Rift Oracle

• 

• 

• 

• 

• 

• 

• 

Inspiration

The draft phase in professional League of Legends is often described as the "silent battle", a chess match where games can be won or lost before a single minion spawns. Yet despite its critical importance, most teams rely on scattered spreadsheets, mental notes, and intuition. We asked ourselves: what if coaches had access to an intelligence system that could process thousands of historical matches and deliver actionable insights in real-time, turn by turn?

The GRID Data Platform offered something unprecedented: granular access to professional esports data. We saw an opportunity to transform raw statistics into strategic foresight, giving teams the analytical edge that separates champions from contenders.

What it does

Rift Oracle is a dual-interface drafting intelligence system that operates in real-time during the pick/ban phase:

Draft Simulator: A full 20-turn draft interface where coaches execute picks and bans with competitive timers, champion filtering by role and class, and an intuitive UI inspired by the professional broadcast experience.

Analysis Dashboard: A synchronized command center that delivers turn-by-turn recommendations, enemy pick predictions with probability distributions, live win probability tracking, and automated composition warnings (damage imbalance, missing roles, structural vulnerabilities).

The system also provides comprehensive pre-match scouting: player champion pools with individual win rates, team objective tendencies, historical match results, and head-to-head records between any two teams. Every recommendation comes with transparent reasoning—no black boxes, just data-driven insights coaches can trust and verify.

How we built it

We architected Rift Oracle on Next.js 15 with the App Router for server-side rendering and API routes, using TypeScript throughout for type safety across our complex data structures.

The data pipeline was our foundation. We integrated with GRID's GraphQL API to retrieve team rosters and series metadata, then leveraged the File Download API to bulk-extract end-state game files containing draft sequences, gold differentials, objective timings, and player performance metrics. A custom caching layer with TTL management ensures responsive performance during live scenarios.

The recommendation engine weighs multiple factors: signature champion proficiency, counter-pick historical performance, team synergy patterns, flex pick value, and deny potential. Win probability calculations integrate team historical performance, player-specific champion metrics, and objective control tendencies.

For the dual-page architecture, we built a custom state synchronization system using localStorage and BroadcastChannel adapters, allowing the draft and analysis interfaces to stay perfectly in sync—mirroring how professional teams operate with coaches and analysts on separate screens.

Challenges we ran into

Data normalization proved far more complex than anticipated. GRID provides incredibly rich data, but champion names, team IDs, and match structures required careful normalization across different API endpoints and historical data formats.

Real-time performance was critical—recommendations must appear instantly as each draft action occurs. We implemented request debouncing, abort controllers for stale requests, and aggressive caching to ensure sub-second response times even when processing extensive historical datasets.

Balancing complexity with usability challenged our UX decisions. The recommendation engine considers dozens of variables, but presenting that information without overwhelming users required iterative design. We settled on ranked recommendations with expandable reasoning, allowing coaches to dig deeper only when needed.

State synchronization across browser tabs without WebSocket infrastructure forced creative solutions. Our adapter pattern allows seamless switching between sync mechanisms while maintaining consistent behavior.

Accomplishments that we're proud of

We built a production-grade tool that could realistically be deployed in a professional team environment. This isn't a prototype—it's a polished system with error handling, loading states, and graceful degradation.

The transparent reasoning behind every recommendation sets Rift Oracle apart. Coaches see exactly why a champion is suggested: "High synergy with current frontline, 67% win rate against enemy mid-laner's champion pool, denies opponent's signature pick." Trust comes from understanding.

Our head-to-head analysis feature delivers genuine competitive intelligence—not just general statistics, but specific insights about what has worked in past meetings between these exact teams.

The synchronized dual-interface architecture genuinely reflects how professional teams operate, making Rift Oracle immediately applicable to real coaching workflows.

What we learned

Working with real esports data revealed the gap between theoretical statistics and actionable intelligence. Raw numbers mean nothing without context—understanding why a champion has a high win rate for a specific team requires deeper analysis of player comfort, team playstyle, and meta positioning.

We gained deep appreciation for draft theory itself. Implementing the recommendation logic forced us to formalize intuitive concepts: what makes a good flex pick, how to quantify "denying" value, when synergy outweighs individual champion strength.

Building for professional use cases demands a different standard. Speed, reliability, and trust are non-negotiable. Every feature must earn its place in the interface.

What's next for Rift Oracle

Live integration is our priority—connecting directly to tournament broadcast feeds to capture draft actions automatically, eliminating manual input during high-pressure moments.

Machine learning models trained on GRID's historical dataset could replace our heuristic-based recommendations with predictive models that identify non-obvious patterns in draft success.

Patch analysis automation would track meta shifts across game updates, automatically adjusting champion valuations as the competitive landscape evolves.

Team-specific training would allow organizations to fine-tune the recommendation engine based on their unique playstyle preferences and strategic philosophy.

Expanded analytics including player matchup matrices, draft tendency reports for upcoming opponents, and tournament-wide meta tracking would transform Rift Oracle from a draft tool into a comprehensive competitive intelligence platform.

The draft is where championships begin. Rift Oracle ensures teams enter that battle with every advantage data can provide.

Built With

• next.js
• tailwindcss