Fast Break — 2v2 Arcade Basketball
Unity · C# · Service Locator · GAS-inspired abilities · event bus
- Build window
- 5 days
- Service layer
- 8 interfaces
- Scope
- 2v2 local multiplayer
Problem
The brief: build the basics of a 2v2 local multiplayer arcade basketball game. Two controllable players on one team, two static opponents, pass and shoot. Then answer extension questions — full shooting, score tracking, AI, online multiplayer — as if the project were going to continue.
Constraint
Five days. The extension questions were the real test: if answering “add AI” meant rewriting the game loop, the architecture had already failed.
Decision
I designed the codebase around a Gameplay Ability System philosophy before writing gameplay:
- Service Locator with dependency resolution — services register interface→concrete bindings and declare their dependencies. A topological sort orders
Initialize / PostInitialize / Shutdown, with circular-dependency detection. - Interface-based service layer (8 services) — an abstract
GameManagerwithLocalMultiplayerGameManagerServiceas the concrete implementation, plusTeamService,EntityPositionService,EntitySpawnerService,PlayerInputManagerService, and a taggedGameplayEventServiceevent bus. - Everything gameplay-facing goes through those interfaces. The manager spawns players and opponents, joins input, assigns teams, and spawns the ball — it does not own the rules.
Verification
The acceptance mechanism is the extension path itself: full shooting, score tracking, AI, and online play fall out as new services and abilities rather than rewrites. That was the claim the take-home was asking me to prove, so it is the claim I designed for.
Result
A prototype that grew from three stubs into a data-driven, GAS-inspired architecture inside five days.
The transferable lesson: a five-day limit is a good architect. It forces you to separate what changes from what stays.