A leaderboard is a discovery surface, not a recommendation engine. High recent PnL can result from one concentrated position, unusual risk or an asset that cannot support another trade at the displayed price.
Compare several windows
Query 24-hour, 7-day and 30-day rankings. Reward consistent participation and treat an extreme short-term result as a prompt for investigation.
An example research score might combine:
25% multi-window consistency
20% risk-adjusted return
15% trade count confidence
15% volume confidence
15% liquidity quality
10% portfolio diversification
Weights are strategy assumptions, not universal truths. Backtest them with realistic observation and execution delay.
Apply minimum evidence filters
| Filter | Why it matters |
|---|---|
| Account age | Reduces emphasis on one-off profiles |
| Number of trades | Provides a larger behavior sample |
| Total volume | Indicates the scale behind reported results |
| Multiple active days | Distinguishes repeated behavior from one event |
| Supported network | Ensures your validator understands the assets |
| Minimum liquidity | Limits impossible or high-impact replication |
Followers and verification can provide social context, but neither proves trading quality.
Measure concentration and drawdown
Use balance snapshots to calculate how much equity sits in the largest position. Use historical aggregate snapshots to estimate peak-to-trough drawdown. Two traders with the same return can have very different risk profiles.
Avoid survivorship and look-ahead bias
Save the leaderboard as it appeared at each research timestamp. Do not backtest today’s top traders as though you knew their identity before they became successful. Apply the same delayed data and selection rules that a live system would have used.
Simulate execution honestly
Include fees, price impact, slippage, failed quotes, unsupported tokens and transaction latency. Reject a candidate when liquidity or token safety checks fail, even if that makes simulated performance less attractive.
Keep the output informational
Present score components and uncertainty rather than a single authoritative badge. A useful research report explains why a trader entered the candidate set, what risks were observed and which data remains unverified.
Responsible selection starts with evidence, continues with independent market validation and ends with explicit portfolio limits. Rank alone should never reach a signer.