valab

Valorant replay analyzer

valab is a Valorant replay analyzer. Upload the .vrf file the game saved and it comes back as scouting reports, per-map heatmaps, opening duels, trade rates and round impact, for every player in the match rather than only for you.

Create a free accountOpen a real match →

It analyzes the enemy team too

This is the thing most tools cannot do. A replay is the recorded network stream of the match, so it contains all ten players. The same file that holds your rounds holds what the other team did during them: which site they defaulted to, who took the first fight, where their sentinel set up, when their utility came out.

Anything reading a screen recording or a post-match API is limited to what one camera saw or what a scoreboard reported. Neither can tell you where an opponent likes to stand.

Scouting reports

Run a report on yourself, a teammate, or anyone you have played against. It states what that player actually does: the reads worth respecting, the habits worth punishing, and the tendencies in between, each with the number it came from rather than as an assertion.

Reports read across every match you have uploaded that the player appears in, which is what makes them worth anything. A tendency is not visible in one game; it is visible in fifteen.

Map reports, map by map

Movement is where habits hide. valab draws presence heatmaps per map and per phase of the round, routes you can play back, and the angles a player holds with the direction they were watching.

Everything here is filtered by agent, and that matters more than it sounds. The same person plays a duelist and a controller from completely different parts of the map, so pooling their agents produces an average that describes nobody. Heatmaps are per map for the same reason: a smear across four maps is not a pattern.

Duels, trades and impact

Kill-death ratio rewards being alive at the end of lost rounds. These are the numbers that actually describe what happened:

Every stat is graded against the whole corpus of matches on the site, so you see where a number sits rather than just what it is. Where a metric is an estimate, the page states its precision rather than printing decimal places the sampling cannot support.

Built for teams

The people you queue with regularly are grouped automatically. Name a stack and it reads as a single record: the setups you default to on each map, what the group does rather than what five individuals do. Named stacks carry through to the match page, so a side you recognise opens under the name you gave it.

Read from the file, not estimated

Positions, view angles, health, shields, credits, purchases, ability casts and damage are all read out of the replay. Nothing is inferred from video and nothing comes from a stats API. When the site says a bullet hit a wall at a particular point, that is what the file recorded.

The limits are stated too, and so is the moment one turns out to be wrong. Flashes, concusses and nearsight are all read per player now, but nearsight was published here as untracked until we found it on another part of the file, and the patch notes still carry the old limitation alongside the entry that closed it rather than quietly deleting it. Where a value has not been measured yet the site shows a dash instead of a zero. That is the sort of thing the about page goes into properly, and it is deliberately public rather than buried.

Getting started

Grab a .vrf from %LOCALAPPDATA%\VALORANT\Saved\Demos and upload it. A free account gets ten replays a day with every feature switched on, and there is no paid plan to move up to.

If you would rather look before signing up, this match is public. For the review routine the numbers are meant to support, read how to analyze your Valorant gameplay. For playback specifically, see the replay viewer.