← MARBLE // STUDIOS
POLYEDGE.SYSTEM
LIVE — COLLECTING DATA
Automated edge discovery engine for binary prediction markets. We collect high-frequency snapshots across 2,000+ Polymarket markets, test statistical hypotheses for pricing inefficiencies, and validate findings through paper trading before any capital is deployed.
HYPOTHESES UNDER TEST
H1 — Mean Reversion
Markets that deviate sharply from equilibrium tend to revert. Testing entry/exit on >15% moves with trailing stops.
▸ +5.26% avg return, 78% win rate (paper trading)
H13 — Post-Event Overshooting
Markets overshoot after news events, creating short-lived pricing errors exploitable within minutes.
▸ 71.8% avg overshoot, 88.9% occurrence rate — STRONGEST SIGNAL
H15 — Volatility Cones
Comparing realized vs implied vol across timeframes to identify mispriced risk.
▸ Insufficient data — re-testing at 7 days
H2 — Liquidity Clustering
Order book density at psychological price levels (5¢, 10¢, 50¢) creates predictable support/resistance.
▸ 21.4% clustering at 5¢ boundary — under observation
H3 — Calendar Effects
Volume and volatility patterns tied to time-of-day and day-of-week across market categories.
▸ Data collection phase
H10 — Expiry Convergence
How markets converge to binary outcomes in final hours. Speed and path of convergence as tradeable signals.
▸ Data collection phase
DATA MILESTONES
Day 7 — Basic hypothesis testing, H15 re-test
Day 14 — Paper trading statistically meaningful
Day 28 — Statistical validity threshold
Day 42 — Walk-forward backtesting, strategy finalization
STACK
PythonDuckDBFastAPINext.jsPolymarket CLOBGamma APIsystemd timers