Sentiment & Reinforcement Learning

andrewkni / bayesian-spike-detector

by Andrew Ni (andrewkni)

What it does

Bayesian model that distinguishes "fake" YES price spikes (low volume + widening spread) from "real" repricings (high volume + tightening spread), then fades the fakes. Enters NO when posterior mu > 0.7. Includes a CSV-based backtester tracking price/volume/spread deltas.

What we learned from it

the volume-and-spread joint signal is much stronger than either alone for distinguishing manipulation from real moves. This is the kind of strategy that works only because most participants react to price alone.

Don't want to maintain your own bot?

Bot for Kalshi is the no-code platform for automated prediction-market trading — visual builder, signals, encrypted credentials, kill switch. The infrastructure these open-source projects build, ready to use.

Start Free Trial →