Remote config and A/B tests
Change values in your game without an update, and split players between versions to see which one does better.
Both are on a game's Config & tests tab. Your game reads them with the SDK's config call (see the SDK reference). They aren't available for Fortnite islands, which can't run the SDK.
Remote config
Config values are things your game reads at runtime, like a price, a drop rate or whether an event is on. Change them in the dashboard and your game picks them up the next time it fetches config, no update needed.
- Key: starts with a letter; letters, numbers,
_,.and-; up to 64 characters. For examplehalloween_enabled. - Value: anything JSON (
true,20,"text",{"json": 1}). Anything that isn't valid JSON is stored as text. - What it does: an optional note of up to 200 characters for your team.
Deleting a key means games fall back to their built-in value, so always ship a sensible default in code.
DevNet caches each game's config for up to 30 seconds, so a change can take that long to reach players, plus however often your game fetches it.
A/B tests
A test splits players between versions that override config values, and compares how many players in each version do your goal event.
- Give the test a name and a key. The key is unique per game and decides how players are split.
- Optionally pick a goal event, such as
purchaseortutorial_done. - Set up two variants, each with a name, a split (a weight from 0 to 100) and its config overrides as a JSON object, like
{"starter_pack_price": 49}. Leave the control's overrides empty to keep the normal config. - Tests start as drafts. Select Start once your game reads the config keys. You can Stop and Resume a test, or delete it with its results.
How it works:
- Each player always lands in the same variant. The split is worked out from the test's key and the player ID, so it doesn't change between sessions or servers.
- While a test runs, the config call returns that player's overrides merged on top of the normal config, plus which variant they're in.
- Results count the players seen since the test started in each variant, and the share of them who did the goal event. The variant with the best rate is marked Leading.
Give a test a few days and a few hundred players per variant before deciding.