Paper-trading research feed. Not investment advice. No real capital is traded on your behalf.
Model vs Market Odds
Edge = Darwin model probability minus Polymarket YES price. Positive edge = model more bullish than market. Not financial advice.
Direction Accuracy
1h predictions logged hourly · 4h every 4 hours · 1d at midnight UTC. Each resolves against actual price after its timeframe elapses. Accuracy = correct directional calls ÷ total resolved.
How Honest Is the Engine?
When the engine says “70% confident,” it should be right about 70% of the time. Bars below actual accuracy mean the model is over-confident. Above = under-confident (conservative).
Not enough data for calibration analysis.
Each bucket compares stated confidence (X) vs actual accuracy (Y). A perfectly calibrated model traces the diagonal. Over-confidence shows below the line, under-confidence above.
Polymarket Simulation · 3 × $10k Wallets
3 independent $10k virtual wallets — one per timeframe. Kelly Criterion bet sizing (half-Kelly, ≤10% balance). Even-money payout. Min 55% confidence to bet. Settles after each timeframe elapses.
Darwin Engine
A self-evolving prediction engine, paper-traded. Currently pre-alpha — the model is still learning. Research output only; not financial advice.
Engine parameters evolve automatically via Bankr LLM (Claude Opus) when win rate drops below 80%. Pre-alpha — not financial advice.
