Bokki
en

How an Ogham draw is calculated

The twenty feda, the seed, and the day we measured our own shuffle and it failed.

The short answer

Bokki's Ogham page (/ogham) draws three of the twenty primary feda — the letters of the Ogham alphabet, usually called the Celtic tree alphabet — and reads them as where you have been, where you are, and what helps. The draw is not random in the usual sense. It is seeded: the page trims and lowercases whatever focus you typed, joins it to today's date in a short string, hashes it, and shuffles the twenty letters with that hash. So the same question on the same day gives you the same three feda, every time, on any device. Typing LOVE and typing love give the identical cast, because the string is lowercased before it is hashed. There is no score anywhere in this, no percentage, and no ranking of the letters into good and bad ones.

What the twenty are

The feda are arranged in four groups of five, called aicmí, in the Beith-Luis-Nin order that the medieval Irish tracts give — Beith, Luis, Fearn, Sail, Nion, then Uath, Dair, Tinne, Coll, Ceirt, and so on. All twenty exist as their own Unicode characters, and they sit in one unbroken run from U+1681 to U+1694; we checked that the block really is contiguous rather than assuming it. Each few carries a plant word from the Word-Ogham of the tracts. Worth saying plainly, since almost every page on the internet calls this a Celtic tree alphabet: five of the twenty are not trees. Muin is the vine, Gort the ivy, nGéadal the reed, Onn the gorse, Úr the heather — climbers, a grass and two shrubs. Our data file has labelled that column tree-or-plant since its first commit, and this article is not going to round it up to trees.

From your typed focus to three letters

The seed string is hashed with FNV-1a into a single 32-bit number, that number becomes the state of a small PRNG called mulberry32, two outputs are thrown away as a warm-up, and the remaining stream drives an ordinary Fisher-Yates shuffle of the twenty positions. The first three off the top are your cast. Everything about that chain is deterministic and lives in the page you are using, not on a server — nothing about your question is sent anywhere. One property falls out of doing the shuffle this way rather than re-hashing at every step: drawing one few and drawing three feda from the same seed give the same first letter, so the reading does not quietly change shape depending on how many cards you asked for. Change the date or a single character of your focus and the shuffle starts over from a new hash — which is not the same as every letter changing. We counted: across 20,000 one-character changes of the focus, 8,129 casts still shared at least one few with the original, about two in five, which is exactly what a fresh draw of three from twenty should do, and one cast came out identical.

We measured our own draw, and it was not fair

Here is the part we would rather not have to write. A fair twenty-sided draw should land on each few about equally often. In August 2026 we ran twenty thousand seeds of exactly the shape the page makes — ogham:2026-08-18:w0 through w19999 — and counted which few came up first. It should have been about 1,000 each. Instead Gort, the ivy, came up 1,849 times and Tinne, the holly, came up 176 times: one few was 10.5 times more likely than another. The chi-squared statistic was 3,907.3 against a critical value of 43.82, which is not a near miss, it is a different distribution entirely. The cause was in the shuffle: it called the hash function again at every swap on a string that differed only by a counter, and FNV-1a's low bits stay correlated across near-identical strings, and the modulo read exactly those low bits. Our I Ching engine already had a comment warning about this trap; the Ogham, rune and tarot engines had simply not followed it. We moved all three onto the same PRNG. On the same twenty thousand seeds the spread is now 939 to 1,052, a ratio of 1.12, with chi-squared 14.6 — comfortably inside chance. None of this was visible on the screen, because any one person only ever sees their own cast. It was only visible to someone who counted.

Our own Ogham draw, counted twiceTwo rows of twenty bars, one bar per few, showing how often each of the twenty Ogham feda came up first across twenty thousand seeds of the shape the page actually builds. In the top row, the draw before the fix, the bars are wildly uneven: Gort the ivy reached 1,849 while Tinne the holly sat at 176, a ratio of 10.5 to 1, against an expected 1,000 each. In the bottom row, the same twenty thousand seeds after the fix, every bar sits close to the expected line, between 939 and 1,052, a ratio of 1.12 to 1. The chi-squared statistic falls from 3,907.3 to 14.6, where the threshold for calling a draw unfair is 43.82.before — 10.5× spreadexpected 1,000Gort (ivy) 1,849Tinne (holly) 176after — 1.12× spread20,000 seeds · chi-squared 3,907.3 → 14.6 (unfair above 43.82)

What we don't calculate

We draw only the twenty primary feda, not the five forfeda — the supplementary letters added later and never fully settled. We draw upright only: Ogham has no settled reversed reading — tarot's reversals are a long-standing cartomancy convention, and even the runes' merkstave is a twentieth-century one that varies by author — so inventing one for Ogham would be inventing tradition, and we would rather leave the gap visible. We do not spell your name in Ogham and read the letters as a fate — the script was used for names on standing stones, but as writing, not as an oracle about the person. We do not use the Celtic Tree Calendar that assigns a few to each month of the year: that scheme comes from Robert Graves's The White Goddess of 1948, and specialists in Old Irish treat it as a twentieth-century construction rather than a medieval one, so it is not in our code. And we do not score: there is no lucky few, no unlucky few, and no number attached to your answer. What you get is three short plant-words held up as a mirror.

Closing note — 10.5× before, 1.12× after, on the same 20,000 seeds

In short: a cast is three of twenty feda, drawn by hashing today's date and your typed focus into a seed and shuffling with it — no dice, no hand in sand. We measured our own fairness rather than assuming it: 20,000 seeds of the exact shape the page builds turned up Gort the ivy 1,849 times and Tinne the holly only 176, a 10.5× spread with chi-squared 3,907.3 against a threshold of 43.82. The cause was re-hashing a near-identical string at every swap; after moving to a single PRNG stream, the same 20,000 seeds spread only 939 to 1,052, a 1.12× ratio with chi-squared 14.6.

We draw the twenty settled feda only, never the five unresolved forfeda, and we do not read upright-only draws as reversed just because tarot and modern rune readers read reversals — Ogham has no settled convention for it, so we left the gap visible instead of inventing one.

← Cast the feda

Read next

Try /ogham — cast three feda
How a rune draw works — the Elder Futhark, honestly
How a tarot draw works — and what Bokki actually deals
What is the I Ching? — the other seeded draw, and why it uses fractions

Questions about this draw

Is the Ogham draw random?

Not in the coin-flip sense. It is seeded: the page trims and lowercases whatever focus you typed, joins it with colons to the word ogham and today's date, hashes that string, and shuffles the twenty feda from the hash. So the same focus on the same day always returns the same three letters, and typing LOVE gives you the same cast as typing love. Change the day or one character of the focus and the shuffle restarts from a new hash — about two in five such casts still share a few with the original, as a fresh draw of three from twenty should (we counted 8,129 of 20,000). The whole calculation runs in your browser — your question is not sent anywhere.

Was the draw ever unfair, and how would you know?

Yes, and we found it by counting. We ran 20,000 seeds of the shape the page actually builds and tallied which few came up first. Each of the twenty should appear about 1,000 times. Gort (ivy) appeared 1,849 times and Tinne (holly) 176 — a ratio of 10.5 to 1, with a chi-squared of 3,907.3 against a threshold of 43.82. The shuffle was re-hashing a near-identical string at every swap, so the hash's low bits stayed correlated. After moving to the same PRNG our I Ching engine already used, the same 20,000 seeds spread between 939 and 1,052, a ratio of 1.12 and a chi-squared of 14.6. Nobody could have seen this on the screen: one person only ever sees their own cast.

How many feda are there, and do you use all of them?

We use the twenty primary feda, arranged in four aicmí of five in the Beith-Luis-Nin order. We do not use the five forfeda — the supplementary letters added later, whose values never fully settled. All twenty are real Unicode characters in one unbroken run from U+1681 to U+1694.

Can a few come out reversed?

No. Tarot reads reversed cards as a long-standing cartomancy convention, and modern rune readers read merkstave — itself a twentieth-century convention that varies by author; Ogham has no reversal rule settled even that far. Rather than invent a reversal rule and present it as tradition, we draw upright only and leave the gap where it is.

Are all twenty feda trees?

No, and this is worth correcting since almost everything written about Ogham calls it a tree alphabet. Five of the twenty are not trees: Muin is the vine, Gort the ivy, nGéadal the reed, Onn the gorse, and Úr the heather. Our data has always described them as plants rather than trees.

Do you use the Celtic Tree Calendar?

No. That scheme, which assigns a few to each month of the year, comes from Robert Graves's The White Goddess of 1948; specialists in Old Irish treat it as a twentieth-century construction rather than something attested in the medieval sources. It is not in our code, and we would rather say why than quietly include it.