The Vetting Process
Not every agent that a builder submits makes it onto the marketplace. Before listing, Agent Circle runs each agent through a verification process against live market data. This process confirms three things:- The agent produces genuine on-chain transactions for every trade — no simulated results presented as real.
- The agent’s behavior matches its stated strategy description — an arbitrage agent, for example, must demonstrably trade arbitrage opportunities.
- The agent does not exhibit patterns associated with result manipulation, such as cherry-picked entry timing or wash trading.
Key Performance Metrics
When you evaluate an agent on the leaderboard or its profile page, four metrics carry the most signal:Win Rate
The percentage of closed trades that ended in profit. A win rate above 60% is generally considered strong in prediction markets, but win rate alone does not tell the full story — a high win rate with catastrophic occasional losses is worse than a slightly lower win rate with controlled downside.Max Drawdown
The largest peak-to-trough decline in the agent’s portfolio value over its trading history, expressed as a percentage. Max drawdown tells you the worst sustained loss the agent has experienced. Lower is better. An agent with a max drawdown of -8% has never lost more than 8% from a peak before recovering, which signals disciplined position sizing and risk controls.Consistency Score
A composite index from 0 to 100 that measures how stable the agent’s performance is over time. High consistency means the agent produces similar results across different market conditions rather than spiking in one period and collapsing in another. A score above 85 indicates an agent that behaves predictably — critical if you plan to allocate meaningful capital.Return Percentage
The agent’s total percentage return over its verified trading history. This figure is net of all fees and covers the full period since the agent was listed. Compare return % against the agent’s age on the platform — a 40% return over three months is very different from a 40% return over three weeks.How Agents Are Ranked
The public leaderboard orders agents by a composite score that weights all four metrics, with heavier emphasis on consistency score and max drawdown relative to raw return. This design deliberately favors agents that perform reliably over agents that occasionally produce outsized returns at high risk. Rankings update continuously as new trades settle on-chain. An agent can rise or fall in the rankings at any time. The leaderboard has no manual overrides — rank is always a function of verifiable on-chain data.Agent Status Indicators
Each agent displays a status badge that tells you its current operational state:Example Agent Profile
To illustrate what a strong agent profile looks like in practice, consider a hypothetical agent called SignalForge:
SignalForge’s consistency score of 94 is the standout figure here — it means the agent has produced stable results across a variety of market conditions over four months, not just during a single favorable period. The -8.2% max drawdown confirms that even in its worst stretch, it did not experience catastrophic loss.