Drift Radar
Paste a Polymarket URL. Get a probabilistic forecast of the YES-price path to resolution, then hand it into Kelly or EV sizing.
This is a statistical analysis of the price series, not a prediction of the event outcome. See the methodology page for how we compute and validate forecasts.
Find exact Polymarket contract
Search Polymarket and choose the specific contract Drift Radar should forecast.
How to read the forecast
The solid line shows the historical YES price, hour by hour.
The shaded cone is the forecast range. The top edge is the 90th percentile (where the price might end up if the trend strengthens), the bottom edge is the 10th percentile (if it reverses), and the line inside is the median.
Driftis the gap between the current spot price and the model's median at resolution. Positive drift means the model thinks the price is heading up. The bigger the drift, the more the model disagrees with the current market.
Confidence reflects the width of the cone relative to spot price. A narrow cone means the model is confident; a wide one means either the market is volatile or the history is short.
Drift Radar FAQ
- What is Drift Radar?
- Drift Radar uses Google's TimesFM 2.5 time-series foundation model to extrapolate the price path of a prediction market contract. Given the historical YES-share price of a Polymarket contract, it returns a probabilistic forecast (10th, 50th, and 90th percentiles) to the market's resolution date.
- Is this a prediction of whether the event will happen?
- No. It is a statistical analysis of the price series, not a forecast of the underlying event. The distinction matters: we are extrapolating market behavior, not modeling the world. Use it as one input alongside your own analysis, never as an oracle.
- What can I do with the forecast?
- Take the model's median forecast at resolution and use it as 'your probability' in the Kelly or EV calculator. If the model sees meaningful drift from the current spot price, and you agree with the model's reasoning, the Kelly fraction estimates a position size. If you disagree with the model, trust your own analysis.
- How accurate is TimesFM on prediction market prices?
- TimesFM was not trained specifically on prediction markets, so treat its output as a statistical baseline, not a proven edge. We track accuracy over time via the Calibration Tracker: for every forecasted market that resolves, we record the median prediction and the actual outcome. Accuracy data is published on the methodology page.
- Why only Polymarket? What about Kalshi?
- v1 is Polymarket-only because Kalshi is CFTC-regulated and the compliance framing is different. We will add Kalshi in v1.1 after validating the forecaster on Polymarket data first.
- Is this financial advice?
- No. Drift Radar is a statistical tool, not investment advice. Prediction markets involve substantial risk including total loss of capital. You are solely responsible for every decision you make.