DnD halfling names
A halfling name here is two separate draws: a first name assembled from the culture's onsets, vowels and ten closings, then a surname compounded out of eight heads and eight tails. Nothing in the second draw can see the first. Over 10,000 draws at seed 20261101 that split is easy to measure — 6,898 different first names, 64 surnames, and the busiest first name meeting five different surnames across eight appearances.
- Riro Teabarrel
- Boser Goodpot
- Minin Rootmead
- Nirer Burrowbough
- Buno Undermead
- Wero Amberleaf
- Nupo Gooddown
- Nomin Underwell
- Homo Burrowbough
- Rerin Underpot
- Delo Honeyleaf
- Belo Goodmead
| Measure | Result |
|---|---|
| Different full names | 9,937 — 63 repeats (0.63%) |
| Different first names | 6,898 |
| First names seen only once | 4,959 (71.9%) |
| Surnames | 64 (8 heads x 8 tails), each drawn 156 or 157 times |
| Different first names per surname | 153 to 157, median 155 |
| Mean first-name length | 5.61 letters, shortest 4, longest 7 |
| Busiest first name | Belo — 8 draws, 5 different surnames |
| Same-surname pairs | 65 against 78.5 expected at the 1/64 baseline |
Two pools with a space between them
The first name is built from three lists: ten onsets (b d f h m n p r t w), five vowels (a e i o u) and ten closings (bo dy lo wy ni po in er o y). Six closings already start on a consonant; the four that start on a vowel take one of five coda consonants (l n r s m) so the stem does not run straight into them, which turns ten closings into 26 possible shapes. Halfling is not one of the long-name cultures, so a first name is one or two opening syllables:
| Piece | Count | Combinations |
|---|---|---|
| Onsets x vowels | 10 x 5 | 50 openings per syllable |
| One-syllable names | 50 x 26 | 1,300 |
| Two-syllable names | 2,500 x 26 | 65,000 |
| First names in total | 1,300 + 65,000 | 66,300 |
| Surnames | 8 heads x 8 tails | 64 |
| Full names | 66,300 x 64 | 4,243,200 |
The surname side is flatter: Under, Tea, Good, Burrow, Honey, Thistle, Amber and Root closing on bough, leaf, barrel, bank, pot, down, mead or well. The two picks run on separate balanced counters — the closing counter is keyed to the race, the surname counter to the culture — and the surname function never receives the first name as an argument.
Do the halves ever meet?
They can, by coincidence, and the run shows how rarely. Of the 1,939 first names that came up at least twice, 1,928 met two or more different surnames; 11 met only one, and 59 (3.0%) ran into the same surname twice somewhere in the batch. Counted as pairs rather than names, the batch produced 65 same-surname pairs against 78.5 expected if the two halves were drawn independently at 1/64 — a gap of 1.5 standard deviations on a square root of 78.5 = 8.9, which is inside the spread two independent picks would produce.
| First name | Draws | Different surnames |
|---|---|---|
| Belo | 8 | 5 |
| Narin | 8 | 7 |
| Molo | 8 | 8 |
| Dolo | 8 | 7 |
| Faly | 8 | 8 |
| Rolo | 8 | 7 |
| Bulo | 7 | 6 |
| Welo | 7 | 7 |
| Telo | 7 | 7 |
| Nuny | 7 | 7 |
| Nilo | 7 | 7 |
| Filo | 7 | 7 |
Molo and Faly reached eight draws with eight different surnames apiece; Narin, Dolo and Rolo reached eight draws with seven. Belo is the odd one — eight draws, five surnames, because one surname took it three times. That is what an independent pairing looks like at this sample size: mostly fresh, occasionally doubled.
Every surname is a crowd
Sixty-four surnames across 10,000 draws is 156.25 apiece on an even split, and the run came back with every one of them used 156 or 157 times — the per-batch cap holds them that tight. What makes a halfling name look like it belongs to a family is not the surname repeating, but the company it keeps: each of the 64 met between 153 and 157 different first names, median 155. A surname repeats 156 times and arrives with a different first name almost every time.
Full names repeat 63 times in 10,000 draws (0.63%), which is the surname pool showing its size: 64 combinations cannot carry 6,898 first names without collisions. The first names on their own come from a pool of 66,300 and repeat far more often — 4,959 of them appeared exactly once.
Lightfoot and Stout: one surname pool, two first-name pools
Both halfling races map to the same culture in the code, so the surname pool is literally the same 64 names for both — Underbough, Tealeaf, Goodbarrel and the rest of the grid. The first names are not shared: the closing tables differ by six entries (Lightfoot runs bo dy lo wy ni po in er o y, Stout runs in o er y el on ic ar od uck), and Stout first names come out longer: 6.30 letters against 5.61.
| Measure | Lightfoot | Stout |
|---|---|---|
| Different full names | 9,937 (63 repeats) | 9,971 (29 repeats) |
| Different first names | 6,898 | 8,477 |
| First names seen once | 4,959 (71.9%) | 7,395 (87.2%) |
| Mean first-name length | 5.61 letters | 6.30 letters |
| Same-surname pairs | 65 vs 78.5 expected | 29 vs 33.1 expected |
| Surnames | 64 | 64 (same pool) |
The pairing test repeats itself on the second run: 29 same-surname pairs against 33.1 expected, a gap well inside a square root of 33.1 = 5.8. Two different first-name pools, one surname pool, the same absence of a link.
How this was measured
Pools (code constants). 10 onsets, 5 vowels, 10 closings, 5 coda consonants, 64 surnames (8 x 8). Halfling is not in the long-name set, so first names are one or two opening syllables.
Sample. 10,000 draws at seed 20261101, gender Neutral, the five styles cycled, one tracker. The Stout column is a second 10,000 draws at the same seed and the same settings.
Counting. The first name is the text before the space, the surname the text after it. A same-surname pair is two draws sharing both halves. Expected pairs = the sum of C(k,2) over first names with k draws, divided by 64 — 78.5 here. The per-batch surname cap keeps the pool spread, so 1/64 is an upper bound rather than an exact expectation.
Reproduce it
var s=document.createElement('script'); s.src='/assets/halfling-stats.js'; document.body.appendChild(s);Run that line on the homepage, not on this page: the generator only loads there. It prints both subrace runs for seed 20261101 — distinct first names, the surname-met buckets, the same-surname count against the 1/64 expectation, and the busiest first names with the surnames they met.
Questions
Does a halfling first name come with a fixed surname?
No. Of the 1,939 first names that came up more than once in 10,000 draws, 1,928 met two or more different surnames and only 59 (3.0%) ever met the same surname twice. Same-surname pairs came to 65 against 78.5 expected at the 1/64 baseline (the sum of C(k,2) over first names with k draws, divided by 64). The gap is 1.5 standard deviations on a square root of 78.5 = 8.9, which is noise rather than a rule.
How many halfling names can the generator make?
66,300 first names times 64 surnames = 4,243,200 full names, computed from the code's own tables. Ten onsets times five vowels is 50 openings; one syllable gives 50 x 26 = 1,300 and two syllables 2,500 x 26 = 65,000, where 26 is what the ten closings expand to once the four vowel-initial ones take one of five coda consonants (6 + 4 x 5 = 26). Halfling is not one of the long-name cultures, so three opening syllables are never used.
How long are halfling first names?
5.61 letters on average for Lightfoot, between 4 and 7. Six letters is the common case: 6,247 of 10,000 draws, then four letters (1,754), five (1,211) and seven (788). Stout first names run longer - 6.30 letters on average, spread over 4 to 8.
Do Lightfoot and Stout halflings share surnames?
Yes, all 64. Both races point at the same halfling surname pool in the code: Under, Tea, Good, Burrow, Honey, Thistle, Amber and Root crossed with bough, leaf, barrel, bank, pot, down, mead and well. The first names are separate pools - 6,898 different ones for Lightfoot against 8,477 for Stout in the same 10,000 draws.
Generate halfling names · Last names and how surnames compound · Gnome names, two subraces measured · All 36 races