How to Test a Fulfilled Prophecy

"Predict how future events will turn out, so we might know you are gods." — Isaiah 41:23, NET¹

"Hosea’s words are not a prediction, but an account of Israel’s origin." — R. T. France, evangelical commentator on Matthew, 1985²

Summary

A claim of fulfilled prophecy says four things at once.  The words were written before the event.  They said something specific enough that the event could have failed to match.  Nobody who knew the words brought the event about on purpose.  And the match was there to be seen, not supplied by a reader looking back.  Each of the four has a natural rival.  The text was written after the event.  It was vague enough to fit many events.  Someone fulfilled it deliberately.  Or a later reader found a pattern and called it a prediction.  A fulfilled prophecy is evidence of foreknowledge only when all four rivals have been ruled out.

The Bible sets the first test itself.  Deuteronomy 18:22 says a prophet whose prediction "is not fulfilled" has spoken presumptuously.¹  Isaiah 41:22–23 asks the idols to "tell us what will happen" and to "predict how future events will turn out, so we might know you are gods."¹  So the question is fair from both sides, and it needs rules fixed before the answers are known.  This article sets out eight, numbered so that the other articles in this section can cite them.

The eight tests.

1. Written before. The words must be shown to exist before the event: by a manuscript, a translation, or a datable quotation older than the event, or by internal evidence that does not assume the answer. A text written after the events it "predicts" is accurate up to the moment of writing and goes wrong after it. That turning point is evidence, and it is not the same thing as assuming that prediction is impossible.
2. Committed in advance. A reader before the event must have been able to say what would count as a failure. Score each element the text commits to: who, what, where, when. Ambiguity, symbols with several readings, and flexible start and end points all lower the score.
3. Read that way before. The text must have been read as a prediction of that kind of event before the event happened. A pattern noticed afterwards, which New Testament scholars call typology, may be good theology, but it is not evidence of foreknowledge.
4. Out of reach. The outcome must lie beyond the power of everyone who knew the text: the claimant, his followers, his enemies, and the writer who reports the fulfilment.
5. Reported independently. The fulfilment must be attested by someone with no stake in the prophecy, or the claim is only as strong as the one interested report.
6. All the misses counted. A hit is weighed against every failed, cancelled, or reinterpreted prediction from the same source, and against the number of texts that were searched to find it. If unfulfilled prophecies are excused as conditional, fulfilled ones must lose weight by the same rule.
7. Priced honestly. A probability may multiply only independent items, must use estimates that can be defended, and must correct for the selection of hits from a larger pool.
8. The same rules for everyone. Any test loose enough to pass Nostradamus or a Babylonian prophecy written after the fact is too loose. Any test strict enough to fail every possible prediction is too strict.

What the tests do to the nonbiblical controls.  The Egyptian Prophecy of Neferti, the Babylonian Dynastic Prophecy, and much of the Sibylline Oracles fail Test 1:  each is accurate up to its own time and no further, and their editors date them by that point.  The Delphic oracle to Croesus fails Tests 1 and 2.  Nostradamus's famous quatrain on the death of Henri II passes Test 1, since it was printed in May 1555, four years before the joust that killed him.³  It fails Tests 2 and 6:  almost every element it commits to does not fit, and it is one of about 940 quatrains.⁴  Josephus's prediction that Vespasian would become emperor is the strongest pagan-era case, and it passes Tests 1 and 5;  its weight is limited by Josephus's motive and by how likely the outcome already was.

What the tests do to Christian apologetics.  The result cuts both ways, and several favourite uses fail.

• Peter Stoner's "1 in 10¹⁷" for eight messianic prophecies fails Test 7. His arithmetic is right, but its inputs are classroom estimates that he said he would make no attempt to defend,⁵ and he never says how the eight were chosen from the "more than three hundred" he counted. All eight also fail Test 3, 4 or 5: the text was not a prediction or was not read as one before Jesus, or the fulfilment lay within reach of people who knew the text, or it is reported only by the Gospels.
• Hosea 11:1, "out of Egypt I called my son," fails Test 3. The leading conservative commentators agree. France: "Hosea’s words are not a prediction."² Walter Kaiser: "it was not a prophecy as was Micah 5:2."⁶ Matthew's use is typology.
• Zechariah 9:9, the king on a donkey, fails Test 4. The Gospels themselves say Jesus sent for the animal.
• The naming of Cyrus in Isaiah 44:28 and of Josiah in 1 Kings 13:2 stand or fall on Test 1, and the manuscripts cannot settle it.
• Ezekiel's oracle against Tyre fails Test 6 as a showpiece. Ezekiel 29:18 itself says Nebuchadnezzar's army "received no wages from Tyre," and the evangelical scholar Robert Chisholm concedes that Nebuchadnezzar "did not devastate the city in the manner described by Ezekiel."⁷

What survives.  Some biblical texts pass Test 1 beyond dispute.  The Great Isaiah Scroll, including Isaiah 53, was copied between 180 and 100 BC.⁸  Daniel 9 predates Jesus by about two centuries even on the latest date any scholar gives the book.  The execution of Jesus under Pontius Pilate is reported by Tacitus, a hostile Roman senator, which passes Test 5 for the bare event.⁹  But passing Test 1 is only the first step.  Isaiah 53 and Daniel 9 still have to be scored on Tests 2, 3 and 6, and they are, in their own articles.

Verdict.  Fulfilled prophecy cannot honestly be argued as a list of hits with a large number at the bottom.  Applied evenly, the tests remove most of the popular list.  What they leave is a short set of texts that were certainly written before the events, that cannot have been arranged by those who knew them, and that must then be judged one at a time on what they actually commit to.  That is a weaker argument than the one usually made.  It is also one that a skeptic cannot dismiss by pointing at Nostradamus.

Part 1: What a prediction claim says, and the Bible's own rules

Most prophecy is not prediction

Much of what the Hebrew prophets said was not forecasting.  It was accusation, warning, and comfort, spoken to a crowd that could still change its ways.  Robert Chisholm, who chairs Old Testament studies at Dallas Theological Seminary, puts the point this way:  prophetic prediction "is often (usually?) dynamic.  It announces God’s intentions conditionally".⁷  The critical scholar Robert Carroll agrees that prophecy "was essentially a preaching for a decision type of activity."¹⁰  The evangelical scholars Gordon Fee and Douglas Stuart, in their guide How to Read the Bible for All Its Worth, warn apologists against reading the prophets as mainly forecasters:  "Less than 2 percent of Old Testament prophecy is messianic.  Less than 5 percent specifically describes the New Covenant age.  Less than 1 percent concerns events yet to come."¹¹  They give no counting rule for those figures, but the warning stands.  So a test of prediction applies only to the part of prophecy that predicts.  It does not tell anyone whether Amos was right about injustice in Samaria.

That narrows the field without emptying it.  Some prophetic texts plainly do predict:  a named king, a city's fall, a span of years.  And the Bible itself makes prediction a proof.  Isaiah's God challenges the idols:  "Tell us about your earlier predictive oracles, so we may examine them and see how they were fulfilled.  Or decree for us some future events!"¹  He claims the opposite for himself:  "I announced them to you beforehand; before they happened, I predicted them for you, so you could never say, ‘My image did these things’".¹  An apologist who appeals to fulfilled prophecy is making the claim Isaiah makes.  A skeptic is entitled to examine it the way Isaiah invites.

Three rules in the Law and the Prophets

The Bible states three rules for judging a prophet's prediction, and each of them carries into the tests below.

1. A failed prediction counts against the prophet. "Whenever a prophet speaks in my name and the prediction is not fulfilled, then I have not spoken it; the prophet has presumed to speak it, so you need not fear him" (Deuteronomy 18:22).¹ Carroll calls this "a straightforward falsification principle allowing the community to distinguish between the prophet with the word of Yahweh and the prophet lacking that word."¹⁰
2. A fulfilled prediction is not enough by itself. Deuteronomy 13:1–3 imagines a prophet whose "sign or wonder should come to pass" and who then says "Let us follow other gods." Israel is told, "You must not listen to the words of that prophet".¹ A hit does not validate a message that contradicts what is already known.
3. Warnings and promises are not judged alike. Jeremiah 18:7–10 says a threatened nation that "stops doing wrong" may see the destruction cancelled, and a promised nation that "does what displeases me" may lose the good.¹ Yet Jeremiah 28:9 adds that a prophet who "prophesied peace and prosperity" was "only known" to be sent by God "when what he prophesied came true."¹

Both Chisholm and Carroll see the same weakness in the first rule.  It can only be applied after the fact.  Carroll says it "put a good deal of emphasis on hindsight in that only by waiting until the prophetic word had come to pass would the community have been able to ascertain the prophet’s authenticity."¹⁰  Chisholm concludes that the test "must apply to short-range prophecies, not prophecies of the distant future."⁷  For a reader today, though, every prophecy is in the past.  The question is no longer whether to wait.  It is whether the waiting was honest:  whether the words really came first, whether they really committed to something, and whether the misses have been counted.  The eight tests below turn those questions into rules.

What a test must not assume

Two shortcuts would make the exercise pointless, and both are common.

The first belongs to some critics.  If prediction is impossible, then every accurate prediction was written after the event, and every text can be dated by what it gets right.  The evangelical commentator John Oswalt names this as the unspoken premise behind many late dates in Isaiah:  "no specific prediction of the future is possible, and whenever such appears to occur (e.g., with the exile to Babylon [39:5—6] or the deliverance by Cyrus [45:1]) these are always either contemporary with, or after, the fact."¹²  A test built on that premise cannot find a prediction, so it cannot be evidence that there are none.

The second belongs to some apologists.  If the Bible is inspired, every fulfilment it reports is real, and every text it calls fulfilled was a prediction.  A test built on that premise cannot find a failure.

So each test below must be one that a biblical text could pass and could fail, and one that a nonbiblical text could pass and could fail.  That is Test 8, and it governs the rest.

Part 2: Test 1, written before the event

Three kinds of evidence for a date

A text can be shown to exist before an event in three ways.

A physical copy.  A manuscript older than the event settles the question for that event.  The Great Isaiah Scroll from Qumran is the standard example.  A 2025 study that combined radiocarbon dates with a machine-learning model of handwriting found that "Enoch dates the two halves consistently between 180–100 BCE."⁸  Whatever Isaiah 53 says about a suffering servant, it said it before Jesus was born and more than a century before his crucifixion.  But a copy only gives an upper limit.  The same scroll contains Isaiah 44:28, which names Cyrus, and a copy made around 150 BC cannot tell anyone whether that verse was written before Cyrus (about 559–530 BC) or during his career.

A translation or a datable quotation.  The grandson of Ben Sira, who arrived in Egypt "in the thirty-eighth year, in the reign of Euergetes the king," that is 132 BC, wrote that the Law, "the Prophets," and "the rest of the books" already existed in Greek.¹³  So the Greek Prophets are a Jewish translation made before Christianity.  Any reading they contain is a pre-Christian reading.

Internal evidence.  A text can betray its own date:  a word that came into the language later, an event it treats as past, a situation it takes for granted.  The most important internal test for prophecy is the turning point.

The turning point

People who wrote history in the form of prophecy have a characteristic pattern.  They are accurate up to their own time, because they are describing what happened.  After that they are guessing, and they usually guess wrong.

The Assyriologist A. K. Grayson, who edited the Babylonian examples, describes the type:  "An Akkadian prophecy is a prose composition consisting in the main of a number of 'predictions' of past events.  It then concludes either with a 'prediction' of phenomena in the writer's day or with a genuine attempt to forecast future events.  The author, in other words, uses vaticinia ex eventu to establish his credibility and then proceeds to his real purpose".¹⁴  Vaticinium ex eventu is Latin for "prophecy from the outcome."

The Dynastic Prophecy, a cuneiform tablet from Babylon, shows the pattern clearly.  It "predicts" a king who will reign two years and be murdered by a eunuch, then a rebel prince who will seize the throne and reign five years.  These are Arses, who "ruled two years and was murdered by the eunuch Bagoas," and Darius III, who "indeed ruled for five years," as Bert van der Spek, who re-edited the tablet, notes.¹⁵  Then troops of "the land of Hani," the Macedonians, invade, and after a break in the text a king helped by the gods of Babylon brings about "the overthrow of the Hanaean troops."¹⁶  Grayson, the first editor, read that king as Darius III, which makes the line the reverse of what happened at Gaugamela in 331 BC:  "For this I have no answer.  It is extremely unlikely that the ‘prophet’ would deliberately falsify the outcome and aftermath of such a famous and well-known battle."¹⁴  Van der Spek states the problem as bluntly, "It is simply not true.  Darius III never defeated Alexander,"¹⁵ and then offers other readings:  a scribal slip, or a genuine forecast of some later saviour, written at any time up to Alexander's death in 323 BC.¹⁵  On every reading the pattern holds.  The tablet is right about reign lengths that could only be known after the reigns ended, and its first unambiguous look ahead either failed or cannot be checked.  That is where the author stood.

The Egyptian Prophecy of Neferti is the same kind of work, and it gives itself away by a name.  A sage at the court of King Snefru of the Fourth Dynasty foretells civil war and then a saviour king.  Miriam Lichtheim, whose translation is standard:  "He calls the redeemer “Ameny”—the short form of Amenemhet—and thereby provides the clue to the understanding of the work.  The readers, ancient and modern, are to understand that the “prophecy” is a literary disguise, veiling the contemporary character of a work composed in the reign of Amenemhet I".¹⁷  A prophecy that names the reigning king, five centuries in advance, is not evidence of foresight.  It is evidence of when it was written.

The Delphic oracle shows the same thing from the side of the reports.  Joseph Fontenrose sorted 535 recorded responses of the oracle by how they are attested.  He called a response "Historical" when it appears in a record written within the lifetime of the person who heard it, and "Quasi-Historical" when it is "first attested by a writer whose lifetime was later than the accepted or supposed date of the response."¹⁸  The difference in content is stark.  "The Historical responses include no ambiguous, unclear, or conditioned commands, no ambiguous prohibitions or warnings, no extraordinary statements about past or present, no ambiguous or conditioned predictions, and only 2 clear predictions".¹⁸  The famous Delphic predictions are all in the later reports.  The closer a record stands to the event, the less prophetic the oracle looks.

Daniel 11 and the turning point

The same test applies to the Bible, and it cuts.  Daniel 11 traces the wars of the Ptolemies and Seleucids in close detail down to Antiochus IV and his desecration of the temple in 167 BC.  Then it says the king of the north will invade Egypt again, "pitch his royal tents between the seas toward the beautiful holy mountain," and "come to his end, with no one to help him."¹  Antiochus IV died at Tabae in Persia after a failed raid on a temple in Elymais.¹⁹  The conservative scholar Joyce Baldwin stated the critics' criterion fairly:  "The author's true standpoint is to be discerned at the moment of transition, when history can no longer be demonstrated to correspond to the events prophesied."²⁰  She granted that "in style and form chapters 7-12 now prove to have Babylonian literary affinities," but denied that Daniel contains disguised history:  "If there are no forged prophecies there is no point in arguing either for a second century or for pseudonymity."²⁰

A physical date now complicates the turning-point argument.  A copy of Daniel 8–11 from Qumran, 4Q114, has a radiocarbon date whose 95% range is either 355–285 BC or 230–160 BC, and whose 68% range puts 34.5 of its 68.3 percentage points in 205–175 BC, before Antiochus IV took the throne.⁸  The editors judge that the scribe made too many errors to be the author.⁸  If the parchment date holds, a copy of Daniel 11 existed before the events of 11:21–35.  The full case on both sides is in Dating the Book of Daniel, and the kingdoms are scored in Daniel 2, 7, and 11.  For this article the point is narrower.  Test 1 can be decided by internal evidence, by physical evidence, or by both, and when they disagree, neither side may quote one and suppress the other.

The burden of a late date

The turning point is evidence.  It does not assume that prediction is impossible.  It would date a genuine prophet's book just as well, by the point where his predictions stopped coming true, if there were such a point.  But not every late date rests on a turning point.

John Oswalt, defending the unity of Isaiah, argues that many critical dates assume their conclusion.  On the oracle against "the king of Babylon" in Isaiah 14:  "So long as the Church accepted the possibility of predictive prophecy, no problem was perceived.  But with the rise of skepticism in that direction, the title was taken to be proof of a postexilic date (e.g., Duhm)."¹²  He also turns the turning-point test against one late date for Isaiah 39.  The scholar Ronald Clements had dated the story of the Babylonian envoys to 598–586 BC, because it foresees the palace treasures carried off but says nothing of the temple being destroyed.  Oswalt replies that the theory "contains the seeds of its own destruction.  If the story of the Babylonian envoys had to be rewritten once to explain how Isaiah’s supposedly inviolable Jerusalem could be forced to surrender to Babylon, how much more it should have been rewritten again a few years later when Jerusalem was utterly destroyed by the Babylonians."¹²  That is a fair use of the test from the conservative side.

Not every critical date is built on the premise Oswalt names.  The case for dating Isaiah 40–48 to the 540s BC, as S. R. Driver put it in 1910, rests on the prophet's standpoint:  "the prophet does not claim foreknowledge of Cyrus, but only of what he will accomplish : he is already "stirred up," and "come"".²¹  Driver concluded that "the prophet's reasoning in c. 41-48 implies that the early successes of Cyrus must have been already historical facts."²¹  Oswalt's answer is that the book makes the Cyrus prediction its proof of God's power:  "If this is an after-the-fact realization, what purpose is served by attempting to make it appear a prediction?"¹²  Both are arguments from the text.  Neither needs, or refutes, the possibility of prediction.

So the rule for Test 1 is symmetrical.  A late date must rest on evidence other than the accuracy of the prediction:  a turning point, a later word, a situation the text takes for granted.  An early date must rest on more than the book's heading.  Where neither side has such evidence, the text is undecided, and an undecided text cannot be offered to a skeptic as a fulfilled prediction.

Two of the most striking predictions in the Bible fall into that undecided class.

• Cyrus by name (Isaiah 44:28, 45:1). The manuscripts cannot decide between the eighth century and the 540s BC, and the internal arguments are the ones just quoted.
• Josiah by name (1 Kings 13:2). A prophet tells Jeroboam I, around 930 BC, that "a son named Josiah will be born to the Davidic dynasty" and will burn bones on the altar at Bethel.¹ Kings as it stands ends with the release of King Jehoiachin "in the first year" of Evil-Merodach of Babylon, which is 562 or 561 BC, about half a century after Josiah died in 609 BC.¹ No copy or quotation shows the verse existed before Josiah. That does not show it was added later. It shows that the verse cannot carry weight with someone who does not already accept the book's date.

Test 1, stated as a rule.  A text passes if a copy, a translation, or a datable quotation older than the event contains the words, or if internal evidence that does not argue from the prediction's accuracy places it before the event.  It fails if it is accurate up to a point and wrong after it, and the event in question falls before the point.  Otherwise it is undecided.

Part 3: Test 2, what the text commits to

Score the elements, not the impression

A prediction is worth something only if it could have failed.  So the second test asks what a reader before the event could have written down as the outcome the text requires.  The working method is a short scoring sheet, filled in before looking at the fulfilment.

1. List the elements. Who acts, what happens, where, when, and how many. Count only what the words say, not what a later reader adds.
2. For each element, ask whether it could have failed. A forecast that a great empire will fall cannot fail if every possible outcome destroys some empire.
3. For each element, ask how likely it was anyway. This is the prior: the chance that an informed person at the time of writing would have guessed it. A sectional war in America was likely in December 1832, which is why the Joseph Smith article scores his Civil War revelation as a reasonable guess even though it came true.
4. Count the free choices. If the text leaves a start date, a unit of time, a referent, or an end point open, every reasonable option counts against the hit, because the interpreter could have chosen another.

A hit can be no stronger than its prior is low.  If an outcome had a one-in-three chance of happening anyway, its fulfilment makes the prophet at most three times more likely to have known than to have guessed.  Only outcomes that were genuinely unlikely, and stated plainly, can carry much weight.

Croesus: a prediction that could not fail

Herodotus tells the classic story.  Croesus of Lydia asked Delphi whether to attack Persia, and the answer was that "if he should march against the Persians he should destroy a great empire."²²  He marched and lost his own.  When he complained, the oracle replied that "if he wished to take counsel well, [he] ought to have sent and asked further whether the god meant his own empire or that of Cyrus".²²  The answer scores zero on Test 2, because no outcome of the war could have falsified it.  It also fails Test 1.  Herodotus wrote about a century after Croesus fell, and Fontenrose classes all the Croesus oracles as Quasi-Historical, reported only after the fact.¹⁸

Nostradamus: a quatrain that fits by its loose parts

The first edition of Nostradamus's Prophéties, finished printing at Lyon on 4 May 1555, contains this as quatrain 35 of the first Century:³

Le lyon ieune le vieux surmontera,
En champ bellique par singulier duelle,
Dans caige d'or les yeux luy creuera:
Deux classes vne, puis mourir, mort cruelle.

James Randi translates it:  "The young lion will overcome the old one, On the field of battle in single combat: He will burst his eyes in a cage of gold, Two fleets one, then to die, a cruel death."⁴  In 1559 King Henri II of France was killed in a jousting accident when a lance splintered into his helmet.⁴  The quatrain is famous as a prediction of that death, and it passes Test 1:  it was in print four years earlier.

Scored element by element, it does much worse.  The quatrain names no king, no country, and no year.  Henri died in a tournament, and Randi notes that "the event at which Henry was killed was not a duel nor was he “on the field of battle” at all.  It was a formal jousting contest in which serious injury was not expected to take place."⁴  His opponent, the Comte de Montgomery, was only a few years younger than the king, and the lion was not a French royal emblem.⁴  The word classes most naturally means "fleets," from Latin classis.  The usual rendering "wounds" comes from the Greek klasis, a derivation Randi calls "convenient but unlikely."⁴  What remains is a wound to the eyes inside something golden, followed by a cruel death.  That is a real resemblance, and it is a small one.  Nostradamus also left a dated forecast for 1559, an almanac for that very year, which does not mention the death of the king.⁴

Daniel 9: the endpoints are part of the claim

Daniel 9:25 counts "seven weeks and sixty-two weeks" from "the issuing of the command to restore and rebuild Jerusalem" until "an anointed one, a prince arrives."¹  Sixty-nine weeks of years are 483 years.  Applied to Jesus, the count needs three choices:  which decree starts it, how long a year is, and which event in Jesus's life ends it.

The choices are not idle.  Robert Anderson, in 1881, started from Nehemiah's commission in 445 BC and counted years of 360 days:  "The seventy weeks are therefore to be computed from the 1st of Nisan B.C. 445," landing on "Sunday the 6th April, A.D. 32."²³  Harold Hoehner of Dallas Theological Seminary rejected both dates:  "the 445 B.C. date is not acceptable for Artaxerxes’ twentieth year; instead the decree was given in Nisan, 444 B.C.," and he moved the end to AD 33 on the same 360-day year.²⁴  Other readers start from Cyrus or from Ezra's commission in 457 BC, and use solar years.

Counting every combination of the four commonly proposed start dates and the two year lengths gives eight landing years, spread from 62 BC to AD 40.  Three of the eight fall in AD 26–36, the span of Jesus's adult ministry and death.²⁵  Two facts follow, and both should be said.  The count lands in the right century from every Persian decree, which is not nothing for a text that predates Jesus by at least two centuries.  But the exact-day fulfilments published for it depend on choices made after the event, and different defenders make different ones.  An honest score fixes the start, the year, and the end before computing, and counts the alternatives that were passed over.  The site's article on Daniel 9 argues for one set of choices;  this test asks only that the choices be stated and priced.

Specificity and dating pull against each other

The most specific predictions in ancient literature are names and exact reign lengths:  Neferti's "Ameny," the Dynastic Prophecy's two-year and five-year kings, Isaiah's Cyrus, the prophet's Josiah in 1 Kings 13.  Specificity is exactly what makes a prediction valuable, and it is also exactly what writers after the event supply.  So a high score on Test 2 makes Test 1 decisive.  A vague text that passes Test 1 proves little;  a precise text that fails Test 1 proves nothing;  only a precise text that passes Test 1 is strong evidence.

Test 2, stated as a rule.  List what the text commits to before looking at the fulfilment.  Score each element for whether it could have failed and how likely it was anyway.  Count every free choice of start, unit, referent, and end, and treat the alternatives as part of the claim.

Part 4: Test 3, read that way before the event

Prediction or pattern?

The third test asks whether anyone read the text as a prediction of this kind of event before the event happened.  If no one did, the "fulfilment" was not foretold and then observed.  It was observed and then matched to a text.  That can still be a true insight about God's ways in history.  It is not evidence that anyone knew the future.

The New Testament uses "fulfil" in both senses, and conservative scholars say so.  The clearest case is Matthew 2:15.  After Joseph takes Mary and the child to Egypt and stays "until Herod died," Matthew writes:  "In this way what was spoken by the Lord through the prophet was fulfilled: “I called my Son out of Egypt.”"¹  The prophet is Hosea, and Hosea's sentence is about the past:  "When Israel was a young man, I loved him like a son, and I summoned my son out of Egypt."¹

Four evangelical scholars, writing across nearly three decades, treat Matthew's use as typology rather than as the fulfilment of a prediction.

• R. T. France: "Hosea’s words are not a prediction, but an account of Israel’s origin. Matthew’s quotation thus depends for its validity on the recognition of Jesus as the true Israel".²
• Walter Kaiser, a leading defender of reading the Old Testament by its author's own meaning: "Did Hosea, you ask finally, intend to give a prophecy when he penned the words ‘Out of Egypt have I called My Son’? No, it was not a prophecy as was Micah 5:2; but on the other hand, it was not merely a historical reminiscence either."⁶ He calls Matthew's use "biblical typology at its best."⁶
• Craig Blomberg: "Hosea 11:1 is a reference to the exodus, pure and simple."²⁶ He classes Matthew's use as "a classic example of pure typology," and adds that "the original event need not have been intentionally viewed as forward-looking by the OT author".²⁶
• G. K. Beale pushes furthest the other way. He argues that "what Matthew sees was already something seen to some degree by Hosea himself," because Hosea 11:10–11 looks to a future exodus, and "the past exodus is seen to foreshadow a later end-time exodus, which is a typological understanding."²⁷

Beale's case is the strongest one for a forward-looking Hosea, and even on his reading what Hosea anticipates is a return of the nation, not a child brought back from Egypt.  None of the four calls Hosea 11:1 a prediction of Jesus.  France defines typology as "the recognition of a correspondence between New and Old Testament events, based on a conviction of the unchanging character of the principles of God's working."²  Recognition happens afterwards.  So typology fails Test 3 by definition.  That is not a criticism of Matthew, who never claimed to be offering a probability argument.  It is a criticism of any apologist who counts Matthew 2:15 as a fulfilled prediction.

Matthew 2 contains two more of these.  Jeremiah 31:15, Rachel weeping for her children, is applied to Herod's massacre in Bethlehem.  Blomberg:  "The text in Jeremiah is not a prediction, nor does it even use the future tense."²⁶  Matthew 2:23 says Jesus's move to Nazareth fulfilled what "had been spoken by the prophets," that "Jesus would be called a Nazarene."¹  Blomberg again:  "But no OT text ever declares that anyone will be called a Nazarene!"²⁶

Texts that were read as predictions beforehand

Some texts pass this test, at least in part.  Micah 5:2 says a ruler of Israel will come from "Bethlehem Ephrathah."¹  Two Gospels report that this was the expected birthplace of the messiah before anyone connected it with Jesus.  In Matthew the chief priests and scribes answer Herod's question with it.  In John it is raised as an objection to Jesus by people who think he comes from Galilee:  "Don’t the scriptures say that the Christ is a descendant of David and comes from Bethlehem, the village where David lived?"¹  Both reports are Christian, but John's has the shape of an objection that the Christians had to answer, not a proof they invented.  Still, both were written after Jesus's life, and no Jewish text from before him cites the verse, so Micah 5:2 passes only in part.  The text, its pre-Christian setting, and the Jewish reading of it belong to the article on Messianic Birth Texts.  For the method, the point is that Micah 5:2 is a different kind of claim from Hosea 11:1.  It was a prediction before it was a fulfilment.

Isaiah 53 does not pass.  The Aramaic Targum opens the passage with "Behold, my servant the messiah shall prosper",²⁸ and then moves the suffering onto other people, line by line.  But the Targum cannot be dated before Jesus, so it shows an early Jewish messianic reading, not one made before the event.  Whether any Jew before Jesus expected a messiah who would suffer and die is the question argued in Isaiah 53 and Second Temple Messianic Expectations.

One oracle, three readers

A text can pass Test 3 and still leave its referent open.  Josephus says that what most drove the Jews to war with Rome "was an ambiguous oracle that was also found in their sacred writings, how, "about that time, one from their country should become governor of the habitable earth." The Jews took this prediction to belong to themselves in particular, and many of the wise men were thereby deceived in their determination. Now this oracle certainly denoted the government of Vespasian, who was appointed emperor in Judea."²⁹  Tacitus reports the same belief and gives the same verdict:  "This mysterious prophecy had in reality pointed to Vespasian and Titus, but the common people, as is the way of human ambition, interpreted these great destinies in their own favour".³⁰  Suetonius says the prediction referred "to the emperor of Rome, as afterwards appeared from the event."³¹

So one text was read before the war as a promise of a Jewish world ruler, after the war by three Roman-era writers as a prediction of Vespasian.  Each reader fixed the referent after the fact.  An oracle that fits whoever wins was read that way beforehand, and it still tells nothing until someone says in advance whom it means.

Test 3, stated as a rule.  A text passes if it was read, before the event, as a prediction of that kind of event, and better still of that referent.  It fails if its predictive reading first appears after the event.  Typology, however illuminating, is a reading after the event and does not count as prediction.

Part 5: Tests 4 and 5, out of reach and reported independently

Who could have made it happen?

A prediction can come true because someone who knew it made it come true.  Test 4 asks who knew the text and what was in their power.  There are four parties to check:  the claimant, his followers, his enemies, and the writer who reports the fulfilment.

The claimant.  Zechariah 9:9 pictures Zion's king "riding on a donkey— on a young donkey, the foal of a female donkey."¹  Mark says Jesus sent two disciples to a village where they would find "a colt tied there that has never been ridden.  Untie it and bring it here."¹  E. P. Sanders, a leading critical historian of Jesus, thinks the event happened and was deliberate:  "I incline to the view that it was Jesus himself who read the prophecy and decided to fulfil it: that here he implicitly declared himself to be ‘king’."³²  That is not a criticism of the Gospels, which say the same thing.  John reports that on the cross Jesus said "I am thirsty!" "in order to fulfill the scripture."¹  A man who knows the Scriptures and believes they speak of him can act them out.  Every such act fails Test 4.  It is evidence of what Jesus believed about himself, which is a different and important question.

The followers.  Jesus was not the only one.  Sanders notes that "first-century Jews sometimes intentionally modelled their own actions on those of biblical figures," and that "in the decade or so after Jesus, other prophets intentionally acted in ways that recalled biblical stories of events that had taken place centuries before."³²  Josephus describes two of them, and Acts names both.¹  Theudas "told them he was a prophet, and that he would, by his own command, divide the river, and afford them an easy passage over it," like Joshua at the Jordan.  The Roman governor's cavalry killed many of his followers, and Theudas was beheaded.³³  A prophet from Egypt led a crowd to the Mount of Olives and promised "that he would show them from hence how, at his command, the walls of Jerusalem would fall down," like Jericho's.  The governor Felix attacked;  four hundred died, and the Egyptian vanished.³³  These men did what was in their power:  they gathered a crowd at the right place.  What was not in their power did not happen.  That is exactly the line Test 4 draws.

The rulers.  Josephus says that Cyrus released the Jews because he had read his own name in Isaiah, and that "an earnest desire and ambition seized upon him to fulfill what was so written."³³  The Cyrus Cylinder gives the conqueror's own account, and it credits a different god:  Marduk "took a righteous king, his favorite, by the hand, he called out his name: Cyrus, king of Anšan".³⁴  If Josephus is right, the decree to rebuild Jerusalem was a fulfilment arranged by a reader of the prophecy.  If he is wrong, he still shows that an ancient Jew saw nothing odd in a king fulfilling a prophecy on purpose.  Either way the decree itself fails Test 4.  Only the naming of Cyrus could count, and that depends on Test 1.

The narrator.  The last party is the easiest to miss.  When the only report of a fulfilment comes from someone who believes the prophecy, the report may have been shaped to fit it.  Sanders, again:  "They believed that Jesus really did fulfil the promises of Hebrew scripture.  If he did so in one case, presumably he did so in another.  There were some genuine overlaps, and this made it very easy for the early Christians to add new facts derived from Jewish scripture."³²  Of the donkey he writes:  "It is possible to think either that the prophecy created the event or that the prophecy created the story and that the event never occurred."³²  Matthew's version of the entry shows the pull of the text.  Zechariah's Hebrew poetry names the animal twice, "on a donkey— on a young donkey."  Mark has one colt.  Matthew has the disciples bring "the donkey and the colt," and Jesus "sat on them."¹

Who else saw it happen?

Test 5 follows from the last point.  A fulfilment reported only by the people who hold the prophecy is only as strong as their report.  A fulfilment reported by someone with no stake in it is much stronger.

The crucifixion is the clearest biblical case that passes.  Tacitus, a Roman senator who despised the new movement, wrote that "Christus, from whom the name had its origin, suffered the extreme penalty during the reign of Tiberius at the hands of one of our procurators, Pontius Pilatus".⁹  The followers did not choose this, and it was the opposite of what they expected of a messiah.  But Test 2 still applies to it.  Rome crucified rebels routinely.  When the governor Varus put down the revolts that followed Herod's death in 4 BC, Josephus says, "the number of those that were crucified on this account were two thousand."³³  For a man executed by Rome as a would-be king of the Jews, crucifixion was the expected method, not an improbable detail.

Most of the details in the passion narrative have no independent witness:  the thirty pieces of silver, the potter's field, the lots for the clothing, the vinegar.  That does not make them false.  It means they rest on Test 4's last party, the narrator, and cannot add weight to a probability argument.

Josephus's own prophecy shows that independent reports can exist even for a pagan-era prediction.  He says that at Jotapata in AD 67 he told the Roman general, "Thou, O Vespasian, art Caesar and emperor, thou, and this thy son."²⁹  Vespasian was acclaimed emperor in the summer of AD 69.  Suetonius tells the story independently:  a prisoner, "Josephus by name, as he was being put in chains, declared most confidently that he would soon be released by the same man, who would then, however, be emperor."³¹  Cassius Dio has him say, "You may imprison me now, but a year from now, when you have become emperor, you will release me."³⁵  It passes Tests 1 and 5 better than most biblical claims.  It still has limits.  Josephus made it to save his life, as Vespasian himself suspected,²⁹ and the historian Tessa Rajak notes that the outcome was not unthinkable in 67, given the conspiracy against Nero in 65 and Vespasian's growing eastern command, though "rational calculation can hardly have led a man ... to conclude that Nero would be toppled and eventually replaced by none other than Vespasian."³⁶  Scored honestly, it is a striking hit by a man with a motive, on an outcome with a low but not tiny prior.  Josephus also claimed other successes, such as having foretold that Jotapata would fall after forty-seven days.³⁶

Test 4, stated as a rule.  List everyone who knew the text:  the claimant, his followers, his enemies, and the narrator.  Any element within their power to bring about, or to report as having happened, does not count as foreknowledge.

Test 5, stated as a rule.  An element counts fully only when someone with no stake in the prophecy reports that it happened.  Elements reported only by believers in the prophecy carry, at most, the weight of that report.

Part 6: Test 6, all the misses counted

The Bible records its own unfulfilled prophecies

A fair test starts by counting what did not happen, and the Bible itself does some of that counting.

• Nineveh. Jonah announced, "At the end of 40 days, Nineveh will be overthrown!" The city repented, and "God relented concerning the judgment he had threatened them with and did not destroy them."¹ Jonah complains that he expected exactly this, because God is "one who relents concerning threatened judgment."
• Jerusalem under Hezekiah. Micah prophesied that "Zion will be plowed up like a field, Jerusalem will become a heap of ruins."¹ A century later the elders of Judah recalled the oracle and explained why it had not come true in Hezekiah's day: "Did not the Lord forgo destroying them as he threatened he would?"¹
• Josiah's peaceful death. The prophetess Huldah told King Josiah, "I will allow you to die and be buried in peace."¹ Josiah was killed by Pharaoh Necho at Megiddo.¹ Robert Chisholm of Dallas Theological Seminary grants that if the oracle is read as unconditional, "we are faced with a problem and probably forced to conclude, with Cogan and Tadmor, “these words of Huldah remain a striking example of unfulfilled prophecy.”"⁷
• Tyre and Egypt. Ezekiel said Nebuchadnezzar would bring "many nations" against Tyre, "scrape her soil from her and make her a bare rock," and that Tyre would "never be built again."¹ In a later oracle, dated to 571 BC, Ezekiel himself records that Nebuchadnezzar's army "received no wages from Tyre for the work he carried out against it," and promises him Egypt instead.¹ Chisholm: "Nebuchadnezzar did indeed besiege Tyre from 585–572 bc, but he did not devastate the city in the manner described by Ezekiel, a fact that is acknowledged in a prophecy delivered in 571 bc (29:17–18)." Of the promised conquest of Egypt he adds, "Even a loosely essential fulfillment is difficult to detect."⁷ He rejects the common apologetic rescue, that Alexander finished the job in 332 BC, calling it "unconvincing."⁷ Did Ezekiel's Prophecy Against Tyre Come True? weighs the case in full.

Chisholm opens his article with a sentence many apologists would not write:  "A close analysis of OT prophecy reveals that many prophecies were not fulfilled either in part or in whole."⁷  His explanation is that prophetic prediction is usually conditional, even when no condition is stated:  "Sometimes conditions are explicitly stated (e.g. Isa 1:19–20), but more often they are unstated and implicit."⁷  Kenneth Kitchen makes a related distinction between "options and predictions":  "Options are conditional; “if you do this, then such and such will happen” (but not if you don’t).  Predictions are meant to stick — this is going to happen — full stop."³⁷

The critical scholar Robert Carroll read the same evidence the other way.  His book on the subject is "an attempt to examine the prophetic traditions in terms of their predictive elements and the responses to the failure of the expectations aroused by them."¹⁰  On his reading, the language of conditions and delays is how believers coped after a prediction failed, much as the modern groups studied by social psychologists did.  Chisholm and Carroll agree on more than either side usually admits:  some oracles did not come true as worded.  They disagree about whether the conditions were part of the prophets' meaning from the start.

The symmetry rule

Conditional prophecy is a real biblical category, and Jeremiah 18:7–10 states it as a general rule.¹  A test that ignored it would fail Jonah, whom the Bible treats as a true prophet.  But the category has a price, and the price must be paid on both sides of the ledger.

Suppose a prophet's doom oracles come true unless the city repents, and suppose the chance that God relents is r.  Suppose a shrewd human forecaster's doom oracles come true with chance q.  Then a fulfilled oracle favours the prophet by a factor of (1 − r)/q, and an unfulfilled one by r/(1 − q).  When r is zero, a miss refutes the prophet and a hit is worth 1/q.  As r rises, misses stop counting against him, and hits count for less.  If the forecaster's chance was one in two and God relents half the time, a hit and a miss are both worth exactly nothing.³⁸

So conditionality cannot be invoked only after a failure.  It counts as a rescue when the text states the condition, or when the record shows the condition was met:  Nineveh's fast and sackcloth in Jonah 3, Hezekiah's repentance remembered in Jeremiah 26.  Where no condition is stated and no repentance is recorded, as with Tyre, a reader may still believe the oracle was conditional.  But then the oracle cannot also be scored as a fulfilled prediction, by Alexander or anyone else.  The ledger must use one rule for both columns.

Count the pool, not just the hits

The second half of Test 6 concerns selection.  A list of hits says nothing until it is compared with the number of texts that could have been searched for hits.  Peter Stoner, whose probability argument is priced in Part 7, wrote that "more than three hundred prophecies from the Old Testament which deal with the first advent of Christ have been listed," and built his calculation on eight of them, with no account of how the eight were chosen.³⁹  The conservative scholar J. Barton Payne, who catalogued every prediction in the Bible, counted 1,817 predictions in 8,352 of the Bible's 31,124 verses.⁴⁰  Nostradamus published about 940 quatrains.⁴  The Joseph Smith article scores twenty-nine items fixed by a rule before scoring began, and finds six fulfilled.

The arithmetic of selection is simple.  If a searcher has a thousand candidate texts, and each has a one-in-a-hundred chance of being matched loosely by something in a given life, the searcher should expect ten matches, and the chance of finding at least eight is about 78%.⁴¹  Loose matches in a large pool are not evidence.  Matches are evidence only when each is specific (Test 2), the pool is small or fixed in advance, and the misses from the same pool are reported.

Test 6, stated as a rule.  Score every checkable prediction from the source, chosen by a rule fixed before scoring, not only the famous ones.  Count conditional rescues only where the condition is stated or its fulfilment recorded, and when a condition is invoked to excuse a miss, apply the same discount to the hits.  Report the size of the pool that was searched.

Part 7: Test 7, probability claims priced

Stoner's argument, in his words

The best-known number in Christian apologetics comes from Peter Stoner, who chaired the departments of mathematics and astronomy at Pasadena City College until 1953³⁹ and published Science Speaks in 1952, with revised editions through 1976.  His method was to ask college classes how likely each prophecy would be to come true in one man by chance.  "The students were asked to be very conservative in their probability estimates.  They discussed each prophecy at length, bringing out various conditions which might affect the probability of any man fulfilling it."³⁹  He then adjusted their figures and multiplied them.  For eight prophecies about Christ the product was "1 in 2.8 x 10²⁸," which he rounded to 10²⁸.³⁹  He divided by the number of people who had ever lived, "about 88 billion or 8.8 x 10¹⁰," rounded to 10¹¹, and concluded that "the chance that any man might have lived down to the present time and fulfilled all eight prophecies is 1 in 10¹⁷."³⁹  To picture it, he spread 10¹⁷ silver dollars over Texas:  "They will cover all of the state two feet deep."³⁹  For forty-eight prophecies he reached 1 in 10¹⁵⁷.³⁹

The book carried a foreword from H. Harold Hartzler of the American Scientific Affiliation, saying that "the mathematical analysis included is based upon principles of probability which are thoroughly sound."⁵  Josh McDowell's Evidence That Demands a Verdict reproduced the figures in 1972, "to show that coincidence is ruled out by the science of probability," and again in The New Evidence That Demands a Verdict in 1999.^{42 43}

Two common criticisms of Stoner are unfair, and should be dropped.  The arithmetic is right:  2.8 × 10²⁸ divided by 8.8 × 10¹⁰ is 3.2 × 10¹⁷, and Stoner's rounding works against him, not for him.⁴⁴  And he did not simply multiply linked events as if they were independent.  He wrote that "if the events are not entirely independent, a somewhat different theory of probability must be used,"³⁹ and he framed the linked items conditionally:  "Of the people who have been betrayed, one in how many has been betrayed for exactly thirty pieces of silver?"³⁹

Where the number fails

The number fails on its inputs, and Stoner was candid about them.  "No attempt will be made, however, to defend these estimates."⁵  Run the eight through the earlier tests and the inputs fall away.

All eight fail Test 3, 4 or 5 for a reader who does not already accept the Gospels' details.⁴⁵  The texts are in the NET.^{1 1}  The crucifixion comes closest.  It fails only Test 3, and it is a real and independently attested event, but for a man condemned by Rome as a would-be king of the Jews it was the expected outcome, and the word "pierced" in Psalm 22:16 is a modern rendering:  the Hebrew reads "like a lion," the Greek "they dug," and the New Testament never quotes the verse.  The full case is in Psalm 22.

Then there is the selection.  Stoner himself said that "more than three hundred prophecies" had been listed, and he chose eight without saying how.³⁹  The question Stoner asked was how likely one man is to fit eight texts that were chosen because they fit him.  The question that matters is how likely it is that some eight fits would turn up when hundreds or thousands of short Old Testament phrases are compared with a life.⁴¹  Selection does not make a specific match likely.  It makes loose matches inevitable.

Even the author's family has stepped back.  Stoner's grandson Don, who maintains the online edition, added a note in 2022:  the numerator in each estimate "was usually presumed to be "1" in my Grandfather's book (as in my "Challenge" above), but this is no longer a satisfactory assumption."  He adds:  "At this point in time, Science Speaks is more likely to be weighed for its historical significance than its usefulness as an evangelical tool."³⁹

Larger numbers, same method

Hugh Ross of Reasons to Believe made the argument bigger.  His 2003 article began:  "Approximately 2500 prophecies appear in the pages of the Bible, about 2000 of which already have been fulfilled to the letter—no errors."  Because each had a chance of fulfilment "less than one in ten" and the prophecies were "for the most part independent of one another," the odds against chance were "less than one in 10²⁰⁰⁰."⁴⁶  His estimates, he wrote, "come from a group of secular research scientists," and his worked example shows how:  "Since the Messiah is God in human form, the possibility of his being killed is considerably low, say less than one chance in 10."⁴⁶  That estimate assumes the conclusion it is meant to support.  The 2026 revision of the same article says "a couple thousand prophecies" and "a number far greater than ~ 1 in 10¹⁰⁰⁰."⁴⁷  Ross and Stoner give very different odds for the same events, Bethlehem at 1 in 10⁴ against Stoner's 1 in 2.8 × 10⁵, and neither explains the difference.⁴⁸  Estimates that move by orders of magnitude between two careful authors are not measurements.

Critics have made the same points from the other side.  Jim Lippard's essay on the Secular Web notes that Stoner's class estimates "did not consider any of the above objections to these prophecies, nor did they consider the possibility of intentional fulfillment.  (For example, a Messiah claimant might hire a John-the-Baptist-style messenger to precede him, or intentionally ride a donkey into the city of Jerusalem.)"⁴⁹  Critics also make errors:  Lippard gives Stoner's figure for the eight as 1 in 10²¹, a number that in the printed editions belongs to a different passage.⁴⁹

What an honest number would need

A probability argument from prophecy is legitimate in principle.  It is the formal version of Isaiah's challenge.  To be honest it needs five things.

1. Only items that pass Tests 1 to 5. Every other item is either not a prediction, or not out of reach, or not independently known to have happened.
2. Priors that can be defended, from data where possible: how many towns a Davidic claimant might come from, how often Rome crucified rebels, how often messianic claimants were executed.
3. Conditional estimates for linked items, as Stoner tried to do.
4. A correction for selection. The pool of candidate texts and the pool of candidate fulfilments both enter the calculation.
5. A comparison, not a lone number. The result is a ratio: how much more likely the evidence is if the author knew the future than if he did not. It must then be weighed against everything else the reader knows.

Built that way, Stoner's eight do not yield 10¹⁷.  For a skeptic they yield little more than the crucifixion of a messianic claimant, which is weak evidence of prediction.  For a reader who already trusts the Gospels' details, items such as Bethlehem and the thirty pieces of silver can add weight, but then the argument rests on the Gospels' reliability, which is a different argument and has to be made on its own terms.  The strongest prophetic arguments are not lists.  They are single texts, like Isaiah 53 or Daniel 9, that pass Test 1 beyond dispute and must then be judged on what they actually commit to.

Test 7, stated as a rule.  Multiply only independent or properly conditioned items.  Use estimates that can be defended from data, not guesses.  Include only items that pass Tests 1 to 5.  Correct for the number of texts and lives that were searched.  Report a ratio, not a lone number.

Part 8: Test 8, the same rules for everyone

Christians have been fooled too

The tests in this article were not invented to embarrass the Bible, and the clearest proof is that early Christians failed them with a pagan text.  The eighth book of the Sibylline Oracles contains a Greek poem whose first letters spell out "Jesus Christ, Son of God, Saviour, Cross."  The emperor Constantine, in his Oration to the Assembly of the Saints, presented it as a pagan prophetess's prediction of Christ, and he answered the obvious objection.  Critics, he said, "hold, in short, that they are a forgery."  His reply was a dating argument:  "there is no room to suspect that this poem was composed after the advent and condemnation of Christ," because "Cicero was acquainted with this poem."⁵⁰  Augustine quoted the acrostic too.⁵¹

Modern scholarship has settled the question against them.  John J. Collins, editor of the Sibylline Oracles in the standard English collection, dates the second half of Book 8 only by the fact that Lactantius quotes it in the early fourth century, and explains the acrostic:  "The use of acrostics here reflects the desire of the Christian author to present his work as a gentile oracle."⁵²  The rest of the Sibylline corpus is dated the way Test 1 dates any text.  Book 4 "refers to the destruction of the Jerusalem Temple (116), the legend of Nero's flight to the Parthians and future return (119-24, 138-39), and the eruption of Vesuvius in A.D. 79 (130-35).  All scholars agree that it was written shortly after the last datable event mentioned—therefore about A.D. 80."⁵²

So the tests cut against apologetic history as well as against apologetic arguments.  That is what makes them usable.  A test that only ever removed pagan claims would be a weapon, not a test.

The scorecard

The table applies the tests to the cases discussed in this article.  "Pass" and "Fail" are for a reader who does not already accept the source's authority.  A dash means the test does not bear on the case or was not scored here.

The pattern is plain.  Every nonbiblical case fails Test 1 or Test 2 except two:  Josephus's prediction, a hit with a motive and a modest prior, and Joseph Smith's Civil War revelation, a reasonable guess in the month it was made.  Several popular biblical cases fail too, and they should be retired:  Hosea 11:1 as prediction, the donkey as foreknowledge, Psalm 22's "pierced," Tyre as a showpiece, and Stoner's number.  What remains is a small set of biblical texts that pass Test 1 beyond dispute, that no follower could have arranged, and whose fulfilments are attested outside the church.  Those texts are the ones worth arguing about, and each needs its own article.

Test 8, stated as a rule.  Before using a test, apply it to a nonbiblical text it should reject and to a biblical text it should accept if the Bible's claim is true.  If it passes the first or cannot pass the second, change the test, not the result.

Objections and responses

1. "These tests are built to reject every prophecy." They are built to reject explanations other than foreknowledge, and they do not always succeed in doing so. Josephus's prediction to Vespasian passes Tests 1, 2 and 5. Isaiah 53 and Daniel 9 pass Test 1 beyond dispute. The tests also reject the skeptic's favourite shortcut, dating a text late because it is accurate, as Part 2 shows with Oswalt's charge. A test that cannot be passed is rejected by Test 8.
2. "Typology is a legitimate kind of fulfilment. The New Testament writers never claimed to predict anything in Hosea." Agreed, and that is the point. France, Kaiser, Blomberg and Beale all describe Matthew 2:15 as typology. The error belongs to the apologist who lists it among fulfilled predictions, not to Matthew.
3. "Conditional prophecy makes the Bible unfalsifiable." Only if it is invoked after the fact and on one side of the ledger. The symmetry rule in Part 6 allows a condition only where the text states it or the record shows it was met, and it discounts hits by the same rule. Conservative scholars already concede the cases that fail on this basis: Chisholm on Tyre, Egypt and Huldah.
4. "Any prophecy could have been written after the event. You can never prove a date." For events after the oldest copy, a manuscript proves it. The Great Isaiah Scroll was copied by about 100 BC.⁸ Anything it says about a figure who died in the AD 30s was written before he was born. The objection has force only for events that fall before the oldest copy, such as Cyrus, and there the article grants it.
5. "The evangelists invented the fulfilments." The tests do not assume invention. Test 5 says only that a detail reported by believers alone cannot carry weight in a probability argument. Some fulfilments are attested outside the church, including the execution under Pilate. And invention has limits that critics themselves recognise: Sanders thinks the entry on the donkey happened, and was Jesus's own act.³²
6. "Stoner was a scientist, and the American Scientific Affiliation endorsed his mathematics." His mathematics was sound. His inputs were not, and he said he would make no attempt to defend them.⁵ His grandson now says the book is "more likely to be weighed for its historical significance than its usefulness as an evangelical tool."³⁹
7. "Ezekiel's prophecy against Tyre was fulfilled by Alexander, who scraped the old city into the sea." Ezekiel 26:7–11 names Nebuchadnezzar, and Ezekiel 29:18 records that his siege won nothing. Chisholm, an evangelical, rejects the Alexander reading as "unconvincing."⁷ And a reader who explains Nebuchadnezzar's failure as a cancelled condition cannot then score the same oracle as fulfilled. The details belong to Prophecies About Nations and Cities.
8. "Deuteronomy 18:22 says one failed prediction makes a false prophet. So Jonah, Huldah and Ezekiel were false prophets." The Bible itself qualifies the rule: Jeremiah 18 states the condition in general, Jonah 4:2 states it for Nineveh, and Jeremiah 26:19 records it for Micah. Chisholm argues that the Deuteronomy test applied to short-range predictions.⁷ The cost of that answer is the symmetry rule. The Joseph Smith article uses the same pattern test for the same reason.
9. "Daniel 9 lands on the very day of the triumphal entry." It does on one set of choices. Anderson's choices give Sunday 6 April AD 32; Hoehner, who rejected Anderson's start date, reached AD 33 from 444 BC.^{23 24} Test 2 asks that the start, the year and the end be fixed before computing, and that the alternatives be counted. It also grants that every Persian decree lands within about a century of Jesus, which is a real feature of a text that predates him by at least two centuries.
10. "Nostradamus is a straw man. No Christian compares the Bible to him." Skeptics do, and the comparison is fair only if the same tests are applied to both. Applied evenly, Nostradamus's famous quatrain passes Test 1 and fails Tests 2 and 6. Some biblical favourites fail the same tests, and some do not. That difference is the only answer to the comparison that a skeptic is obliged to accept.

Issues and unanswered questions

• Priors are rarely computable. Test 2 asks how likely an outcome was to an informed contemporary. For a few cases there is data, such as Rome's routine crucifixion of rebels. For most, such as the chance that a Davidic claimant came from Bethlehem, the prior is a judgment. The tests make the judgment explicit; they do not remove it.
• Pre-Christian readings of Daniel 9 were not assessed here. Test 3 for the seventy weeks depends on whether any Jewish reader before Jesus expected an anointed one at the end of 483 years. That belongs to the Daniel 9 and Second Temple articles.
• The radiocarbon date of 4Q114 dates the parchment. The scribe wrote on it later, by an unknown interval, and the study found that its radiocarbon results "indicate older date ranges for individual manuscripts" than palaeography had suggested.⁸ How much weight one sample can bear against a clear turning point is argued in Dating the Book of Daniel.
• The Roman reports of Josephus's prediction may not be independent of him. Suetonius and Dio wrote after Josephus's War was published. If they drew on it, Test 5 is weaker than the scorecard shows.
• Ancient people tested prophecy before acting, not after. At Mari a prophet's message was "often controlled by extispicy," with the prophet's hair and garment fringe sent along to represent him.⁵³ That checked the source of a message, not its outcome. Apart from Deuteronomy 18 and Jeremiah 28, no ancient text read for this article scores prophecies against what happened.
• Isaiah 53 and Daniel 9 are not scored here. They pass Test 1, and their scores on Tests 2, 3 and 6 are argued in their own articles. This article supplies the tests; it does not pre-judge those results.

Verdict

A fulfilled prophecy is evidence that someone knew the future only when the words were demonstrably written first, committed to an outcome that could have failed, were read as a prediction before the event, lay beyond the reach of everyone who knew them, are reported by someone without a stake in them, and survive a fair count of the misses.  Probability figures add nothing unless they are built from items that pass those tests.  Applied the same way to every text, the tests do what a skeptic would expect to Nostradamus, the Delphic oracle, and the Babylonian and Egyptian prophecies written after the fact.  They also remove much of the popular Christian case:  Stoner's 10¹⁷, Hosea 11:1 as prediction, the donkey as foreknowledge, Psalm 22's "pierced," and Tyre as a showpiece.  Conservative scholars have already conceded most of these points in print.

What survives is narrower and harder to dismiss.  Some biblical texts were certainly written centuries before the events Christians claim they foretold, and some of those events were outside any follower's control and are attested by hostile witnesses.  Whether those texts actually commit to the events claimed is the real question, and it cannot be answered by a list.  It has to be answered one text at a time, by the tests above, in the articles that follow.

Sources