from the scholarly literature; not read by us.