Methodology

Killstreaks records statistics from official Activision and first-party studio publications: the Call of Duty patch notes, the official blog, and in-game text. Every value on an entity page links to the source it came from, with the date we read it.

We do not take statistics, descriptions, rankings, loadouts, measurements or page structures from competing Call of Duty sites. Where a fact is only available from a secondary outlet, it is classified as secondary reporting and labelled as such.

Currently recorded: 11 sources, last verified 2026-07-28. The full list is on the sources page.

Every value carries an authority classification, shown publicly next to it. This is the difference between “Activision published this” and “we worked it out”.

When a figure has not been published, we leave it blank and say so. There is no default of zero, no carried-forward guess, and no placeholder that could be mistaken for a measurement. A weapon page showing “not verified” against rate of fire means exactly that: Activision has not published it, so we cannot calculate a time to kill for that weapon, and we say so rather than producing a number.

The same rule applies to hit-region multipliers. An absent head multiplier does not mean 1.0 — it means we have not seen one published.

Call of Duty multiplayer and Warzone apply different damage models. The same weapon can be nerfed in one and buffed in the other on the same day — the AN-94 at Season 05 is a worked example. Every statistic on this site therefore belongs to a ruleset: a specific product, game, mode family, patch and season.

Statistics are never merged across rulesets. Where a context shares another’s values — Warzone Resurgence uses Battle Royale weapon values, as the official notes state — that inheritance is declared in the data and is only ever resolved within the same product. The code that resolves it throws rather than crossing the multiplayer/Warzone boundary.

Shots to kill, time to kill, range breakpoints and attachment-adjusted statistics are computed by a single tested calculation engine, shared by the weapon pages and every tool, so the numbers always agree.

Every calculated result states its assumptions. Specifically, we always disclose:

• whether the first shot occurs at time zero;
• whether bullet travel time is included, and at what velocity;
• whether burst fire delay is included, and how many burst boundaries were crossed;
• whether open-bolt delay is included;
• the assumed health and armour;
• the assumed hit region;
• that all shots are assumed to hit.

Base health and armour are documented assumptions, not officially published figures. They are adjustable in the calculator, and every result that used them says so.

Historical statistics are never overwritten. When a patch changes a value, a new weapon version is recorded and the previous one stays readable on the weapon’s page. Each change is stored as a structured record — entity, field, previous value, new value, unit, direction, and the official scope qualifier (“All Modes”, “BR/RES Only”) verbatim.

Data is published in immutable, checksummed releases. A release is created, validated, diffed against the active one, approved by a named human, and only then activated. Automation can prepare a release; it cannot approve or publish one.

We do not have telemetry. We have no pick rates, no win rates and no usage data of any kind, and we will not use that language about our own opinions.

Our meta rankings are labelled Editor-rated everywhere they appear. They are our reading of published statistics, and the reasoning is written out so you can disagree with it. The data validator refuses to build if a ranking is classified as anything other than editorial judgement.

Official notes are occasionally internally inconsistent. When that happens we record what was published, flag the inconsistency, and lower our confidence — we do not quietly reconcile it.

Two live examples, both visible on the site today:

• The Black Ops 7 Season 05 notes head the AN-94’s damage table “VOYAK KT-3 DAMAGE ADJUSTMENTS”, while the surrounding section is headed “WEAPON: AN-94”. We record the values against the AN-94 per the structural field, at medium confidence, with the discrepancy noted on the version.
• The Warzone Season 05 pre-patch figures for the VST list the medium band ending at 22m and the minimum band beginning above 21m — a one-metre overlap. Recorded as published, with a validator warning.

When we get something wrong we fix it and say so. Corrections are listed on the corrections page and on the article or entity they affect, with the date. We do not silently edit a page and leave the original claim unacknowledged.

Every entity page states how much we actually hold: complete, partial, identity-only, or referenced-only. A page is only invited into search results when there is enough behind it to be worth reading. That is why some weapons appear in a directory but have no page of their own — we know their name and class and nothing more, and a page saying so would waste your time.