Street Fighter Underground Project Report
1. The Concept
Street Fighter Underground is a web-based 2.5D fighting game where players battle through an illegal underground tournament across unique arenas including subway stations, neon alleys, rooftops, warehouses, underground clubs, docks, factories, construction sites, championship arenas, and final boss stages. Each level introduces different opponents, AI behaviors, environmental hazards, rewards, and increasing combat difficulty.
The project was chosen as a one-day sprint to evaluate how well modern AI tools could collaborate to build a complete action game—from planning and game design to combat systems, UI, animations, progression, and gameplay logic—using only HTML, CSS, and JavaScript without external frameworks.
---
2. The Stack
The Stack: HTML5, Vanilla CSS, and Vanilla JavaScript.
The AI Assist: ChatGPT handled the project planning, gameplay systems, UI/UX design direction, combat mechanics, level progression, AI behavior, and technical documentation. OpenAI Codex was responsible for implementing the game, generating source code, integrating assets, building combat systems, creating menus, and connecting the gameplay mechanics based on the provided specifications.
---
3. The 1 Day Sprint
#
Morning (Hours 1–3)
The project started by defining the overall vision of a premium mobile fighting game. We designed the Home screen, Tournament system, fighter progression, combat controls, HUD, reward system, and level structure. Multiple iterations were spent refining the UI layout and creating a complete gameplay flow before development began.
#
Afternoon (Hours 4–6)
The development phase focused on implementing the fighting engine. Core systems included movement, attacks, blocking, dodge, jump, EX meter, combo logic, enemy AI, collision detection, animations, level progression, and reward management. Considerable time was spent refining gameplay because the initial implementation resulted in static fighters, weak AI behavior, repeated levels, and inconsistent combat animations.
#
The Finish Line (Hours 7–8)
The final hours were dedicated to bug fixing, balancing, and polishing. We improved the tournament flow, level unlocking, progression saving, UI responsiveness, combat effects, hit reactions, animation timing, and overall user experience. Although the game became fully playable, additional polishing remains necessary to reach commercial-quality standards.
---
4. The Roadblocks
The largest challenge throughout the project was translating high-level game design into working gameplay. While the planning phase defined a complete fighting game system, Codex frequently produced incomplete implementations that required multiple refinement passes.
Several issues repeatedly appeared during development:
- Combat animations remained static despite implemented state logic.
- Enemy AI often stood idle or attacked without realistic timing.
- Level progression occasionally restarted the same stage instead of advancing.
- Generated UI layouts did not consistently match the intended premium mobile design.
- Character artwork required repeated regeneration due to cropping, scaling, and positioning issues.
- Maintaining responsive 9:16 layouts while preserving visual quality required significant manual iteration.
Much of the development time was spent improving gameplay feel rather than adding new features.
---
5. Key Takeaways
#
What worked best?
The planning process proved extremely effective. ChatGPT rapidly generated a complete game design covering combat mechanics, AI systems, progression, UI hierarchy, level structure, reward systems, and technical architecture. This provided a strong foundation before implementation began.
#
What worked worst?
The implementation phase required significant iteration. Codex successfully generated the project structure and core gameplay systems but often struggled with premium UI composition, animation quality, gameplay polish, AI behavior, and complex progression logic. Achieving smooth combat required repeated debugging and continuous prompt refinement.
My final thought:
If I had one additional day, I would focus entirely on gameplay polish. The highest priorities would be improving combat responsiveness, creating smoother sprite animations, expanding enemy AI with more advanced behaviors, adding cinematic visual effects, improving audio feedback, and further refining the Home screen into a commercial-quality mobile game interface. While AI significantly accelerated development, creating a polished fighting game still required extensive testing, iteration, and manual direction to achieve satisfying gameplay.