Fortnite Battle Royale

Competitive genre, but the audience was twelve-year-olds and ex-pro CS players in the same lobby. Tune for the top 5% and retention craters. Tune for casuals and the competitive ceiling collapses.

A meta legible to a new player but still rewarding mastery at the top. Tuned weekly. Neither group melted.

Both audiences can't win on a single change. You ship a direction, watch the data, accept one side will be loud. The lead work is deciding which side wins this week, and defending it in the patch notes.

Layered, not switched. High-mastery items (movement, building, headshots) carried the skill ceiling. Lower-mastery items carried accessibility. When the meta ran hot at the top we'd nerf the mastery item, not buff the casual one. Adding power to the casual layer is reversible. Pulling it from the competitive one isn't.

How often the right call was smaller than the team wanted. Big swings made patch notes louder and the meta less stable. Tuning in 5% weekly increments produced a healthier season than any single decisive move.

Build an event that feels global to every player on earth. No patch window.

Global feel at the individual server level. A player on a 100-person server should still feel millions pulling in the same direction.

Allow failure cases. If players didn't engage, progression stalled. They could troll the event by refusing to participate. We let that risk stand. The bet was that most players would want to advance it. It could have failed publicly.

Keep all interaction simple. This was a side layer running inside an active competitive match. You never knew if the player approaching you was there to fight or to help. Required attention had to be minimal. Pickaxe an object. Hold an interact key. The interaction primitives changed between stages, but the complexity ceiling stayed deliberately low.

The unvaulting itself rode on the hotfix infrastructure I'd built two years earlier. The moment the vote closed, the chosen weapon's loot row flipped via hotfix, and every server got the change before its next match.

The players chose the Drum Gun. We had biplanes and other options on the ballot, and most of us internally were betting on the planes. The Drum Gun was polarizing. The vote said something about what the community valued (chaos, accessibility, a known meta-warping weapon) that none of us had fully predicted.

Design lead, end to end. Led a team of designers on execution. The synthesis was the part nobody else could split with me.

Among Us had a specific texture: small player count, role asymmetry, communication windows, voting rituals, a task system. Fortnite had its own conventions for movement, item interaction, lobbies, and season fit. Respecting both lineages was a single-owner judgment call.

Player count. Communication mechanics. Role-reveal pacing. Task-system flavor. Battle Royale integration. Each was a tradeoff between an Among Us beat and a Fortnite expectation. The moves that worked were the ones where the answer was visibly neither one.

Heavily praised internally. Leadership specifically flagged it as having done justice to both source experiences. That recognition is the closest thing I have to a public record of the design judgment behind the mode.

A real-money tournament inside a creator-tool environment, accessible to every player on the platform, at platform scale. Transparent. Skill-rewarding. Room for luck. Produces a single defensible winner with no appeals. The winner had to be obviously the winner.

One ruleset. Solo-or-multi fair. Not cheatable. A single uncontestable champion the moment the timer hits zero.

The scoring system. Had to be transparent: players knew exactly what scored and when. Reward skill: the better player wins more often. Leave room for luck: somebody could break in against expectation. And make tied scores at the top mathematically impossible, so the winner could be crowned instantly.

Each account got a maximum of 20 total attempts across three days. The final score was the average of your three best attempts. That rewarded consistency over lucky one-shots. Reaching the top required hitting the ceiling three times.

Every mode used lower-mastery Fortnite mechanics (movement, pickaxing, environment interaction) rather than the shooting layer. Strategy decided it, not raw aim.

Per match, the maximum possible score was guaranteed equal across all matches. Some rounds had 200 points available, some had 100. Across ten rounds, the totals always summed to the same number. A perfect run always equaled the same score. But playing perfectly was extremely hard, and how to play perfectly was randomized per match. Nobody hit a perfect score across the event. The path was always visible. Execution was the wall.

How clean the resolution was. 12M unique participants. Zero contested results. No event-caused downtime. A single uncontested champion the second the timer ran out.

A creator-facing Unreal editor has roughly the same complexity as the professional one. You can't expose all of it. You also can't lock it down so hard that creators only build five flavors of the same game. Nothing new could break what already shipped. The cost of breaking an existing creator's island was always higher than delaying a new feature.

A toolset that let creators build high-fidelity, mechanically distinctive experiences without engine-level access. A roadmap that prioritized the tools with the highest creative leverage at each stage.

We could not ship everything creators asked for. Every cycle we picked which capabilities went next, knowing each addition meant a surface we'd have to maintain forever.

Stood up a dedicated internal team to dogfood the UGC tools. Two effects. One: the tools were built usable from the start, not retrofitted. Two: we had an informed internal team that could tell leadership, from first-hand experience, which tools would unlock the largest creator experiences. That second part made prioritization arguments actually useful.

How creative the community got at bridging gaps. The most interesting creator experiences often combined our primitives in ways our internal team hadn't anticipated. The dogfooding team learned more from watching creators ship than from any spec we wrote.