⌘K
How we identify potentially suspicious trading activity on prediction markets
Our insider detection algorithm analyzes trading patterns to identify potentially suspicious activity before markets resolve. We use a multi-factor scoring system that considers various behavioral signals, similar to how credit rating agencies evaluate risk.
Like Moody's or S&P, we share our general approach and the factors we consider, while keeping the exact weights and thresholds proprietary. This allows you to understand how we work while preventing bad actors from gaming the system.
Our algorithm evaluates multiple signals when assessing whether a trade might indicate insider knowledge. No single factor is definitive—we look for patterns across multiple dimensions.
Large bets relative to the trader's typical position sizes or overall trading history can indicate high conviction based on non-public information.
New wallets making outsized bets are more suspicious than established traders with long histories. Fresh wallets are often created specifically for insider trades.
Positions taken close to market resolution, especially combined with other signals, warrant additional scrutiny. The closer to resolution, the higher the signal strength.
Contrarian bets at extreme prices (buying at 10¢ or 90¢) that turn out to be correct suggest the trader knew something the market didn't.
Multiple wallets betting the same direction around the same time may indicate coordinated activity. We analyze wallet correlation patterns to detect potential rings.
Wallets we've flagged before that turned out to be correct get increased scrutiny on future trades. Repeat offenders face higher suspicion scores.
A wallet flagged in multiple different markets is more suspicious than one flagged only once. We aggregate signals across a trader's entire activity.
Sudden changes in trading behavior—like a trader who usually bets $5K suddenly betting $50K—are flagged for review compared to their historical patterns.
Detection is only half the equation. After markets resolve, we track whether our flagged wallets actually made the correct bet. This feedback loop is critical for validating our methodology.
We continuously refine our detection based on real outcomes:
Current Status: We are actively collecting outcome data to calculate our detection accuracy. As markets resolve and we gather more data, we will publish our hit rate and notable case studies.
We track significant trades ("whale trades") across prediction markets to provide real-time insights into smart money movement. Here's how we determine what qualifies as a whale trade.
We prioritize markets by trading volume to focus computational resources where whale activity is most likely:
| Coverage | Markets | % of Total Volume |
|---|---|---|
| Top 100 | 100 | 71.5% |
| Top 200 | 200 | 89.2% |
| Top 300 | 300 | 96.7% |
| Top 500 (tracked) | 500 | 99.8% |
Markets ranked 501+ represent just 0.2% of total volume. A $5,000 trade in a market with $50,000 total volume would be 10% of the entire market—statistically improbable.
We define "whale trades" using percentile-based analysis of actual trade data:
| Percentile | 7-Day | 30-Day |
|---|---|---|
| P50 (median) | $1,015 | $1,243 |
| P75 | $1,995 | $3,216 |
| P90 (threshold) | $4,634 | $5,535 |
| P95 | $6,583 | $9,015 |
| P99 | $17,582 | $17,501 |
Our $5,000 threshold corresponds to approximately the 90th percentile (P90)—meaning we track the top 10% of trades by size. This is a defensible, data-driven definition of "significant" trading activity.
Update Frequency: Whale trades are synced every 60 seconds from venue APIs with a 7-day lookback window.
In the interest of transparency, here's what we deliberately don't disclose:
This approach mirrors how credit rating agencies operate—they publish methodologies explaining what factors matter, without revealing the exact formulas. It builds trust while maintaining detection effectiveness.
Our methodology is designed to catch obvious patterns, but no detection system is perfect. We're transparent about our limitations:
We're continuously working to improve detection while minimizing false positives. Our outcome tracking helps us iterate on what works.
See which wallets our algorithm has flagged and track their outcomes.