FSP is a fishing game I’m building in Unreal Engine 5.8, mostly solo. It was a test with procedurally generated landscapes and real water physics. The whole thing runs off data: every number in the game (fish weights, bite chances, prices) lives in a spreadsheet, gets checked by a validator, and is then loaded into the engine. I wanted to see if I could rebuild RF4 with today’s new game engines.
Getting there took a few false starts as always. I tried three different Unreal setups in four days, and every one fell over at the same spot: wiring my data through someone else’s kit. I brute force my way through it all with Claude and GPT. The project owns its core systems, and store-bought kits are only allowed to make things look nice. The one exception is the fishing plugin, which still handles the actual fight, because the feel was already right.
I run it like a small studio with strict habits. Code isn’t “done” until it builds. Balance gets tuned from real play numbers, not “by vibes”. Every change gets a second pair of eyes and has to come with proof. One example: players were reacting in 424 ms on average, so the hook window went from 500 to 1,000 ms without touching a line of engine code.
Specs
| Status | In development. First Validated Catch gate passed Jul 2, 2026 |
| Engine | Unreal Engine 5.8 |
| Languages | C++ (engine modules), Python (data tooling), CSV + JSON Schema (content) |
| Dates | Jun 28 – Jul 26, 2026 |
| Repos | 2 private: fsp (310 commits on main), fsp-unreal (271 commits on main) |
| Code | ~49,000 lines of project C++ across 8 Fishing* modules; ~10,800 lines of Python |
| Tests | 139 engine automation tests; 369 Python test functions; UI suite 83 passed / 0 failed |
| Content | 31 data tables, 16 invented species |
| Process | Separate build, data and review roles; evidence required on every PR; AI-assisted |
| Pipeline | CSV → validate_content.py → export → DataTable import → telemetry → balance report |
| Not yet | Playable demo map, backend, final display names, HUD art pass |