-
-
Daily matches schedule
-
Metrics and historical analytics
-
Leaderboard competition - Gamer vs AI
-
Live match analysis
-
AI metric semanatic analysis
Inspiration
Bet Studio was born from a desire to merge the nostalgic appeal of 8-bit retro gaming with the analytical power of modern AI. Our goal was to create a high-engagement PWA for football analytics where the complex world of data is simplified through a fun, vintage user experience. Furthermore, we wanted to challenge the status quo of cloud-dependent analytics which often suffers from latency and high operational costs.
What it does
Bet Studio is a Progressive Web App (PWA) that provides in-depth semantic and metric analysis of football matches through a sophisticated Hybrid AI model.
It allows users to: 1. View Advanced AI Predictions: Get data-driven insights with a retro graphical overlay. Our predictions result from combining server-side metric calculation with client-side semantic analysis. 2. Challenge the AI: Users submit their own predictions and compete against the AI's analysis for that day's matches. 3. Climb the Leaderboards: Points are awarded based on prediction accuracy, fostering weekly gamification and community engagement.
How we built it
The existing application utilizes a Hybrid AI model for maximum precision. Deterministic logic (metrics calculation, rule-based output) is managed server-side, while Semantic Analysis was previously outsourced to an external LLM.
The core technical achievement for this challenge was the critical migration of the semantic component to Chrome's Built-in AI APIs powered by Gemini Nano.
- Hybrid Architecture: The server (PHP/MySQL) continues to manage all Big Data processing, user authentication, and the competitive leaderboard (Deterministic Logic). This ensures data integrity and scalability.
- Core AI Shift (Cloud to On-Device): We successfully replaced the latency-prone OpenAI/cURL endpoint with a direct call to the Chrome Prompt API. The raw metric data is now passed to Gemini Nano to perform the complex semantic/stochastic analysis (e.g., generating the narrative analysis and risk assessment) directly on the user's device.
- PWA/TWA Standard: The entire application is built as an installable PWA/TWA, ensuring a seamless, native-app experience across Chrome platforms.
Challenges we ran into
The primary challenge was prompt engineering for the local model. Gemini Nano is optimized for on-device speed, but we had to carefully fine-tune our detailed system prompts to ensure the small, local model could maintain the sophisticated analysis quality previously delivered by larger cloud models. This required intensive testing to maximize performance under local memory constraints.
Accomplishments that we're proud of
- Achieving sub-100ms response times for core analytical queries by eliminating network latency via Gemini Nano.
- Successfully transforming a core operating cost (Cloud API fees) into a user benefit (speed and unrivaled privacy), as data analysis never leaves the browser.
- Creating a unique, fully functional PWA that successfully marries the 8-bit aesthetic with advanced, cutting-edge on-device AI.
What we learned
We learned the immense potential and limitations of local models. Specifically, we gained deep insight into optimizing prompts and data inputs for Gemini Nano to maximize analytical depth without sacrificing the speed essential for a real-time analytics platform. This shift has redefined our approach to client-side data processing.
What's next for Bet Studio: ⚽ Retro Game Analytics AI (PWA & Chrome AI)
We plan to expand the use of other Chrome AI APIs (like Summarizer or Rewriter) to enhance the user's personalized summary reports. Our next step is to finalize the launch of the TWA version on the Google Play Store and continue promoting the benefits of private, cost-free, on-device AI to our growing user base.
Log in or sign up for Devpost to join the conversation.