Before the match
Build a focused GamePlan.
Go in with clearer intent.
Your League of Legends learning loop. Prepare with a GamePlan, stay grounded in the facts, and review what matters after the match.
ONE CONTINUOUS LEARNING LOOP
Build a focused GamePlan.
Go in with clearer intent.
Stay grounded in the facts.
Keep the decisions in your hands.
Review what matters.
Turn games into practice.
INSIDE THE MVP
A Windows prototype connecting preparation, factual live data and post-game review.

01 / BEFORE THE MATCH
Your champion, role, visible picks and bans in one place. Read-only Champion Select context, with hidden information kept hidden.
Available in the MVP

02 / PREPARE WITH CONTEXT
Carry your preparation from Champion Select into the match: composition context, supported skill plans, power windows and role-specific considerations.
Available in the MVP · seeded rules

03 / PLAN YOUR OPENING
Compare opening routes and conditional options around farming, resets, ganks, invades and objectives. Keep the final decision in your hands.
Available in the MVP · sensor-dependent timers

04 / STAY GROUNDED IN THE FACTS
A lightweight view of your own champion, level, game time, gold, K/D/A and CS. Information for awareness, with gameplay remaining under your control.
Basic overlay available in the MVP

05 / REFLECT AND IMPROVE
Review a locally recorded match summary and mode-aware observations. Start with a concrete moment to reflect on before your next game.
Basic review available · deeper coaching planned
An early working prototype, with an honest view of what is still in development. Expand a feature for details.
The Windows prototype reads Champion Select through the local League Client API. It shows assigned role, champion selections, visible team composition, bans and the selection timer. The MVP does not auto-pick, auto-ban or reveal deliberately hidden players.
What’s next. Broader champion coverage and richer preparation are planned. Current composition profiles use explicit seeded rules, not statistical win-rate predictions.
A persistent GamePlan keeps pre-match context through loading and into play. Partial plans are enriched when the game client makes more information available. Champion coverage is shown explicitly; unsupported information stays unknown.
What’s next. Data-driven rune options and build branches are planned, pending a validated data pipeline. The MVP visibly marks these fields as dataset pending.
The planner includes opening routes, power windows and multiple branches with reasons and conditions. The runtime uses known player facts such as level, inventory and unspent gold. It labels camp information as exact, estimated or unavailable.
What’s next. Exact camp timers require a supported source and an observed event. The standalone Live Client feed does not provide individual basic-camp deaths. The sensor bridge has been exercised with simulated data; production Overwolf integration and distribution remain pending.
The desktop prototype reads the local Live Client Data API and includes a small factual overlay. Pre-game planning and your own performance data form the basis of the live experience. The image shows the MVP’s Live panel; the current overlay uses a smaller layout.
What’s next. A modular native in-game overlay is planned. Hidden enemy information, enemy ability or summoner-spell cooldown tracking, automated gameplay and single-action commands are outside the intended product scope.
The current coach provides rule-based findings from local match snapshots. It applies different rules for Summoner’s Rift, ARAM and other modes, rather than imposing the same targets on every role or queue.
What’s next. A detailed decision timeline, richer alternative-choice analysis, personalized goals and long-term AI coaching are planned. Account-V1 and Match-V5 integration for opt-in identity and the player’s own history depends on Riot approval and implementation.
Public download is not available yet. Riot review, production API access and Overwolf integration are pending. This website presents the prototype and its development direction.
BUILT AROUND PLAYER AGENCY
Preparation before the match. Factual information during play. Reflection after it.
Designed to support learning without hidden information, automated gameplay or commands that replace your judgement.
Independent project. No Riot endorsement or approval is claimed. Features remain subject to the Riot developer policies and League of Legends policies.
CONTACT
Product questions, review enquiries or privacy requests.
Email the developer mimirrift.app@outlook.comProject contact: Mimir Rift developer