Probability Stability (“Sticky” Moves)
Probability stability—often described as “sticky” moves—is the tendency for a prediction market’s implied probability to stay near its new level after changing, rather than snapping back toward the prior level.
In other words, it answers a practical question in time-series prediction market data:
- Did the market learn something real (the move holds)?
- Or was it a temporary spike (the move reverts)?
A probability path can be volatile yet still show sticky moves (big jumps that persist), or it can be calm but fragile (small moves that keep reverting).
Why It Matters
Probability stability helps you treat prediction market probabilities as a signal:
- Signal vs noise filtering: Stable moves are more likely to reflect durable information; unstable moves are often microstructure noise, attention bursts, or thin liquidity.
- Alerting & monitoring: You can trigger alerts on moves that hold for a minimum window, reducing false positives.
- Model features & weighting: Persistence after a move can be used to weight markets/outcomes when building forecasting or decision systems.
- Market quality assessment: Markets that frequently revert after moves may be less informative (or simply less liquid).
“Hold” vs “Revert”: What It Means
A move is stable (sticky) when, after the probability changes, it remains close to the new level for a defined period.
A move reverts when the probability drifts back toward the pre-move level soon after the change.
Because “hold” and “revert” depend on your use case, stability is usually defined with two parameters:
- A time window (e.g., 10 minutes, 1 hour, 1 day)
- A tolerance band around the post-move probability (e.g., ±1–3 percentage points)
If the probability stays within the band for most of the window, the move is treated as sticky.
Common Causes of Sticky vs Non-Sticky Moves
Sticky moves are more common when:
- New information is credible and widely shared (the market updates and doesn’t look back).
- Liquidity is sufficient for price discovery (moves reflect real trading conviction, not a single small order).
- There is broad participation and fast arbitrage (mispricings get corrected and the new level becomes “anchored”).
Reverting moves are more common when:
- Liquidity is thin and small trades move the price disproportionately.
- Attention spikes fade (brief hype, rumor cycles).
- Participants disagree and the order flow mean-reverts.
How to Measure Probability Stability with the Prediction Markets API
With time-stamped probability series from FinFeedAPI’s Prediction Markets API, you can quantify stability in several practical ways:
- Post-move retention: After a threshold move (e.g., ≥5 points), measure how often the probability remains near the new level after T minutes/hours.
- Mean reversion score: Compare the post-move probability to the pre-move baseline; larger “pullback” implies less stability.
- Autocorrelation / half-life (advanced): Estimate how quickly deviations decay. Longer half-life implies stickier moves.
- Stability conditioned on liquidity: Evaluate whether moves are stickier when liquidity is higher (a common pattern).
Example
An outcome trades around 40% most of the morning. A verified announcement hits and the probability jumps to 55%.
- If it stays around 55% for the next hour (with normal fluctuations), that’s a sticky move and suggests the market incorporated durable information.
- If it quickly slides back toward 40%, the move likely reflected temporary flow, low liquidity, or overreaction.
Probability Stability and FinFeedAPI
Probability stability is easiest to analyze with consistent, historical probability data.
FinFeedAPI’s Prediction Markets API provides the time-series probability data you need to detect sticky vs reverting moves, build stability-aware alerts, and evaluate which markets produce more reliable probability signals.