Conditional Probability Modeling for Short-Horizon Bitcoin Binary Markets
Summary
The document presents a framework for estimating the probability that Bitcoin will finish above a round’s benchmark, conditional on its current price distance, time remaining, volatility, and recent drift. It separates this forecast from Polymarket prices, then compares the forecast with executable order-book VWAP after fees and slippage to assess expected value. Volatility blends realized estimates over several time windows, while drift is shrunk toward zero and capped; early-round estimates receive reduced confidence.
It also discusses OpenMarket’s synchronized Binance and Polymarket dataset and reports that its public feature model did not consistently beat the market-implied probability out of sample or show a clear edge after costs. The proposed system stores features and settlement labels for later stage-specific calibration and validation. The author recommends paper trading and data collection until probabilities and executable returns are supported by out-of-sample evidence. The described probability baseline is a transparent starting point, not proof of a profitable strategy; it may also miss jumps, changing regimes, and execution risks.
Key ideas
- Estimate the terminal event probability from current market conditions rather than forecasting direction alone.
- Compare the probability with executable VWAP after fees and slippage before entering a trade.
- Keep forecast inputs separate from the prediction-market order book to avoid circular reasoning.
- Blend volatility across multiple horizons, shrink drift, and reduce confidence when information is sparse.
- Validate calibration and executable returns out of sample by time stage before using complex features live.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.