Seed data structure

TL;DR: Inspect the 403-byte seed-record layout, decoded fields, and population ranges. Full explanation at the end ↓

Population statistics

The table above shows one seed’s decoded values. This section instead summarizes every metric across all seeds, from the offline histogram scan in scripts/compute-max.js (shipped as maxStats.json). For each metric: min/max/mean/std and the percentiles are the exact population distribution, and top is the record-holding seed.

Two kinds of rows appear:

Per-seed metrics (summed: false) — the value is read straight from one field in the record. The axis.* rows are split by world (West PW / Main / East PW); spell totals, spell types, damage types, horde and faction creatures are each a single aggregate count for the whole seed (no per-world breakdown).
Summed metrics (summed: true, the axisSum.* group) — synthetic: the three per-world axis.* values added together. The record-holder’s per-world parts are shown as e.g. 1+2+5.

Raw decoded record

The literal object returned by reader.readSeed(seed) — expand to inspect.

Full explanation and methodology

Each seed is stored as one fixed 403-byte record and fetched with a single HTTP Range request. Multi-byte values are decoded little-endian according to the shipped seed scheme.

The structure table lists one row per leaf field, including offsets, widths, types, example values, and available population ranges. The raw decoded object at the bottom shows the literal result returned by the reader.