Data & Methodology

Probable analyzes how prediction markets behave as information emerges. This page explains the data we use and how our insights are formed.

Data Sources

Probable analyzes publicly available prediction markets, including Kalshi and Polymarket. We focus on active, open contracts where market behavior is visible through participation, pricing, and trading activity.

• • Kalshi: CFTC regulated exchange dedicated to trading on the outcome of future events.
• • Polymarket: Prediction market platform where users trade on the outcomes of real-world events.

To keep insights meaningful, we exclude markets that are inactive, closed, or lack sufficient trading activity. Our goal is not to cover every possible contract, but to analyze markets where signals are most informative.

What We Analyze

Rather than predicting outcomes, Probable looks at how markets respond. Our analysis focuses on three core elements:

• Participation How many traders are active, how broadly activity is distributed, and how attention shifts across related markets.
• Price Behavior How stable or volatile prices are, how smoothly they move, and how responsive they are to new information.
• Agreement Across Markets Whether related markets tell a consistent story or imply conflicting views across scenarios.

Together, these patterns help us understand how confidently and coherently markets are expressing information at a given moment.

Mood Labels & Rail

The headline mood (Calm, Stable, Focused, Cautious, Active) is based on a composite mood index from 0-100. When both sources are available, Probable blends Kalshi and Polymarket using source freshness and market coverage. If one source is unavailable, the available source is used.

These labels describe current market behavior and tempo, not a forecast.

• • Calm (0-34): slower repricing and lower urgency across active markets.
• • Stable (35-51): balanced flow with orderly price adjustments.
• • Focused (52-66): attention concentrated in a smaller set of themes.
• • Cautious (67-81): defensive positioning with higher sensitivity to new information.
• • Active (82-100): fast repricing and rapid rotation of attention.

Bias and Conviction Map

Each point is a time-stamped market snapshot. The x-axis shows directional bias from bearish to bullish, and the y-axis shows conviction strength from lower to higher conviction. Dot color reflects heat (speed of repricing/rotation): Low, Elevated, or High.

Conviction here means how consistently related markets are positioned and how persistent that positioning is over time. It does not mean an outcome is more likely to be correct.

The source toggle (Kalshi / Polymarket / Both) switches the plotted series. The 5h, 1d, 2d, 3d, and 7d controls highlight selected windows, with source-specific ellipses showing clustering over each interval.

Breadth and Rotation Chart

This chart shows two normalized indices over time:

• • Breadth (0-100): how diversified conviction is across active markets.
• • Rotation (0-100): how quickly top markets and attention change.

It uses the same source toggle and 5h/1d/2d/3d/7d window controls so users can compare short-term versus broader behavior.

“Markets Drawing the Most Attention” List

The list includes only active/open markets with non-extreme current pricing (between 0.5% and 99.5%) and non-zero recent trading activity. Rows are sorted by impact, then volume.

Conviction is derived from probability level and recent movement. These are behavior buckets around current pricing skew, not outcome predictions:

• • High: probability ≤5% or ≥95% (very one-sided current pricing).
• • Strong: probability ≤20% or ≥80% (clearly tilted pricing).
• • Moderate: all other active probabilities (less one-sided pricing).

Δ24h Driver Tags

• • Rising: shown when a market’s probability increased by more than 5 points in the last 24 hours.
• • Fading: shown when a market’s probability decreased by more than 5 points in the last 24 hours.
• • For Kalshi, Probable computes the 24h baseline from stored hourly snapshots matched by market ticker.
• • For Polymarket, Probable uses source-provided one-day change fields where available.
• • No tag: either the move stayed within ±5 points or a 24h baseline was unavailable for that row.
• • These tags describe recent repricing speed, not expected outcome direction.

Impact starts as a numeric score based on potential move size weighted by participation (volume/liquidity). Display tiers are assigned relative to the category’s impact distribution (quantile bands) when sample size is sufficient, with source-specific fallback thresholds otherwise.

Insights & Signals

Probable surfaces insights such as market confidence, fragility, and disagreement by combining multiple behavioral signals. These insights describe market structure and behavior, not correctness or likelihood.

Signals are designed to answer questions like:

• • Is participation broad or thin?
• • Are prices stable or easily moved?
• • Do related markets agree, or are signals fragmented?

They are descriptive, not prescriptive.

Time & Updates

All insights on Probable are time-stamped snapshots. They reflect how markets were behaving at the moment they were evaluated and may change as new information arrives or participation shifts.

Markets are dynamic, and insight stability matters. For this reason, Probable updates insights periodically rather than continuously, favoring clarity and context over rapid fluctuation.

What Probable Is — and Is Not

Limitations

Prediction markets reflect the views of their participants — not the full universe of opinions. Participation levels, liquidity, and market design can influence prices and signals. Probable aims to make these dynamics more visible, not to remove uncertainty.

Our Approach

Nobody can predict the future. Probable is built on the belief that how markets behave is often more informative than where prices happen to land.

Our goal is simple: clearer context for reading markets as they react to uncertainty.