Messianic prophecy: a whole-text test against cherry-picking

The short version

1. The famous odds come from cherry-picking.

   Stoner’s “1 in 10¹⁷” multiplies eight hand-picked hits. On our blind list, scoring only the hits gives 10¹²; scoring everything gives about 0.3.

   The evidence ↓
2. So we ran a fair test, blind.

   A protocol fixed in advance; every feature Jewish sources before AD 70 expected of the coming deliverer (82 of them), pulled out and scored by separate readers who were not told the purpose; 15 candidates scored the same way. A perfect match would have scored about 10⁶⁵.

   The evidence ↓
3. Jesus fits best, and misses every military and political feature.

   23 hits and 59 misses, 0 of 19 on war, rule and liberation. He ranks first of 12 claimants, but no higher than the best of 12 ordinary claimants with his records would by chance.

   The evidence ↓
4. Scored whole: no margin, and not evidence against.

   A likelihood ratio of about 0.3 (0.07 to 0.4), at most about 2; the test needed 100. Isaiah 53, scored whole, adds about 1 to 2. Daniel 9 gives a window, not a date.

   The evidence ↓
5. It turns on one number nobody has measured.

   How often followers report a hoped-for feature as fulfilled when it was not. The test reaches 100 only if that happened for fewer than about 1 in 25 claims nobody could check. The weight of the case lies in the events, on the resurrection page.

   The evidence ↓

The question

Do the Hebrew Scriptures carry a predictive signal about Jesus that is stronger than what a skeptic would expect anyway? A skeptic expects some matches from three ordinary sources: (a) mining a large text for anything that fits, (b) a messianic claimant who knows the texts and acts on them, and (c) followers who tell his story in the texts' own words.

We compare two hypotheses.

• Foreknowledge (H_T): some passages were written with knowledge of the person who would be Jesus. It predicts that among the features the texts led readers to expect, Jesus has far more than chance, including features no one could arrange, and that he outranks other candidates on the same scoring by more than chance allows.
• The cherry-picking null (H₀): the texts are human. Jesus was one of many claimants; some features were enacted on purpose, some reported or shaped by narrators who knew the texts, and the list of “fulfilled prophecies” was assembled afterwards from the hits. It predicts that hits cluster in features that are easy to stage or to report, misses cluster in public outcomes no one controls, Christian lists leave the misses out, and rivals fit other strands just as well.

A null hypothesis is the boring explanation a test has to beat. The result is reported as a likelihood ratio (LR): how much more probable the evidence is if H_T is true than if H₀ is. An LR of 1 is no evidence; 10 means the evidence is ten times likelier under foreknowledge; 0.3 means it is a little likelier under the null. We often give its base-10 logarithm: +1 means 10, 0 means 1, −1 means 1/10.

One honest complication. Christians answer the misses with two comings: the political features belong to a second advent. That answer is coherent, but it is chosen after the fact. The model allows it and charges for it (see the rules below). The strict Jewish reading, in which one appearance fulfils everything, would make any miss fatal; we do not adopt it either.

Rules of evidence

These were written down in a protocol before any new data were extracted or any number computed. Changes after the run were allowed only as logged corrections, reported beside the pre-registered value.¹

1. The list comes from Jews before AD 70, read by someone who did not know why. A feature enters if and only if a Jewish, non-Christian source composed before AD 70 attaches it to a future anointed or end-time deliverer, or to the events of his coming. A separate AI agent extracted the list from the Dead Sea Scrolls, the Psalms of Solomon, Ben Sira, the Sibylline Oracles, Jubilees, the Testaments of the Twelve Patriarchs, 1 Enoch, Philo, the Septuagint and Josephus, and was not told the purpose of the study, the hypotheses or the candidates. Its brief told it to be exhaustive: “features that seem odd, militant, or unfulfilled must be included”. After fixed mechanical edits (later and Christian sources dropped, duplicates merged), the list has 82 features.
2. A second reader coded the candidates without knowing which one mattered. Fifteen figures in alphabetical order: Jesus, eleven other claimants from Judas son of Hezekiah (4 BC) to Menachem Mendel Schneerson (died 1994), and three biblical figures read as fulfilling such texts (Cyrus, Hezekiah, Zerubbabel), who are reported but not counted as claimants.
3. A hit is a feature realised in the candidate's lifetime, as reported by any source. A spiritual, figurative or future re-reading (“the temple is his body”, “he reigns from heaven”, “at his return”) is a miss. Unknown is a miss, which favours the best-documented candidate.
4. The events are counted once, elsewhere. Every feature about the figure's death, the manner of his death, or his vindication after death is removed for every candidate: those events are the resurrection page's evidence. Features about the nations stay, but are scored in his lifetime, so the later spread of the church can never count as a hit here.
5. Every hit is graded by who attests it: H, a hostile or neutral non-follower source; M, several independent early follower sources; G, a single follower tradition; T, theological and not checkable. The grade sets how easily a narrator could have supplied the hit. Who could arrange it is marked too: J, the candidate could enact it; N, narrators could supply it.
6. Base rates come from the other claimants. The chance that an ordinary claimant of the era has a feature is estimated from how many of the other eleven have it.
7. Storytelling (f) and staging (g) are unknown, so the result averages over them (each anywhere from 0 to 0.5), and the page shows the whole curve for f.
8. Two comings are allowed and priced. Each feature is, with an unknown probability r, a genuine foretelling; r gets a flat prior from 0 to 1. Misses push r down; that is the explicit cost of deciding afterwards which features belong to a later coming.
9. Jesus was not picked blind, so his result is divided by a look-elsewhere factor: he is tested because his followers read him into these texts, and among 12 claimants someone scores best by chance.
10. The bar was set in advance. A corrected likelihood ratio above 100 (in this list and in our first, unblinded list) for a strong result; 10 to 100 a weak positive; 1 to 10 no margin; below 1 no evidence of foreknowledge.
11. The test could have succeeded. A hypothetical Jesus who hit all 82 features would have scored a corrected ratio of about 10⁶⁵. The design was not too weak to find a signal.

P(reported hit) = 1 - (1 - p0) * (1 - f * inv) * (1 - g * J)

p0  = base rate for a claimant of the era (from the other eleven claimants)
inv = how inventable the hit is: H 0.05, M 0.35, G 0.7, T 1.0
f   = storytelling: chance a narrator reports a known expectation as fulfilled when it was not
g   = staging: chance a claimant who knows the expectation enacts it (only rows marked J)
J   = 1 if the candidate could choose it, else 0
f, g ~ Uniform(0, 0.5); a hidden claimant type (ordinary, prophet, king) links the features of a strand

P(hit | H_T)  = r + (1 - r) * p_null
P(miss | H_T) = (1 - r) * (1 - p_null)
LR = E[ P(profile | H_T) ] / E[ P(profile | H_0) ]   (averaged over r, f, g, p0)
corrected LR = LR / trials factor   (simulated: 2,000 null worlds of 12 claimants)

The evidence

What Jews expected before AD 70

Jewish writings from the two centuries before Jesus do describe a coming anointed figure, but not one figure. Collins counts four paradigms: a king, a priest, a prophet and a heavenly figure.³ The Psalms of Solomon, written after Pompey took Jerusalem in 63 BC, expect a son of David who purges Jerusalem of gentiles, gathers the tribes and rules the nations.⁴ A Qumran text, 4Q521, expects the messianic age to bring healing, the raising of the dead and good news for the poor.⁵ Another, 11QMelchizedek, already counts Daniel's weeks toward an anointed figure.⁶ And Josephus, writing after the war with Rome, says what drove the revolt 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.’”

Josephus, Jewish War 6.312 (Whiston tr.)⁷

The blind extraction found 82 features. By the first paradigm each source names, 42 are royal, 7 prophetic and 33 heavenly, priestly or about the era. The Psalms of Solomon 17-18 are the first-listed source for 26 of them. The sweep found no feature, before AD 70, about the figure's death: no pre-Christian Jewish reader is known to have expected the messiah to die.¹

Every feature scored, misses included

On the agreed coding Jesus has 23 hits and 59 misses of 82. He hits most of what the texts say about the figure's character, titles and teaching, and the prophet-herald-healer features. He hits none of the 19 military and political features: no victory over the nations, no purged Jerusalem, no ingathering of the tribes, no division of the land, no universal peace.^{1, 2}

Look at who attests the hits: 15 rest on several early follower sources, 4 are theological claims no one could check, 2 rest on a single follower tradition, and 2 are attested by hostile or neutral sources (that he was called the Messiah, and that he was opposed and slandered). That is the pattern the null expects: the hits are mostly the kind of thing followers report.

How the Christian lists differ

The popular lists are not drawn from what Jews expected. They are drawn mostly from the New Testament's own citations of the Old, which means the fulfilment reports chose them. None of them lists a single miss. Of Stoner's eight famous prophecies, only one (the forerunner, Elijah's return) is among the 82 features of the blind list.

The control group

If the texts point at one man, he should stand out from other men who claimed or were given the role. Both lists score the same 14 rivals on the same rows: the rebels and sign-prophets Josephus describes,^{14, 15} Simon bar Kosiba, whom Rabbi Akiva greeted with “this is King Messias”,¹⁶ Sabbatai Zevi (1665-66)¹⁷ and Menachem Mendel Schneerson (died 1994), and three biblical figures: Hezekiah, of whom Rabbi Hillel said that “all the prophecies relating to the Messiah were already fulfilled during the days of Hezekiah”,¹⁸ Zerubbabel, and Cyrus, whom Isaiah himself calls God's anointed:

“Yahweh says to his anointed, to Cyrus, whose right hand I have held to subdue nations before him”

Isaiah 45:1, WEB¹⁹

The analysis

Why multiplying selected hits is invalid

Peter Stoner's Science Speaks gives eight prophecies a probability each (Bethlehem 1 in 2.8 × 10⁵, a forerunner 1 in 10³, the donkey 1 in 10², “wounds in the hands” from a friend 10³, thirty pieces of silver 10³, the money thrown to the potter 10⁵, silence before accusers 10³, crucifixion 10⁴), multiplies them, and divides by about 10¹¹ people.¹³ We re-ran his arithmetic from his own inputs and it reproduces: 1 in 2.8 × 10²⁸, so 1 in 2.8 × 10¹⁷.² The arithmetic is right. The method fails five times over:

1. Selection. The eight were chosen because they are reported fulfilled. A list of hits says nothing about the misses.
2. Wrong comparison. It asks how likely a random man is to match. The real question is how likely a claimant who knows the texts, with followers who tell his story in the texts' words, is to be reported as matching.
3. Staging and storytelling ignored. The donkey can be staged. The thirty silver, the potter and the silence are Gospel-only. Three of the eight come from texts no one read as messianic before the Gospels; the Zechariah 13:6 “wounds” passage is about false prophets.
4. Independence assumed. Crucified, pierced and executed with criminals are one event, not three.
5. Guessed inputs. Classes of college students (some 600 in all) agreed on the figures, and Stoner adjusted several. That is not a measurement.

The blind result: no margin

Here is everything the run produced, with the value the review agreed to headline. Every figure is a likelihood ratio for foreknowledge of Jesus against an ordinary claimant plus followers.¹

Why a number below 1 is not evidence against. The data are close to flat. Most of the distance below 1 is the price of the two-comings freedom: under a flat prior on r, a candidate with no hits at all would score 1 in 83. The rest is the look-elsewhere correction, whose method is unorthodox. That is why the pre-registered 0.03 is reported but not headlined: a lone number below 1 reads as “evidence against”, and neither side of the review signs that.

The better rule for misses. The protocol treated every miss as publicly checkable. But some expected features (a white bull, a figure in the divine council) are things no one could check either way. If followers could have claimed such a feature and did not, that tells us something about how much they embellished. The review agreed this is the better likelihood; it raises the headline from the pre-registered model's 0.07 to about 0.3.

The one number that decides it

Everything turns on f: how often followers report a hoped-for feature as fulfilled when it was not. If f is set to a fixed value, the result traces one curve, whichever rule for misses is used.

• A corrected 100 needs f at or below about 0.04 (about 0.03 if the look-elsewhere factor is at its upper bound). In words: followers report an unrealised expected feature as fulfilled for fewer than about 1 in 25 claims nobody could check, and about 1 in 70 claims that several independent follower sources make.
• An uncorrected 100 needs f at or below about 0.06; an uncorrected 10, about 0.1.
• Nobody has measured f. The test cannot: f and foreknowledge explain the same hits. We know of no study that measures it; the work on how the first Christians and later messianic movements read their scriptures gives no number.
• What we can see points higher. Matthew visibly shapes his story to the texts: he has Jesus ride two animals by reading the poetic parallel of Zechariah 9:9 literally (Matthew 21:2-7), assigns Zechariah's thirty pieces of silver to Jeremiah (27:9), and cites a “Nazarene” prophecy that no Old Testament text contains (2:23). A rate below 1 in 10 is hard to square with that, and 1 in 25 harder still. The same f is the Gospels page's question.
• How it could be measured. Movements that left both believers' records and hostile records, scored the same way: the Sabbatean movement of 1665-66,¹⁷ or the Lubavitch movement after 1994. If such a study found f at or below about 0.04, this line would change.

The corrections, and the best case for foreknowledge

1. Linked features. Features within a cluster travel together. Merging only the miss-rich military cluster into one feature would help Jesus, but it is a one-sided choice; merging every cluster the same way gives about 0.26.
2. A richer record. Jesus has four Gospels and Paul; Theudas has three sentences in Josephus. Scored only on what hostile or neutral sources attest, the kind of evidence the controls have, Jesus keeps 2 hits and the ratio is 0.020.
3. The two-comings models. Leaving the whole royal strand to a second coming, with the choice priced, or giving each strand its own foretold fraction, does not reach 100 on either list (at best about 2.7 after correction, on the first list).
4. The one live route is the prophet-herald-healer strand on its own: a ratio of about 4.7, about 1.2 after paying for choosing that strand after seeing the data, and about 0.26 after the look-elsewhere correction. To reach 100 it would need new evidence worth a factor of 85 to 380, from a list fixed in advance by human readers.

Who ranks where

On the blind list, Jesus ranks first of the 12 claimants, and first of all 15 candidates, by a clear margin (0.8 to 1.4 in log₁₀), under every variant of the base rates. In that sense he “dominates any other combination”. But the comparison that matters is with chance: his score lies below what the best of 12 ordinary claimants with his documentation would score in the null simulation (median −0.42, against his −0.84). He outranks the real controls largely because they are known from a few sentences of Josephus. On our first, unblinded list he was second of 12, behind Bar Kokhba. The rank is real; it is not evidence of foreknowledge.¹

This is the “any archetype would fit” objection, measured. The texts do not describe one consistent profile that only one man could fill: the expected king reigns, the servant of Isaiah 53 is killed; 4Q521's healer is not 4Q285's warrior. Different candidates fit different strands: Cyrus, whom Isaiah calls “anointed”, fits the political strand better than anyone.

Daniel 9: a window, not a date

“from the going out of the commandment to restore and build Jerusalem to the Anointed One, the prince, will be seven weeks and sixty-two weeks ... After the sixty-two weeks the Anointed One will be cut off, and will have nothing.”

Daniel 9:25-26, WEB¹⁹

Daniel 9 is the one prophecy of timing in the case, and timing cannot be staged. It is also famous for being computed whichever way gives the wanted answer. So we computed every way: 7 proposed start points (605, 597, 587, 538, 520, 457 and 444 BC) × 3 year lengths (solar, 360-day “prophetic”, lunar) × 5 intervals (49, 434, 483, 486.5 and 490 years) = 105 chronologies, whose endpoints span 603 years.²

So the honest size of the timing hit is a factor of about 1.5 to 4 over chance, not “the exact year”. “Exact” claims need a particular start and a particular year length chosen per start; the mixed scheme sometimes used (457 BC with 360-day years) gives AD 20.1, outside the window. Three facts keep it from being nothing. Daniel predates Jesus even on the critical date (c. 165 BC, with copies at Qumran), so this is not prophecy after the event.²¹ 11QMelchizedek shows Jews before Jesus already counting Daniel's weeks toward an anointed figure.⁶ And the hitting start points are the ones the text names, decrees to rebuild the city.²² But the same public expectation explains why so many claimants appeared then: all 8 first-century claimants in our control group (4 BC to AD 70) are coded as appearing in the expected period. That broad period separates Jesus from Bar Kokhba, Sabbatai and Schneerson, not from Theudas or Menahem.

How Daniel 9 counts. Once, inside the blind list. There it is the feature “appears at a calculable time” (the end of the tenth jubilee in 11QMelchizedek; Josephus's “about that time”), which the blind reader coded as a hit only for Sabbatai Zevi, who announced a calculated year, and as a miss for Jesus. The count of chronologies above describes the text; it is not multiplied in a second time.^{1, 2}

The light to the nations: not counted as prophecy

“It is too light a thing that you should be my servant to raise up the tribes of Jacob, and to restore the preserved of Israel. I will also give you as a light to the nations, that you may be my salvation to the end of the earth.”

Isaiah 49:6, WEB; compare 42:6¹⁹

This was a real expectation before Jesus, made when Israel was a minor client people: pre-Christian Jewish texts expected the nations to come (Psalms of Solomon 17:31; Tobit 14:6).⁴ And the outcome is striking: in 2020 Israel's God was worshipped by about 2.3 billion Christians, 28.8% of humanity, through a Jewish man, while Jews number 14.8 million (0.2%).²³

We do not count it as prophecy, here or anywhere in the model. Three reasons. First, the spread follows largely from the belief that Jesus was raised, and that belief is the resurrection page's evidence; counting the spread again here would count one cause twice. Second, comparable spreads have happened elsewhere: Islam, with 25.6% of humanity, also brought the God of Abraham to the nations, and other movements have grown from small beginnings, so “a small people's God worshipped worldwide” has no base rate small enough to turn into a factor. Third, under the blind test's lifetime rule every feature about the nations (he rules them, he is their light, they come to see his glory) is a miss for Jesus, because none happened in his lifetime. Our earlier version gave it a judged factor of 2 to 20; that factor is withdrawn.

The result, with its uncertainty

What would change it. (1) An independent measurement of f from a movement with both believers' and hostile records, finding it at or below about 1 in 25. (2) A blind, human-built list of one paradigm (the prophet-herald-healer strand, the only one that leans toward Jesus), fixed before extraction, giving a corrected ratio above 100. (3) Hostile or neutral attestation of hits on checkable features (today there are 2). (4) A Jewish text from before AD 70 that reads a death-and-vindication passage of the messiah. (5) For robustness, a two-team human replication of the extraction and coding.

Isaiah 53 in depth

Isaiah 52:13-53:12 is the single best text in the case, and the most argued over. Here is what it says, in part:

“He was despised and rejected by men, a man of suffering and acquainted with disease. ... But he was pierced for our transgressions. He was crushed for our iniquities. ... As a lamb that is led to the slaughter, and as a sheep that before its shearers is silent, so he didn't open his mouth. ... he was cut off out of the land of the living and stricken for the disobedience of my people. They made his grave with the wicked, and with a rich man in his death ... When you make his soul an offering for sin, he will see his offspring. He will prolong his days ... After the suffering of his soul, he will see the light and be satisfied.”

Isaiah 53:3, 5, 7-11, WEB (“the light” in 53:11 follows the Dead Sea Scrolls and the Septuagint)¹⁹

What is established

The text predates Jesus by more than a century: the Great Isaiah Scroll from Qumran contains it whole, and is dated to about 125-100 BC by its script.⁸ It describes an individual-sounding servant who is despised, “pierced” or “wounded” (Hebrew meḥolal, from a root meaning to pierce or wound; both translations are defensible), led like a lamb, cut off, buried with the wicked and the rich, whose suffering is “for our transgressions”, and who then sees offspring, prolongs his days and is exalted so that kings are astonished (52:15). In Scroll and Septuagint he also “sees light” after his suffering.²⁴ Jesus was executed by Rome,¹¹ his followers were proclaiming his death as “for our sins according to the Scriptures” within a few years (1 Corinthians 15:3), and his movement reached the nations.

What is contested

1. Did anyone before Jesus read it of the Messiah? There is no clear pre-Christian evidence, and the blind sweep found no pre-70 feature about the messiah's death. A “killed messiah” reading of 4Q285 is a minority view;¹⁰ the suffering figure of 4Q541 is priestly and fragmentary. The first explicitly messianic reading, the Targum, comes after Christianity and moves the suffering away from the messiah.¹²
2. Who is the servant? The earlier Servant Songs name Israel as the servant (Isaiah 41:8; 44:1; 49:3), and a Jewish reading of the servant as Israel is attested by about AD 248.²⁵ A collective reading is older still: Daniel, written about 165 BC, applies the servant's language to “the wise” (Daniel 11:33-35; 12:3).²⁶ Medieval commentators such as Rashi read the servant as Israel, and Jewish critics today press it hard, adding that the speakers in 53:1-9 are the astonished kings of 52:15.²⁷ The individual reading also has textual support: in 49:5-6 the servant brings Jacob back to God, so is distinct from Jacob-Israel there, and 53:8 has him stricken “for the disobedience of my people” (Singer replies that the Hebrew there, lamo, is plural: “to them”).
3. The fit is partly constructed. The first Christians used Isaiah 53 to make sense of a crucifixion they had not expected. A skeptic's null, a shocked group re-reading its scriptures, predicts exactly that.

Scored whole: about 1 to 2

The rule that governs the whole page, score everything and not just the famous lines, applies to this passage too. Because no one before Jesus read it of the messiah, it sits outside the blind list, so it was scored on its own: all 15 checkable claims of 52:13-53:12, hits and misses, given what happened to Jesus (the events themselves are the resurrection page's evidence), with a price for reading it as a future individual who dies and for having picked this passage out of several of its kind.²⁶

• Three claims are misses: the servant is himself “plagued, struck by God” (53:4), “he will see his offspring” (53:10), and he divides the plunder with the strong (53:12).
• Four hits are Gospel-only details: silent before his accusers, counted with transgressors, a grave “with a rich man”, intercession for transgressors. They add the most, exactly where shaping is visible: only Matthew calls Joseph of Arimathea rich, and Luke has Jesus quote 53:12 on the night of his arrest.
• The result: a prophecy-specific likelihood ratio of about 1 to 2: median 0.84 for Isaiah on its own, 0.57 to 2.22 across the modelling variants. No variant gives a margin.

Why so low, when the fit looks so striking? Three reasons. Once the events are given, the features they entail are close to certain under the null too. A text can cause its own match: the first Christians proclaimed his death “for our sins according to the Scriptures”, reading it through texts like this one, so part of the fit is their doing. And the shape of the passage is a Second Temple pattern found in pre-Christian texts that owe nothing to Isaiah 53. The Wisdom of Solomon's righteous man “names himself servant of the Lord” (2:13); his enemies say,

“For if the righteous man is God’s son, he will uphold him, And he will deliver him out of the hand of his adversaries. ... Let us condemn him to a shameful death”

Wisdom of Solomon 2:18, 20, WEB¹⁹

and later confess they misjudged him (5:4-5). A martyr in 2 Maccabees prays “that in me and my kindred you may stay the wrath of the Almighty” (7:38). Jeremiah says, “But I was like a gentle lamb that is led to the slaughter” (11:19).¹⁹ Only two features of Isaiah 53 appear in none of these: an innocent one who bears others' guilt, the onlookers' included, and the nations and kings. Both are matched by things the text itself could cause.

What survives

A real, pre-Christian text that describes a servant who is killed and then vindicated before the nations, and a Jewish teacher who was killed and whose movement reached the nations. Scored whole, the text adds about 1 to 2. The weight of the case is not in the text but in the events: why did a movement proclaim one man raised, before the general resurrection? That question belongs to the resurrection page. Isaiah 53 is real evidence of how the first Christians understood a death they did not expect. It is not a sealed prediction a skeptic must accept.

Objections

What this shows, and what it doesn't

What we don't claim

• “1 in 10¹⁷” (Stoner), or any product of selected prophecies (10¹⁵⁷, 10⁴⁰, 7.5 × 10²³). These multiply guessed odds of hand-picked hits, assume independence, and ignore misses, staging and storytelling. On the blind list the same move turns about 0.3 into 10¹².
• “Over 300 prophecies fulfilled.” Most come from the New Testament's own citations. The blind sweep of everything Jews before AD 70 expected found 82 features, and Jesus misses 59 of them in his lifetime. The rabbinic passages Edersheim counted (“They amount in all to 456”) come almost all from writings later than Jesus.³⁴
• “Prophecy shows a statistically impossible match”, “there are no anti-prophecies”, or “the texts' author stands outside time”. Scored whole, there is no margin, and the misses are many and public.
• “Prophecy is evidence against Jesus”, or “the blind test refuted prophecy”. The result is no margin, not evidence against; the pre-registered 0.03 is mostly the price of the look-elsewhere correction and must not be quoted alone.
• “Jesus is the only one who fits.” He ranks first among the claimants, but not above what chance and his richer records allow; Cyrus fits the political strand better (Isaiah 45:1 calls him “anointed”); and every first-century claimant in our control group appeared in the same expected period.
• “Daniel 9 predicted the exact day or year.” That needs a chosen start and a chosen year length. The honest claim is a window, AD 27-33. And 457 BC with 360-day years gives AD 20, not AD 27.
• “Daniel 9 and the light to the nations add extra odds on top.” Daniel 9 is counted once, inside the list; the light to the nations is not counted as prophecy.
• “Psalm 22:16 certainly says ‘pierced’.” The Masoretic text reads “like a lion”; the alternative is debated.
• “Zechariah 13:6, ‘wounds in your hands’, is about Jesus.” Its context (13:2-6) is about false prophets.
• “The rabbis read Isaiah 53 as messianic until Rashi invented the Israel reading.” The Israel reading is attested by about AD 248; the messianic readings are post-Christian, and the earliest, the Targum, moves most of the suffering off the messiah.
• “The Dead Sea Scrolls and the Septuagint confirm the prophecies.” They show the texts are old. They say nothing about fulfilment.
• Isaiah 7:14 (“virgin”), Hosea 11:1 (“out of Egypt”), Jeremiah 31:15 (Rachel weeping), “He shall be called a Nazarene” as predictions. None was read as messianic before Christianity; the first two have clear non-messianic contexts, and the last has no Old Testament text.
• “No Jew expected a messiah in the first century; Jesus alone appeared in the window.” Josephus documents a public expectation “about that time” and many claimants.
• “Crucifixion was unknown when Psalm 22 was written.” Impalement-type executions are older, and the psalm's key word is disputed anyway.
• Models of the atonement or the Incarnation “in quantum terms”, as physics. They are analogies, and wrong as physics. The doctrines can be stated as doctrines; they are not evidence on this page.

Sources

1. The blind, pre-registered run (omega.logos.so, ceiling route c10): omega/research/ceiling/c10-salvation/PROTOCOL.md (written before any new data), RECORD.md (the review's rulings, the numbers on this page and the agreed wording), the feature list omega/research/calc/ceiling-c10-UB-features-edited.csv, the coding ceiling-c10-UB-coding.csv, and the model and checks ceiling-c10-prophecy-blind.py, ceiling-c10-S2-checks.py with their outputs. Every script was re-run and reproduced its output exactly.
2. Analysis files for this page: truth/research/calc/09-blind-c10.py (recounts every count quoted from the blind run's data, and the roll-up arithmetic); the first run's 09-universe.csv (33 features × 15 candidates, with the basis of every base rate), 09-prophecy_test.py and 09-daniel9.py, and their outputs. Numbers without another citation come from these runs.
3. J. J. Collins, The Scepter and the Star: Messianism in Light of the Dead Sea Scrolls, 2nd ed. (Eerdmans, 2010). Cited from the scholarly literature.
4. Psalms of Solomon 17-18, tr. K. Atkinson, in A New English Translation of the Septuagint (Oxford, 2007). Phrases checked against secondary summaries.
5. 4Q521 (“Messianic Apocalypse”), tr. G. Strathearn (2011); L. Novakovic, “4Q521: The Works of the Messiah or the Signs of Messianic Times?” (Eerdmans, 2007). Cited from secondary sources.
6. 11QMelchizedek (11Q13), in G. Vermes, The Complete Dead Sea Scrolls in English (Penguin, 2004). Cited via a secondary summary.
7. Josephus, Jewish War 6.5.4 (6.312-313), tr. W. Whiston. penelope.uchicago.edu. Quotation checked.
8. A. J. T. Jull et al., “Radiocarbon Dating of Scrolls and Linen Fragments from the Judean Desert”, Radiocarbon 37 (1995); E. Ulrich, P. Flint, M. Abegg, Qumran Cave 1.II: The Isaiah Scrolls (DJD 32, Oxford, 2010). Cited via a secondary summary.
9. J. J. Collins, “The Background of the ‘Son of God’ Text”, Bulletin for Biblical Research 7 (1997). Cited from the abstract.
10. R. Eisenman and M. Wise; G. Vermes, exchange on 4Q285, Biblical Archaeology Review 18:6 (1992) and 19:1 (1993). Cited from secondary sources.
11. Tacitus, Annals 15.44 and Histories 5.13; Suetonius, Vespasian 4.5. Standard references.
12. Targum Jonathan on Isaiah 52:13-53:12, tr. C. W. H. Pauli, The Chaldee Paraphrase on the Prophet Isaiah (London, 1871), Sefaria; Targum Onkelos, Genesis 49:10, Sefaria. Quotation checked.
13. P. W. Stoner, Science Speaks, ch. 3, “The Christ of Prophecy”. sciencespeaks.dstoner.net. Read; arithmetic recomputed.
14. Josephus, Antiquities 20.97-102 (Theudas; the sons of Judas), 20.169-172 (the Egyptian), tr. W. Whiston. penelope.uchicago.edu. Read.
15. Josephus, Antiquities 17.271-285, 18.4-10, 18.63-64, 18.116-119; Jewish War 2.261-263, 2.433-448, 4.503-584, 7.153-155. Standard references.
16. Jerusalem Talmud, Taanit 4:5, tr. H. Guggenheimer. Sefaria. Quotation checked.
17. G. Scholem, Sabbatai Sevi: The Mystical Messiah (Princeton, 1973). Cited from the scholarly literature.
18. Babylonian Talmud, Sanhedrin 98b-99a, William Davidson ed. (Steinsaltz). Sefaria. Quotations checked.
19. World English Bible (public domain), including the Wisdom of Solomon and 2 Maccabees. ebible.org/eng-web; deuterocanonical quotations checked via bible-api.com.
20. R. A. Horsley and J. S. Hanson, Bandits, Prophets, and Messiahs (1985). Cited from the scholarly literature.
21. J. J. Collins, Daniel (Hermeneia, Fortress, 1993). Cited from the scholarly literature.
22. H. W. Hoehner, Chronological Aspects of the Life of Christ (Zondervan, 1977). Cited from the scholarly literature.
23. Pew Research Center, “How the Global Religious Landscape Changed From 2010 to 2020” (9 June 2025). pewresearch.org. Read.
24. P. W. Flint, The Dead Sea Scrolls (Abingdon, 2013), 66-67, on Isaiah 53:11. Cited from the scholarly literature.
25. Origen, Against Celsus 1.55, tr. F. Crombie, Ante-Nicene Fathers vol. 4. New Advent. Quotation checked.
26. Isaiah 52:13-53:12 scored whole: omega.logos.so, salvation cluster record, omega/research/dev/salvation/RECORD.md §7.3 and Ruling 2, with omega/research/calc/salvation-n61-framework-vector.py (section 5 of its output). Daniel's collective reading of the servant there follows H. L. Ginsberg (1953) and M. Hengel with D. P. Bailey (2004), via M. Z. Brettler and A.-J. Levine (2019); cited from secondary sources.
27. T. Singer, “Who is God’s Suffering Servant? The Rabbinic Interpretation of Isaiah 53”, Outreach Judaism (30 April 2014). outreachjudaism.org. Read.
28. Babylonian Talmud, Sukkah 52a, William Davidson ed. (Steinsaltz). Sefaria. Quotation checked.
29. B. Janowski and P. Stuhlmacher (eds), The Suffering Servant: Isaiah 53 in Jewish and Christian Sources (Eerdmans, 2004).
30. B. A. Strawn, “Psalm 22:17b: More Guessing”, Journal of Biblical Literature 119 (2000). Cited from the scholarly literature.
31. Maimonides, Mishneh Torah, Laws of Kings and Wars 11:4, tr. E. Touger. Sefaria. Quotation checked.
32. J. D. Crossan, Who Killed Jesus? (HarperSanFrancisco, 1995). Cited from the scholarly literature.
33. H. J. Schonfield, The Passover Plot (1965). Cited from the scholarly literature.
34. A. Edersheim, The Life and Times of Jesus the Messiah (1883), Appendix IX. juchre.org. Quotation checked.