Update 3.3.0 — The best players on every division

AAR Labs · published

Division win rates tell you what a division is worth. They do not tell you who is good with it. This update adds that board — plus a look at your own replays that nothing else will give you: how many command points you left sitting in the bank.

Division top 5

Every division on the Global Meta board now shows the players who win with it, beside the matchup breakdown that says what it beats.

You need real games on a division to make its board — a good week does not put you there, a long record does. Each board tells you how many players qualified for it at all, and when each of them last actually played the division, so you can tell an active specialist from a record set two patches ago.

It follows the page filters, so you can ask the question by rank, by ELO or by season.

On profiles

Top-5 placements now appear as chips at the top of a player's profile, and of Your Stats. It is the one thing a profile could never say before: not how someone does overall, but which divisions they are among the best in the game with.

Two new Leaderboard columns

  • Form — the last ten games for every player on the board, oldest to newest, tall is a win. Every row has it, including players nobody has ever uploaded a game of.
  • Divisions — what they field lately, how much of the time, and the tier each one sits in. Hover to see how they do with a division against how everyone else does with it.

That second column is worth more than a win rate. 60% with the strongest division in the game and 60% with the weakest are not the same result, so what you see here is what a player does beyond what their divisions hand them — measured against opponents at their own level.

Float — the points you never spent

Everything on the map has to be bought, and the points arrive on a clock. Buy late and you are permanently behind on the field even when every decision is otherwise sound. It is the one macro mistake a replay can actually prove happened.

Open any of your own replays and the build order now shows it:

  • Idle gaps — the stretches where you stopped buying, marked between the unit you bought before and the one you bought after. Nobody reviews a match second by second, so only the lapses worth your attention are called out.
  • Late buys — each purchase is shaded by how much was already waiting when you made it, measured against your own average. Your normal is the baseline, not somebody else's.

The end-of-match summary, kept

WARNO shows the after-action table when a match ends, and a few seconds later it is gone for good — it is the one part of the report a replay file does not contain.

Leave AAR Labs running in the background while you play and it now catches that screen the moment the match ends, reads the table off it, and stores the numbers with the replay: conquest, kills and losses for both sides, and the kill ratio. They are then in your Match History for good, next to everything else about that game.

It needs that summary to actually be on screen, so if you alt-tab away the instant the game ends there is nothing to catch — that match simply has no numbers, and the app carries on as normal.

And since this involves a picture of your screen, here is exactly what happens to it. There is no AI anywhere in this, and nothing is uploaded to be looked at. The table is ten digits in one fixed game font, so AAR Labs compares each digit against a reference picture of that digit — a plain, predictable comparison that gives the same answer every time and runs entirely on your own PC. The captured image is held in memory for the moment it takes to read those numbers and is thrown away immediately afterwards. No screenshot is ever written to your disk, kept, or sent anywhere.

Verified matches

Every number on the meta boards is built out of replays that players upload — and a replay is just a file on somebody's PC. Nothing stopped a determined person from feeding us invented matches and quietly bending the win rates everyone else reads.

So every ranked match is now cross-checked against our own ladder database, which we build independently of anybody's upload. A match that lines up with a real recorded game gets a shield next to it in Match History. A fabricated one has nothing to line up with, and cannot be passed off as the real thing.

The record does not reach us the instant a match ends, so your shield turns up a little while after you play rather than straight away. It needs sharing to be on, since it is a check against the shared pool.

Sharing

The Global Meta boards and the shared player directory now require sharing to be on. They are built entirely out of players' uploaded matches, so they are for the players who contribute them. Everything read off your own PC, and everything we collect from the public ladder — Match History, Your Stats, Leaderboard, Ranked Pulse — stays available either way.

Fixes

  • Games against unranked players were being thrown away — which cost established players every game they happened to play against a newcomer.
  • Instant surrenders were missing. An opponent conceding during deployment ends the match before the timer runs, and those games were never making it into your record.
  • One bad replay no longer fails a whole upload.
  • Divisions and maps no longer stop being recognised when the game is patched.
  • Players who deliberately have no name are handled properly.

Supporting AAR Labs

AAR Labs is free and always will be. No paid tier, no ads, nothing behind a paywall.

Your own replays are analysed on your own PC — that costs nothing to run. The shared side is not: the live ladder tracking, the leaderboards, and the meta statistics built out of everyone's matches run on servers that cost money every month. So there are now three voluntary donation tiers on the store:

They are listed as DLC because that is the only shape Steam has for this. They contain no content and unlock nothing — no supporter badge, no extra tab, no early access, and no feature will ever be moved behind one.

If you would rather not, please just keep using the app. Nothing changes either