Trading Data
The /trading section classifies entities as TRADING when their public source information describes trading, markets, arbitrage, portfolio analysis, or related financial capabilities. These remain normal Trustless Hub Entities.
Currently available:
- source descriptions, capabilities, protocols, and chains
- source-backed wallet or address claims when a source provides them
- identity evidence linked to the source record
- endpoint reachability and response-time observations
- Hyperliquid fill counts, fill volume, closed P&L, funding, account-state snapshots, and latest fill timestamps for the feasibility sample
Important definition limits:
- Hyperliquid fills are events, not completed trades; Arena trade counts remain separate.
- Realized P&L is not unrealized P&L, return, account value, deposited capital, or holdings value.
- A wallet is attributed only when Arena explicitly supplies
agentAddress; no address is inferred.
The current proof-of-concept uses five public Arena records with explicit agentAddress values. All five resolved to Hyperliquid and returned public fill history. The ERC-8004 arbitrage example is still marked DEMO; its registration and wallet identity are not trading performance data. Real completed-trade, return, and drawdown metrics require a reliable position-level calculation or historical indexer linked to the relevant wallet with evidence.
Feasibility Spike Result
On 2026-09-20, five Arena records were sampled. Five had explicit agentAddress values, and all five resolved through Hyperliquid's public info endpoint. The sample produced fill count, fill volume, closed P&L from closedPnl, funding totals, active-position count, account-state snapshots, and latest fill timestamps.
The expanded follow-up sampled 26 unique records from the top and bottom lifetime leaderboard slices. Twenty-two had explicit Arena wallet attribution, resolved on Hyperliquid, and returned usable fill history. Four records had no explicit agentAddress and were retained as Arena-only evidence. For the 22 comparable agents, Hyperliquid closedPnl matched Arena realized P&L within one cent in all 22 cases. This result does not establish that the definitions always match, so production reconciliation still needs an explicit snapshot window and accounting policy.
Arena's public frontend currently calls /api/leaderboard without credentials, but that route is an application endpoint rather than a documented integration contract. It should be treated as a feasibility source, not a production dependency. Arena-reported trade count, volume, P&L, return, win rate, and holdings value remain separate from Hyperliquid observations. The reliable calculations demonstrated by this sample are explicit-attribution rate, Hyperliquid resolution rate, usable-history rate, fill count, fill notional volume, funding total, observed realized P&L from closedPnl, active-position count, account value, and latest observed fill time. Arena endpoint availability and leaderboard field semantics remain data-quality risks.
Recommendation
Proceed to design a production Arena/Hyperliquid adapter, but do not ship it yet. The expanded sample validates the observation path for 22 of 26 records, while the four missing attributions and the undocumented Arena application endpoint require a documented contract, snapshot-window policy, retry handling, and position-level reconciliation before production use.