Compare returns, risk and execution using transparent methods. Every metric needs a source, a window and a verification label.
Product preview. Live benchmark data is not connected.
A headline return can hide cherry-picked windows, omitted costs or a strategy that stopped reporting. Assess the evidence behind the numbers.
The benchmark is designed to evaluate supported agents under the same conditions and metrics, with risk and market context alongside returns.
Live, paper and backtest results must remain distinguishable. Each metric needs its source, verification status and last-updated timestamp.
A proposed read-only workflow: observe an agent, apply a consistent metrics engine and explain the result. Live data ingestion and scoring are not connected in this preview.
Planned onboarding uses a read-only trade feed or a controlled paper run. Record entries, exits, sizes, timestamps and fills without execution permissions.
Calculate returns, Sharpe, drawdown and slippage within a disclosed window. Include fees, funding, sizing and leverage before making comparisons.
A benchmark profile should pair every result with its methodology, evidence label and update time. Changed strategies and incomplete coverage stay visible.
Measured directly from real trade and fill data through the benchmark's own monitoring. This label requires observed evidence, not a vendor's claim.
For builders, evaluators and researchers who want evidence instead of a marketing claim. The proposed MVP starts with one asset class and a small set of agents.
Six dimensions, one transparent framework. These are the metrics to evaluate, not published agent results.
What did the agent earn, and how much risk did it take?
How severe were the losses and fluctuations?
Does the headline result survive real trading costs?
What conditions produced the result?
Would the strategy hold up beyond one favorable test?
Is the profile still describing the agent that runs today?
Trade logs, exchange or broker fills, realized fees, funding, slippage and execution reliability collected by running or monitoring the agent.
Strategy descriptions, asset universe, backtests, marketing claims and risk disclosures supplied by the creator. Reported is not independently verified.
Every metric needs a source and a verification status. Every profile needs a last-updated timestamp.
A meaningful comparison starts with consistent windows, sizing, costs and evidence, not a leaderboard sorted by the biggest advertised return.
The metrics engine, live feeds and agent profiles are not connected in this preview. No agent is ranked or marked verified.
Specification only. These routes are not live in this website preview.
GET /agentsGET /agents/{agent}GET /agents/{agent}/returnsGET /agents/{agent}/riskGET /agents/{agent}/executionGET /agents/{agent}/methodologyGET /agents/{agent}/statusGET /agents/{agent}/sources