timesPM
Find mispriced prediction markets.
We analyze full price history and surfaces 24-hour forecasts.
TRY “”, “”, “”, “” — ANY ACTIVE POLYMARKET EVENT
Find a market
Search thousands of active Polymarket events by keyword. Results are ranked by 24-hour trading volume so you see the most liquid markets first.
Run the forecast
We pull the complete hourly price history from Polymarket's CLOB and feed it into timesPM, our time-series forecasting engine.
Get the verdict
The model returns a 24-hour price forecast with 80% and 90% confidence intervals. If the delta is large enough and the CI is tight, you get a signal.
24H PRICE FORECAST
Hour-by-hour predicted YES price for the next 24 hours, generated by timesPM from the market's full trading history.
CONFIDENCE INTERVALS
Every forecast comes with 80% and 90% confidence bands so you can see how certain the model is — not just where it thinks the price is going.
ACTIONABLE SIGNALS
When the forecasted move exceeds 5% and the 80% CI width is under 20%, timesPM flags it as a BUY YES or BUY NO signal with the exact mispricing delta.
INTERACTIVE CHART
Visual overlay of recent price history and the forecast with shaded confidence bands. See exactly where the model diverges from the current market price.
Not every forecast produces a signal. timesPM requires two conditions to be met simultaneously before it flags a trade:
CONDITION 1 — MAGNITUDE
Δ > 5%
The absolute difference between the current YES price and the 24-hour forecast must exceed 5 percentage points. Smaller moves aren't worth the spread.
CONDITION 2 — CONFIDENCE
CI < 20%
The width of the 80% confidence interval must be under 20 percentage points. Wide intervals mean the model isn't sure enough to act on.
Polymarket CLOB API
Full hourly price history fetched in 14-day windowed chunks. Deduplicated and aggregated to clean hourly candles.
timesPM
Foundation model pre-trained on 100B+ time-series data points. Automatically selects long-horizon mode for markets with 720+ hours of data.
24 hours ahead
Hourly granularity predictions with 80% and 90% confidence levels. Outputs are clamped to [0, 1] to respect probability bounds.
Forecast quality depends directly on how much history the model has to work with. Markets that have been live for 30+ days with consistent trading activity produce the tightest confidence intervals and most reliable signals.
New markets or those with sparse trading will show wider confidence bands — the model is honestly telling you it doesn't know. timesPM won't flag a signal unless both the magnitude and confidence thresholds are met.
Some markets are inherently news-driven — elections, court rulings, policy announcements. Price moves in these markets are reactions to discrete events that no time-series model can anticipate. timesPM only sees the price curve, not the news behind it, so forecasts on event-reactive markets should be taken with extra caution.
This is a research tool, not financial advice. Prediction markets carry risk and model forecasts are probabilistic, not deterministic.
WANT DEEPER SIMULATIONS?
MiroShark is our multi-agent simulation engine. Upload any document and it generates hundreds of AI agents that simulate public reaction on social media — posts, arguments, opinion shifts — hour by hour. Run it locally or with any cloud API.
MIROSHARK ON GITHUB ↗