dnd name forge

Tavern name generator

A tavern sign is an article and two words. This page runs that frame 10,000 times to answer the question a city map actually asks: how many inns can you name before two of them match? The median run gets there on draw 115.

What this page is counting. The pool is 58 adjectives × 50 creatures × 56 objects, used in three shapes: The Gilded Griffin, The Broken Wheel, The Boar and Barrel — 8,948 possible signs in all. One run of 10,000 mixed draws at seed 20261023; then 2,000 draws per shape; then 500 fresh sessions of up to 400 draws each, counting where the first repeat lands; then 20,000 simulated cities at each size. A draw is one call to the tavern mode with no duplicate guard — what you do across sessions, not inside one fifty-name batch. A repeat is the same full string. Every figure below is counted from that run; the only arithmetic is the birthday column, which is labelled as such.

Fifty tavern names from the run

Seed 20261023, tavern mode, fifty names, duplicate guard on so the batch holds no repeats. The register moves between a sign you could paint (The Porcelain Crown) and one that sounds like a rumour (The Splintered Wolf), because the adjective list runs from painted to wounded.

The pool is 8,948 signs, and the run used 67.3% of it

10,000 mixed draws produced 6,021 different signs, so 3,979 of them (39.8%) were names already drawn. That is 67.3% of the 8,948 possible strings, close to what the birthday problem predicts for 10,000 draws from a pool that size: 6,022. The run landed 1 name from that figure, and the three shapes individually land within 0.6% of their own prediction — so the repetition is not a bias in the generator, it is the size of the bank.

The first repeat lands at draw 115

Five hundred fresh sessions, each seeded differently, each drawing until something repeats. The median session hit its first duplicate on draw 115; the mean was 118.9, the tenth percentile 48 and the ninetieth 200. None of the 500 sessions repeated inside ten draws; 216 of them (43%) repeated inside a hundred. The birthday model for a pool that size says 119.2, so the generator is sitting on that line rather than above it.

Draw number of the first repeat, 500 sessions 1-24 draws12 (2.4%) 25-49 draws42 (8.4%) 50-99 draws154 (30.8%) 100-199 draws242 (48.4%) 200-299 draws48 (9.6%) 300+ draws2 (0.4%)

This is the number that matters when you prep in a notebook: past roughly a hundred signs, half of what you write down is already on an earlier page. Below that, the pool still feels bottomless, which is why tavern names seem to run dry on the third page of notes rather than the tenth.

Naming a city: twenty inns or fifty

Each city below is a draw of k names from the measured distribution; the trial counts a failure if any two match. Twenty inns collide in 3.09% of cities, fifty in 18.27%. The mean number of duplicate names in a fifty-inn city is 0.20.

Repeats when naming k taverns from the same pool, 20,000 simulated cities per row
Taverns in the cityCities simulatedCities with a repeatShareMean duplicates
520,000340.17%0.002
1020,0001570.78%0.008
2020,0006193.09%0.031
5020,0003,65318.27%0.200
Cities where two taverns share a name 5 taverns0.17% 10 taverns0.78% 20 taverns3.09% 50 taverns18.27%

The three shapes, counted separately

Each shape drew 2,000 times on its own seed. Adjective + object has the largest bank (3,248 strings) and the best new-name rate at 75.2% per draw; the two shapes that put a creature in the sign repeat first, and the gap between the winner and the loser is 4.0 names in a hundred.

2,000 draws per shape; the last column is the birthday prediction for the same draw count
ShapePossibleDrawnDistinctNew per drawBirthday
The adjective + creature2,9002,0001,43771.8%1,445
The adjective + object3,2482,0001,50375.2%1,494
The creature and object2,8002,0001,42471.2%1,429

Read the last two columns together: every shape lands within a fraction of a percent of the closed-form answer, which is the useful part. Repetition cannot be fixed by changing the rhythm of the name — only by adding words to the thin list.

The shortest word list is the busiest

Two of the three shapes put a creature in the sign, so the creature list takes two thirds of all draws: 133.3 draws per creature word against 114.9 per adjective and 119.0 per object. Every word in all three lists turned up at least once in the 10,000 (58 of 58, 50 of 50, 56 of 56), so nothing is dead weight — the creature list is simply the shortest and the busiest. Busiest words: creatures Worm 153, Sparrow 152, Raven 151, Manticore 148, Bear 146; adjectives Smoking 137, Empty 134, Crowing 130, Burning 129, Laughing 129; objects Bottle 148, Horn 138, Crown 137, Tower 137, Window 134.

Practical version: if two inns in one district should feel unrelated, change the creature, not the adjective. The Silver Boar and The Rusted Boar read as the same house with a repaint.

Using these at the table

Questions this page answers

How many DnD tavern names can a pool like this make?

8,948 combinations: 58 adjectives against 50 creatures (2,900), 58 against 56 objects (3,248), and 50 creatures against 56 objects (2,800). A run of 10,000 draws returned 6,021 of them, 67.3%, and 500 fresh sessions repeated for the first time at a median of 115 draws, against a birthday prediction of 119.2 for a pool this size.

How many taverns can one city have before two share a name?

In 20,000 simulated cities per size: five inns collide in 0.17% of cities, ten in 0.78%, twenty in 3.09% and fifty in 18.27%. Twenty is comfortable, fifty is not — at fifty the average city carries 0.20 duplicate names.

Which shape gives the most different names?

Adjective + object, with 3,248 possible strings and 75.2% of 2,000 draws new. Adjective + creature came in at 71.8% and creature + object at 71.2%, so the gap is small: the shape is a tone choice, not a volume choice.

Why do so many tavern names start with “The”?

All three shapes in this pool start with it, so the article never carries information — every distinguishing word sits in the two slots after it. That is also why the pool needs two lists rather than one: an article plus a single noun would give 50 or 58 possible signs, and the run would repeat far sooner — the median in this run was draw 115 for two slots.

Are the fifty names above generated or hand-picked?

Generated in one pass, seed 20261023, tavern mode, duplicate guard on so the batch itself holds no repeats. The script at the bottom of this page reproduces the same fifty and every table above, because the draw is seeded.

Reproduce this page

var s=document.createElement('script'); s.src='/assets/tavern-stats.js'; document.body.appendChild(s);

The script draws through the site's own makeTavernName, so the three word lists, the three shapes and the retry rule are the shipped ones. Parameters: seed 20261023, 10,000 mixed draws, 2,000 per shape, 500 sessions of up to 400 draws, 20,000 cities per size, fifty-name shortlist. It prints the same numbers to the console and leaves them on window.__TAVERN_STATS.

Generate a roster · Town names to hang them in · Campaign titles · All measured pages