Methodology

The bulk of the site is read from VALORANT's published game data rather than typed in by hand. That covers the 29 agents and their ability names and descriptions, the 20 weapons and every statistic on them, the 12 maps, all 1348 weapon skins with their content tiers, colour variants and Radianite upgrade levels, and the 209 bundles.

Because it is read rather than transcribed, it is exactly as accurate as Riot's data and updates when that data does. It also means gaps are real gaps: where Riot does not publish a value, we do not have it either.

The clearest example is ability costs. VALORANT's public agent data carries ability names, descriptions and icons, but not credit costs, cooldowns, damage numbers or ultimate point requirements. Community wikis list those, usually accurately, but we cannot verify them against a source and will not present a community figure as game-file data. So the agent pages show what the files contain and stop there.

Some of the most useful numbers on the site are not published by anyone — they are arithmetic on numbers that are. These are ours, and they are labelled as ours.

Shots to kill is the damage a weapon deals at a given range divided into the target's effective health, rounded up. Head, body and leg damage are taken separately from the weapon's damage profile, and each shield state is worked out independently.

Time to kill is the gap between the first shot and the shot that kills — that is, one fewer interval than the number of shots, divided by the weapon's fire rate. It deliberately assumes every shot lands where you aimed and that recoil and first-shot spread do not intervene, which makes it a floor rather than a prediction. Two weapons with the same theoretical time to kill can feel very different.

Bundle component totals add up the individual store prices of the items in a bundle. Where Riot has not published a price for an item, that item is left out of the total and the page says so rather than filling the gap with an assumption.

Aggregate settings statistics on best settings describe the 497 players VALDB tracks. They are not a claim about the professional population as a whole, and the sample size is stated wherever a percentage is.

Settings and crosshair codes for the 497 players on the site are recorded from ProSettings, and every player page links to the specific source record it was taken from. The current snapshot was captured on 5 August 2026.

This is the least certain data on the site and we would rather say so than imply otherwise. Professionals change sensitivity, resolution and crosshairs often, sometimes between matches, and nothing notifies us when they do. A settings page is an accurate record of what was published at a point in time, not a statement about what someone is using right now — which is why the pages are worded that way.

Callout names come from VALORANT's own map data — they are the names the game uses internally, not a community naming scheme — and their positions are plotted using the minimap calibration values Riot publishes alongside each map.

That has one consequence worth knowing: where the community name and the official name differ, the map pages show the official one. It is the more defensible choice for a database, even where it is not what you would say in voice comms.

VALDB compares consecutive builds of VALORANT's data and records what moved. That feed is at detected changes, and it is a different kind of claim from a patch note.

Riot patch notes are changes Riot has announced. They are on patch notes and attributed to Riot.

Detected file changes are differences we observed between two builds. They are real — the data genuinely changed — but Riot has not necessarily said anything about them, and a value appearing in the files is not a promise that it is live, final, or ever shipping. Content can sit staged for months and be cut.

We keep those two labels apart deliberately. Presenting a datamined change as an announced one is the fastest way for a database to publish something false.

The interaction database is the one dataset here that is not derived from a feed. No published source describes what happens when one agent's ability meets another's, so its 17 entries are curated from observable in-game behaviour.

Each entry carries a confidence marker. “Confirmed” means reliably reproducible and stable across patches. “Reported” means widely observed but worth checking yourself before you rely on it in a match — usually because the interaction is subtle, or recent enough that it may have changed. The database is deliberately incomplete: a partial set of interactions that are true is worth more than a complete one padded with guesses.

Skin market or resale prices. VALORANT skins cannot be traded or sold between players. Any figure presenting itself as a skin's market value is describing something that does not exist. We publish store prices where Riot has published them, and nothing else.

Ranked statistics. Pick rates, win rates and rank distribution for public matchmaking are frequently asked for and we do not have a permitted source for them. Riot's official API grants that access selectively and we do not hold it. Rather than infer numbers from a partial sample and present them as the ranked meta, we publish nothing — and the tier list is built from professional match data, with that stated plainly on the page. Professional and ranked pick rates are different numbers that behave differently, and conflating them would make both useless.

Release dates we cannot source. Riot's data does not carry a reliable release date for every skin. Where we do not have one, the page omits it instead of estimating.

Game-file data is checked twice daily and rebuilt when something changes. A build only runs when the data actually differs, so an unchanged page keeps its existing timestamp rather than claiming freshness it has not earned.

Professional settings refresh on their own schedule and carry the capture date above. Server status is fetched live from Riot's public status feed on each request. Professional match data behind the tier list updates as matches are played.

Different clocks, deliberately: a file scan finding nothing should not restamp an editorial page, and an editorial edit should not imply the game data moved.