Ride the whales you choose. A remora fish attaches to a shark and goes where it goes. Remora attaches to a small, hand-picked set of whale traders — it finds each whale's biggest-conviction bet and mirrors the strongest one, with a consensus boost when several whales pile into the same trade.
Why this strategy exists
The broad trader-followers (Raptor, Jackal, Spider) scan a leaderboard universe and synthesize a pick. That's powerful but opaque — you don't control whom you're following. Remora is the opposite: you name the whales. It mirrors their single highest-conviction position, so your exposure tracks specific traders you trust, and it leans hardest when those whales agree.
CRITICAL RULES
RULE 1: You pick the whales
Remora trades nothing until config.whales is set to a list of trader_ids / wallets. There is no default whale set — these are your choices.
RULE 2: Highest conviction = largest notional
For each whale, Remora takes their single largest-notional open position (size × entry, falling back to margin), above minNotionalUsd (default $5,000, to skip dust). That's the whale's strongest live bet.
RULE 3: Consensus is the edge multiplier
Top positions are aggregated into (asset, direction) candidates across whales. The more whales independently hold the same asset+direction, the higher the score (2 whales → +2, 3+ → +3). One whale can be wrong; three agreeing is a much stronger signal.
RULE 4: Producer enters. DSL exits.
No close_position call site. Whales hold winners, so the DSL is the let-winners-run preset — wide ladder (T0 +10% / lock 0), max_loss 18%, time-cuts OFF except a 120h hard_timeout staleness cap. ⚠️ Remora does not yet mirror the whale's EXIT — the hard_timeout prevents holding a stale mirror indefinitely. A whale-exit mirror is a planned v1.1 enhancement.
How Remora scores a trade
Gate: at least one configured whale holds a qualifying top position (above minNotionalUsd).
Score components (max ~7):
| Signal | Points |
|---|---|
| A tracked whale's top conviction position | +3 |
| Consensus: 2 whales agree | +2 |
| Consensus: 3+ whales agree | +3 |
| Any agreeing whale is ELITE / RELIABLE tier | +1 |
Floor: minScore: 4 (so a single whale's top position clears the floor only if it is ELITE-tier; otherwise consensus is required).
DSL preset (let-winners-run — with a staleness cap)
| Phase | Component | Setting |
|---|---|---|
| Phase 1 | max_loss_pct | 18% |
| Phase 1 | retrace_threshold | 10 |
| Time cuts | hard_timeout | 120h (staleness cap) |
| Time cuts | weak_peak_cut | DISABLED |
| Time cuts | dead_weight_cut | DISABLED |
| Phase 2 | T0 → T4 | +10/0 · +20/45 · +35/65 · +55/78 · +90/88 |
Scanner pattern
A focused variant of the Trader-follower / hot-streak archetype (#5) — see senpi-trading-runtime/references/producer-patterns.md. Primary MCP calls: leaderboard_get_trader_positions (per whale — the producer signature; unwraps the nested data.positions.positions shape), discovery_get_trader_state (whale quality tier, optional). The pure functions (position_notional, mirror_direction, top_position, consensus_bonus) are unit-tested in tests/test_signal.py.
Operator install
See README.md.
Changelog
v1.0.0 (2026-05-26) — initial release
A focused single-position mirror of a hand-picked whale set, with a consensus multiplier — a deliberate contrast to the universe-scanning trader-followers. Let-winners-run DSL class (wide ladder, time-cuts off except a 120h staleness cap), taker-true entry, no null numeric signal fields, defensive nested-shape unwrapping on leaderboard_get_trader_positions, and unit-tested pure functions. Whale-exit mirroring is a planned v1.1 enhancement.
License
MIT — Copyright 2026 Senpi (https://senpi.ai).
Skill Attribution
When creating a strategy, include skill_name and skill_version. See references/skill-attribution.md.