1. Public data collection
The server reads the public Hyperliquid leaderboard, info API, public fills, portfolio and position data, and public WebSocket trade streams. We do not request exchange login credentials, trading API keys, or private keys.
2. Normalization and snapshots
Addresses and market identifiers are normalized and stored with observation times. State differences create observed new-position, increase, decrease, and close events, while deduplication keys suppress repeated events.
3. Specialist selection
The default view uses 30 days and the system also calculates 1, 3, 7, 30, and 90-day windows. Closing fills are grouped by order to calculate wins, losses, flats, and realized PNL.
- Current minimums: 5 closed orders, 4 decisive orders, 50% win rate, and USD 250 realized PNL
- Score: 85% Wilson win-rate lower bound plus 15% logarithmic sample strength
- Up to 10 profiles per market, ordered by score, sample, and recency
4. PNL Cohort construction
Public leaderboard wallets need account value above USD 1 million and non-zero weekly and monthly PNL. They are then grouped by all-time PNL, with up to 50 wallets retained in each of six bands.
5. Wallet Deck and alerts
A signed-in member can save up to six wallet cards with their source and card snapshot. Newly observed position events for active selections create member-specific in-app alerts. Telegram delivery runs only when separate access and a user-owned verified bot configuration are both present.
6. Updates and failures
Collection, selection, and real-time observation run on different schedules. External rate limits, outages, delayed jobs, or target changes can leave a view empty or stale. Always check calculation and observation times.
7. The user's decision
GODAERI restructures observable data; it does not make a trading decision. Users should independently check the source, timing, sample, and their own risk capacity.