SuperAPI AI Directory

eel-strategy

General & other

View repository

Updated 10-16 08:46

About this skill

⚡ EEL v1.0 — Electrons vs Hydrocarbons (the AI-power pair)

Eel bets that AI datacenters are a structural new source of electricity demand — so the power complex (uranium, gas, grid, fuel cells) outperforms crude oil, the legacy transport fuel barely levered to AI. It expresses that as an energy-sector-neutral long/short:

  • LONG the power complex — URNM (uranium), NATGAS (gas-fired power), COPPER (grid/electrification), BE (Bloom Energy fuel cells), USAR (rare earth).
  • SHORT crude oil — BRENTOIL + CL (WTI).

Why it's hedged: both legs are energy, so a broad energy crash hurts both and cancels — the P&L is the electrons-vs-barrels spread, not energy direction. One producer (eel-producer.py) serves both books; EEL_LEG selects.

Book Holds Wallet env Runtime Scanner
long AI-power complex (LONG) EEL_LONG_WALLET runtime-long.yaml eel_long_signals
short Brent + WTI crude (SHORT) EEL_SHORT_WALLET runtime-short.yaml eel_short_signals

The thesis, asset by asset

Sleeve Direction Names Rationale
Nuclear power LONG URNM (1.2×) uranium — the purest AI-power play
Grid / electrification LONG COPPER (1.1×) the metal of the electrification buildout
Gas-fired power LONG NATGAS the bridge fuel for datacenter load
Distributed power LONG USAR (0.7×), BE (0.6×) rare-earth magnets, on-site fuel cells (smaller/speculative)
Crude oil SHORT BRENTOIL, CL legacy transport fuel — the hedge leg

How it picks (producer-side)

  1. Curated universe — config.universe, intersected with the live board + a relative-to-market liquidity gate (24h vol ≥ volFloorPctOfMedian × the universe median — no hardcoded $ floor).
  2. ABSOLUTE trend is the gate — long power only while its 4h is not bearish; short oil only while its 4h is not bullish (+ capitulation guard so it never shorts an exhausted oil bottom).
  3. Relative strength is a tiebreaker — tilts ranking/size, never benches a genuinely-trending name.
  4. Conviction sizing — margin = account_value × marginPct × sizingWeights[name]. Uranium/copper big, fuel-cells small. No hardcoded $.

The long/short balance is YOUR dial

The split is the operator's funding across the two wallets (see config _hedge_note) — not hardcoded. Default a slight long-power tilt (~55/45) since the thesis is directionally pro-power (long book: 4 slots / 18% / 5x; oil book: 2 slots / 15% / 4x). Fund 50/50 for a cleaner market-neutral spread, or tilt to the oil book if you expect a crude shock.

Fleet-standard rules (enforced)

  • Max leverage 5x power / 4x oil (strict clamp + runtime gate; venue caps below).
  • Per-position margin ≤ 18% / 15% (× conviction weight, ≤ 25% cap).
  • Drawdown halt 20% / 18%, reset_on_day_rollover.
  • Mandatory DSL; entries + exits use FEE_OPTIMIZED_LIMIT with taker fallback.
  • Verbose per-tick JSON; sizes off max(main, xyz) account value.
  • Oil book runs tighter (lower leverage, tighter max-loss, faster stall-cuts) — a crude supply-shock spike is violent.

Caveat — say it honestly

The pair can invert short-term on an oil-supply shock (Mideast escalation, an OPEC cut) or a cold-winter natgas spike — events that move hydrocarbons independent of the electrification thesis. The DSL owns those drawdowns; the edge is the structural multi-month electrons-beat-barrels drift, not any single week.

Hard rule — user-conversation sessions are READ-ONLY

A Claude session conversing with a user MUST NOT call create_position, close_position, edit_position, ratchet_stop_*, cancel_order, or any strategy_close* tool against Eel's wallets. Entries are emitted only by the producer daemon; exits are owned only by the runtime DSL.

v1.0 — initial build

A Lion-family two-book long/short (the proven helpers-native pattern: one leg-parameterized producer, two wallets, absolute-trend gate, RS tiebreaker, per-group conviction sizing, relative-to-market liquidity gate) pointed at an energy dispersion — the cleanest hedge of the thesis-fund family, since both legs sit in one sector.