KIMI_CLAW_CHATBOT_2026-02-17

What are the 81 algorithms? — KIMI runs 81 different trading strategies simultaneously — like 81 expert traders each with their own style. Some watch for whale wallet movements, others track funding rates, some follow momentum, others look for oversold bounces. Each algorithm scans the market independently every 15 minutes.

What is the ML Signal Ranker? — Not all 81 algorithms are equally good. The ML Signal Ranker uses a Random Forest model to learn which algorithms are actually winning trades in the current market. Think of it like a sports coach who tracks which players score the most goals, then gives them more playing time.

What is Random Forest? — Imagine you ask 100 different decision trees (like flowcharts) to vote on whether an algorithm's signal will win. Each tree was trained on slightly different data, so they each have a unique perspective. The majority vote becomes the prediction. It's called a "forest" because it's many trees working together.

Heuristic vs ML mode: — KIMI starts in heuristic mode (rule-based scoring using win rate + Sharpe ratio). Once 50+ trades are closed, it automatically trains the Random Forest model and switches to ML mode — learning from real results instead of just rules. Current status: Heuristic (accumulating trade data).

How does ranking affect picks? — Each algorithm gets a weight from 0.1 (weak) to 0.75 (strong). Higher-weighted algorithms get larger position sizes. Algorithms that keep losing get their weight reduced, and eventually get eliminated from the competition entirely.

The Elimination Engine: — Algorithms that perform badly go through: Active → Danger Zone (3 failures) → Probation (5 failures) → Eliminated. New challenger algorithms can replace eliminated ones. It's survival of the fittest.