Quant research · tooling
BTC Threshold Probability Dashboard
Interactive dashboard estimating short-horizon probabilities that Bitcoin crosses a price threshold, with a backtester for Kalshi-implied edges.
- Year
- 2026
- Status
- Working prototype
- Stack
- Python · Streamlit · scikit-learn · pandas
Kalshi lists yes/no contracts on whether Bitcoin will be above a given price at a given time — which makes them, in effect, tradeable probabilities. The question this project asks: are those implied probabilities ever measurably off over short horizons?
Model, then compare
A KNN model estimates the probability that BTC crosses a chosen threshold within a chosen horizon, using features built from recent price action. A Streamlit dashboard makes the model interactive — pick a threshold, pick a horizon, see the estimate — and a backtester replays history to compare model probabilities against market-implied ones, flagging where an edge would have existed.
What I would stress-test before trusting it
- Regime shifts — a KNN fitted on calm markets says little about violent ones.
- Fees, spread, and slippage — a paper edge has to survive the costs of being real.
- Calibration — a model that says 70% should be right about 70% of the time, and that deserves its own test.