Is gematria accurate?

Two different questions

“Is gematria accurate?” usually means one of two things: is the number right? or does the number mean anything? The first has a clean answer. The second has a clean answer too, but it is a statistical one, and it is rarely given.

This page separates them. It also takes the question seriously rather than dismissively: gematria has a two-thousand-year history as an interpretive tool, and the people who practised it were not fools. They also, as it happens, said much the same thing about its limits that the statistics below say.

The arithmetic is exact

A cipher is a fixed table. Given the table and the spelling, a word’s value is determined, repeatable and checkable by hand. In that sense every gematria value on this site is accurate to the last digit, and the methodology page verifies each table against traditional results at build time — if אליעזר stopped equalling 318, the site would not publish.

Where “accuracy” problems genuinely arise, they are almost always label problems: two calculators using different tables under the same name (“English gematria” as A = 1 versus A = 6; “Jewish gematria” as a Latin table; Hebrew finals as 40 versus 600). These are not errors of arithmetic but of communication, and the cure is to name the cipher. Everything on this site does.

How often matches happen by chance

The interesting question is the second one. When two words share a value, how surprised should you be? The way to answer is to take a body of ordinary words that nobody chose for their gematria and count how often they collide.

Below is that experiment on a fixed list of 284 common English words — everyday nouns, verbs and adjectives, no names — published in full in the site’s source so anyone can rerun it. For each cipher: how many distinct values the list produces, what fraction of word pairs share a value, how many words have a value nobody else has, and the most crowded value.

Read the English Ordinal row. 284 unrelated words produce only 101 distinct values, so roughly 1.1% of all pairs — about one pair in 91 — share a value by pure chance. Put differently: pick any common word, and on average several other common words already share its number before you go looking. Full Reduction is worse by design: 47 distinct values, 2.9% of pairs matching — one in 34. The Agrippa (“Jewish”) table, with its values up to 900, is the least crowded of the English ciphers, and even there 0.1% of pairs collide.

The famous 74 makes the point concretely. In this small corpus alone, 4 words total 74 in English Ordinal — point, parent, beauty, cross — and none of them was chosen for it. Jesus, cross and Messiah joining that list is not evidence of design; it is what a value in the crowded middle of the range looks like. The same test applied to the “Kill = 44” trope is worked through on the page for 44.

Why more ciphers do not help

A natural response is to require a match in several ciphers at once. This helps far less than it seems, for two reasons.

First, several ciphers are not independent. Sumerian is exactly six times English Ordinal, so it agrees with Ordinal on every match. Reverse Ordinal is 27 × (letter count) − Ordinal, so two words of equal length that match in Ordinal must match in Reverse. The reductions are functions of the tables they reduce. “A match in Ordinal, Reverse and Sumerian” is one coincidence described three ways. The cipher comparison works through which of the nine English tables genuinely stand alone (five do).

Second, the search is usually flexible. If a phrase can be spelled several ways, split into several sub-phrases, and checked in a dozen tables against a dozen target numbers, the number of chances for some match runs into the hundreds, and the base rates above guarantee that some will appear. This is not a criticism of any person’s honesty; it is how selection works. The only defence is to fix the phrase, the spelling, the cipher and the target before looking.

What the tradition itself says

It is worth knowing that the classical Jewish sources, which practised gematria seriously, also ranked it modestly. The Mishnah calls gematria and astronomy “side-dishes of wisdom” — פרפראות לחכמה, appetisers, not the main course^{[1] Source: Pirkei Avot 3:18}. The Jewish Encyclopedia notes that the one legal ruling ostensibly derived from gematria (the thirty-day Nazirite term) was probably an asmakta, a mnemonic support for a law that had other grounds, and records that from Ibn Ezra onward mainstream commentators spoke of gematria “more or less derogatorily” when it was used to establish rather than to illustrate^{[2] Source: Gematria}. Gershom Scholem’s survey describes its kabbalistic flowering as an interpretive art whose results are as varied as the methods that produce them^{[3] Source: Gematria}.

Even the most famous Christian use shows the same caution. Irenaeus, trying names against 666 around 180 CE, offers Euanthas, Lateinos and Teitan — and then says he will not pronounce which is meant, because many names fit^{[4] Source: Against Heresies, Book V, chapter 30}. He had discovered the base-rate problem in the second century.

Prediction

No classical tradition claims that gematria predicts events; its home is the reading of texts already written. Modern predictive claims — that a date, a score or a headline was “coded” — are made after the fact by finding, among many candidate phrases, spellings and ciphers, one that fits a number already known. A method that can be made to fit any outcome cannot forecast one, and none has, under conditions fixed in advance. This site does not offer date calculators or “decodes” of news for that reason, though it explains how such calculations are done in its article on the gematria calendar.

Using gematria honestly

• Name the cipher with every value, and use the letter values rather than the label when comparing with another source.
• Fix the search before looking: the phrase, the spelling, the cipher, the target.
• Count independent ciphers only, and treat reductions as weak.
• Ask the base-rate question: how many other words share this value?
• Use it as its tradition did — to notice, compare and illustrate, not to prove or predict.

Used that way, gematria is what it has always been at its best: a way of paying very close attention to words. The calculator is built to support exactly that — every number labelled, every letter shown, every table sourced.