Mapping the Empty Tail of Minecraft Usernames

Overview

Short Minecraft usernames have long been assumed to be effectively exhausted, especially three‑letter names and, to a lesser extent, four‑letter names.

This project conducted a comprehensive scan of the Minecraft username namespace using Mojang’s public API to measure the actual occupancy and availability of short names. The scan systematically checked every possible three‑ and four‑letter lowercase username, recording availability status and rechecking names multiple times to confirm stability.

The scan has now completed, with all four‑letter names checked at least once and many rechecked multiple times. The resulting dataset provides a reliable, stable snapshot of the namespace as of early January 2026. The data shows a clear structural pattern of availability, with a substantial and persistent tail of unclaimed four‑letter usernames.

Motivation

Short usernames have persistent value across platforms. They are easier to remember, faster to type, and often signal early adoption or scarcity.

In the Minecraft community, conventional wisdom holds that:

• 3‑letter usernames are fully taken
• 4‑letter usernames were claimed years ago
• Any remaining short names are edge cases or errors

Despite how often this is repeated, it is rarely backed by measurement. This project exists to replace assumption with observation.

Methodology

Name generation

All lowercase usernames of length 3 and 4 were generated from the character set:

1

abcdefghijklmnopqrstuvwxyz

This produces:

• 26³ = 17,576 three‑letter names
• 26⁴ = 456,976 four‑letter names

For a total of 474,552 candidate usernames.

Availability checks

Each username was queried against Mojang’s public API:

1

GET https://api.mojang.com/users/profiles/minecraft/{name}

Observed behavior:

• HTTP 200 with JSON → name is taken
• HTTP 204 / 404 → name appears unclaimed
• HTTP 429 → rate limited (retried with backoff)

Each request was recorded in a SQLite database with:

• the username
• taken / not taken status
• timestamp of the check
• number of times the name was checked

This enabled rechecking, stability analysis, and pattern discovery.

Rate limiting and retries

To avoid abusive behavior and reduce false negatives:

• Requests were paced to ~1 per second
• Exponential backoff with jitter was applied on HTTP 429
• Names were checked multiple times to confirm status

Results

Three‑letter names

Three‑letter usernames are effectively saturated.

Across thousands of checks, fewer than 0.1% appeared open, and those are likely transient or reserved edge cases.

For practical purposes, the three‑letter namespace is exhausted.

Four‑letter names

Four‑letter usernames present a different pattern.

All 456,976 four‑letter names were checked at least once, with many rechecked multiple times for confirmation. The overall open rate stabilized around 25%, with higher confidence in names surviving multiple rechecks.

Rechecks enriched for genuinely open names, with confirmation buckets showing increasing open rates. Names confirmed open through multiple checks are highly likely to be truly unclaimed.

No username was ever observed to flip between taken and open states across checks, indicating high data stability.

Confirmation phase behavior

The dataset includes names with varying checked_count levels:

• checked_count = 1 ≈ 18% open
• checked_count = 2 ≈ 68% open
• checked_count ≥ 3 ≈ 75% open

This pattern reflects the confirmation pipeline’s enrichment of open names, not a change in actual availability.

Snapshot summary (final)

All values are final as of January 7, 2026.

Structure of the empty space

Availability is not evenly distributed across the namespace.

Vowel density

Names with fewer vowels are significantly more likely to be open.

Current findings indicate:

• 0 vowels ≈ 31.6% open
• 1 vowel ≈ 24.0% open
• 2 vowels ≈ 11.5% open
• 3 vowels ≈ 5.6% open
• 4 vowels ≈ 1.3% open

Grouping names by vowel count:

• Names with 0–1 vowels are approximately 28% open
• Names with 2 or more vowels are approximately 10–11% open

This nearly threefold difference reflects a dense core of human-attractive names contrasted with a long tail of less desirable or less claimed names.

Pronounceability correlates strongly with prior human interest.

Rare letters

Letters such as q, j, x, z, k, and v are associated with much higher availability.

Prefixes containing these rare letters consistently show availability rates between 50% and 70%, even after hundreds of checks and multiple rechecks.

First-letter bias

There is a notable bias in availability based on the first letter of usernames, reflecting human preference shaping exhaustion patterns.

High availability starters include:

• u (~45%)
• q (~36%)
• v (~33%)

Low availability starters include:

• s (~11%)
• a (~11%)
• m (~16%)

This distribution further supports the idea that human selection has strongly influenced which names are claimed.

Human behavior explains the pattern

The data strongly suggests that:

• Humans preferentially claimed pronounceable or meaningful names
• Bots focused on symmetrical or resale‑friendly patterns
• Large regions of the “ugly” name space were never fully claimed

Additionally, alternating consonant–vowel four-letter names, which are typically human-readable, are almost entirely exhausted, with approximately 0.78% availability (886 of 113,388 open names). In contrast, the remainder of the namespace remains around 26–28% open. This contrast demonstrates that the namespace exhaustion is not uniform but shaped by targeted human and bot behavior.

The four‑letter namespace is not saturated; it is polarized.

Finding usable names

Once the structure was clear, targeted searches became possible.

High‑confidence, low‑interest names

Names with no vowels and multiple rare letters were frequently unclaimed even after multiple checks. These are statistically likely to remain available but are intentionally unattractive.

Examples include combinations such as:

1
2
3
4

jkqj
jmzq
qzpj
gzxk

Human‑friendly surprises

Occasionally, readable and human‑friendly names appeared unclaimed. These were rare and likely to be contested quickly, but they do exist.

Examples observed during the scan include names like:

1
2
3
4
5

nego
dowa
guti
kuyi
pixu

These represent the boundary between the dense core and the unexplored tail.

What surprised me most

No username ever flipped state across multiple rechecks, underscoring the dataset’s stability. The namespace exhibits a sharp polarization between human-readable and “ugly” names. Rechecks rapidly concentrate confidence in genuinely open names, and the long tail of unclaimed four-letter usernames is large, stable, and measurable.