The Insider Detection tool flags trading patterns that may indicate informed trading — that is, trades placed by participants who appear to have non-public information. It does NOT confirm insider trading; it surfaces anomalies that deserve investigation.
## The 5 Anomaly Types
Predite's insider detection runs five separate models, each looking for a different signature.
### 1. Fresh Wallet Large Bet
Pattern: A wallet created in the last 7 days places a single bet larger than $5,000.
Why it's suspicious: Legitimate large bettors typically have a history. A brand-new wallet showing up with significant size has only a few plausible explanations:
- Someone tunneling funds from another platform to bet anonymously
- A test trade by an institutional player
- Someone acting on time-sensitive information they didn't want associated with their existing identity
False positive rate: ~40%. Many fresh wallets are legitimate (new users to Polymarket). The flag is most informative when combined with other signals.
### 2. Pre-News Accumulation
Pattern: Multiple wallets accumulate substantial positions in a market within 2-6 hours before a major news event resolves that market.
Why it's suspicious: If the news was truly unexpected, no one should have been able to position ahead of it. Pre-news accumulation suggests someone knew.
Detection method: We index timestamps of major news events against position-change snapshots. When position size grows >50% in the 6 hours before a relevant headline, we flag it.
False positive rate: ~25%. Some markets attract attention from informed observers (e.g., professional traders pricing in expected announcements) even without true insider information.
### 3. Unusual Volume Spike
Pattern: A market's volume in the last hour exceeds 5x its 24-hour average, without an obvious news catalyst.
Why it's suspicious: Volume usually correlates with news. Volume without news = trading on a non-public signal.
Detection method: We compute rolling 1h/24h volume ratios for every market. Anything above 5x with no matching news in the last 30 minutes is flagged.
False positive rate: ~35%. Markets sometimes spike on social-media virality that hasn't yet hit news feeds. Worth investigating but not auto-trading on.
### 4. Cluster Coordinated Buy
Pattern: Multiple wallets identified (by our clustering algorithm) as controlled by the same entity all buy the same direction in a market within a short time window.
Why it's suspicious: A single entity spreading purchases across wallets suggests they're trying to avoid detection. Doing so simultaneously suggests urgency — they want their full position on before the price moves.
Detection method: Predite runs a graph-clustering algorithm over Polymarket wallets, linking addresses that share deposit sources, withdrawal destinations, or transaction timing patterns. When multiple wallets in the same cluster place same-direction bets within a 30-minute window, it flags.
False positive rate: ~30%. Clustering isn't perfect. Some "clusters" are actually legitimate sub-wallets of the same individual (totally legal, just hiding from price-impact).
### 5. Resolution Front-Running
Pattern: Large new positions opened within 24 hours of a market's official resolution.
Why it's suspicious: As resolution approaches, the outcome becomes more certain. Late-stage large bets in one direction often mean the bettor has confirmed information that the resolution will go that way.
Detection method: We index time-to-resolution for every market. Trades over $5000 occurring in the last 24 hours before resolution are flagged.
False positive rate: ~50%. Many late-stage trades are legitimate hedging or arbitrage. This flag is informative only when combined with other signals.
## Severity Levels
The Insider page combines anomaly counts into severity ratings:
- 🔴 **Critical:** Multiple anomaly types matched on the same market within the same time window. Highest-confidence flag.
- •🟠 **High:** Single strong anomaly (e.g., fresh wallet + $50k bet + small market).
- •🟡 **Medium:** Single moderate anomaly.
- •⚪ **Low:** Below threshold, not displayed by default.
The page shows Critical and High by default. Use the severity filter to see Medium and Low.
## How to Use Insider Flags
Do:
- Investigate manually before acting. Read the market, check news feeds, look at the wallet's history.
- Use flags as a tiebreaker, not a primary signal. "I was already considering this trade; insider flag pushes me to take it."
- Treat flags as evidence of information asymmetry, not necessarily corruption. Many flags are legitimate informed traders.
Don't:
- Blindly trade in the direction of the flagged wallet. They might be wrong. Insider trading also catches scams (people pretending to have information they don't).
- Treat the flag as a guaranteed signal. Predite's false-positive rates are documented above for a reason.
- Use this data to publicly accuse anyone of insider trading. The flags are statistical patterns, not evidence.
## Privacy and Ethics
A note on what we do and don't share:
- Predite **does not publicly identify** flagged wallets. We show the trade pattern and let you investigate.
- •We **do not share** insider flag data with Polymarket, Kalshi, or law enforcement. The flags are for your trading awareness.
- •We **do not** claim to detect actual insider trading. The flags are anomaly detection, not legal determination.
If you have evidence of actual insider trading (e.g., a tip-off), report it to the platform directly (Polymarket has compliance contact) or to the relevant securities regulator.
## Related Docs
- [Whale Tracker](/docs/whale-tracker)
- •[Trader Profiles](/docs/traders)
- •[Reading Signals](/docs/reading-signals)