DIRECTORY

METHODOLOGY LIBRARY

Find the right tool or article for the question you’re asking.

Tools for thinking about uncertainty, and articles on the methodological choices that shape how the 40k metagame gets measured. Use the tools to compute. Use the articles to understand why the computation matters.

TOOLS
Glossary
Terminology and definitions — 40k vocabulary, statistical terms, Archive-internal language.
you encounter terminology you want defined precisely.
OPEN ›
CI Explorer
Wilson 95% confidence interval around a reported win rate, single-rate or paired comparison.
you need to know how uncertain a reported rate is given its sample size.
OPEN ›
Two-Proportion Test
Newcombe hybrid score interval on the difference of two rates, with a three-tier plain-language verdict.
you have two rates and need to decide whether they’re meaningfully different.
OPEN ›
Power Analysis
Sample-size and minimum-detectable-effect calculator, with a cadence-calibrated verdict.
you need to know how much data it would take to settle a question — translated into tournament-cadence reality.
OPEN ›
Damage Distribution Calculator
The exact full probability distribution of an attack profile’s damage — mean, median, mode, and P(kills the unit) on the same picture.
you want to know how reliably a unit clears its target, not just “on average.”
OPEN ›
ARTICLES

Each article tackles a specific way that aggregate metagame numbers can mislead — what they measure, what they don’t, and what to read instead. The plain-language question each one answers is on the card.

The Big Soup Problem
Treating each tournament weekend as a replicate gives wider — and more honest — uncertainty bounds than pooling every game from every event into one number.
two factions sit a few points apart in a pooled table and you want to know how stable the gap is.
READ ›
Six Bins
A 5-round event has only six possible personal win rates. The inherent binomial noise floor at that sample size is roughly ±22 percentage points.
a single event’s “60% win rate” is being read as a measurement rather than a story.
READ ›
Swiss Isn’t Random
Going 5–0 means playing the top of the bracket; going 1–4 means the bottom. Swiss-paired observations are structurally correlated; standard significance tests assume they aren’t.
you want to know how much a Swiss tournament’s sample size is actually worth.
READ ›
On Average
What “average damage” actually means: the difference between expected value, the most likely single outcome, and the median.
the heuristic says “on average this kills the unit” and you want to know the actual probability.
READ ›

More tools and articles are planned. The current set is calibrated to the questions that come up most in week-to-week metagame discussion.