🐉 HYDRA v2.0 — Single-Coin Conviction Fund (cross-asset hedge)

Hydra is a long-term conviction book on one coin, cushioned by a diversified
short of other assets — so the hedge protects against the market risk that
would also hurt your coin, without ever betting against your own thesis. One
producer; HYDRA_COIN sets the thesis asset, HYDRA_LEG sets the head.

The two v2.0 ideas:

1. The thesis rides. The core is a multi-month conviction position on a loose,
   catastrophic-only DSL — you're not shaken out by normal volatility. The hedge,
   not a tight stop, handles the interim drawdown.
2. The hedge is cross-asset. Shorting your own thesis coin fights your conviction
   and whipsaws. Instead the hedge shorts a diversified blend of other assets
   (BTC, other majors, indices, high-beta names) — only the ones actually in a
   confirmed downtrend, vol-parity sized, and sized up when the thesis coin itself
   is breaking down. It auto-scales: ~nothing to short in a calm uptrend (light),
   many names breaking in a broad selloff (heavy). Long the conviction, short the
   carnage elsewhere.

Net: a long-term conviction long that you hold through volatility, cushioned
by a diversified short that grows as the market breaks — without ever shorting the
coin you believe in.

v1.0 → v2.0: v1.0's hedge was a same-asset stress short on the thesis coin
(zero basis risk, but it lagged, whipsawed on V-recoveries, and fought the thesis),
and the core ran a tighter ratcheting DSL. v2.0 makes the core a long-term hold and
the hedge cross-asset.

Shipped as named variants over ONE shared engine

Hydra is published as three named variants — Hydra-ETH, Hydra-SOL,
Hydra-HYPE — each a self-contained, deployable fund the picker recommends
directly (no "which coin?" prompt needed). They share ONE producer + runtimes;
only the per-coin config sets differ. HYDRA_COIN selects the variant,
HYDRA_LEG the head — three wallets per variant:

Each variant loads config/hydra-<coin>-<leg>-config.json. In v2.0 the variants
differ only by the thesis coin (HYDRA_COIN) — the hedge params are uniform,
because the hedge shorts a blend of other assets, not the thesis coin (the
producer auto-excludes the thesis coin from hedgeUniverse at runtime). Run all
three = nine wallets, one codebase. Extensible to any major.

How each head scores (producer-side)

• core — single-coin. 4h trend is the hard gate (sit out a neutral tape); LONG
  if bullish / SHORT if bearish; 1h confirmation, RSI blow-off (long) / capitulation
  (short) guard, funding tailwind; conviction → leverage tier (stdLeverage →
  maxLeverage at apexScore). Ridden on a catastrophic-only DSL (long-term hold).
• dip — single-coin. Requires a confirmed 4h uptrend AND a pullback (1h not
  bullish, or 1h RSI ≤ dipRsiMax); LONG only; tactical add inside the uptrend.
• hedge (v2.0 — cross-asset) — scans hedgeUniverse (a blend of OTHER assets,
  thesis coin excluded) and shorts only the names in a confirmed 4h downtrend
  (per-asset drawdown over stressLookback ≥ stressDropPct, or a 1h breakdown),
  capitulation-guarded. Vol-parity sized (margin inversely ∝ each asset's ATR%,
  clamped) and multiplied UP by a thesis-stress factor (thesisStressFullPct,
  stressMultMax) when the thesis coin itself is breaking down. A basket (up to
  maxSlots), total margin capped at hedgeMaxTotalPct. Auto-scales: ~nothing to
  short in a calm uptrend, heavy in a broad selloff. Never shorts the thesis coin.

Sizing — the funding split is YOUR dial

The split across a coin's three head wallets is the operator's funding decision.
Default ~50 core / 25 dip / 25 hedge of the coin's combined budget — net-long
the thesis with a real cushion. Fund only core+dip to drop the hedge; raise hedge
for more protection.

Fleet-standard rules (enforced)

• Max leverage 5x core / 4x dip / 3x hedge (strict clamp then venue max; hedge
  lowest — short squeezes are violent). core/dip = one position on the thesis coin;
  hedge = a basket (up to maxSlots), total margin capped at hedgeMaxTotalPct.
• Sizing — core/dip marginPct of equity; hedge vol-parity (no hardcoded $).
• Drawdown halt 22/20/20% from rolling peak.
• Mandatory DSL; entries + exits FEE_OPTIMIZED_LIMIT with taker fallback.
• DSL profiles — core = catastrophic-only (wide disaster stop, weak_peak OFF,
  late Phase 2; ride the long-term thesis); dip = moderate; hedge = tight-ish (bank
  the cushion, let it run in a sustained selloff). Time-cuts OFF on all.
• Budget-scaling notional floor max(account_value × minNotionalPctOfEquity, venueMinNotionalUsd); sizes off max(main, xyz) account value.
• Signature-adaptive daemon launch (passes wallet=/scanner= only if the
  installed helpers signature accepts them).

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 Hydra's wallets. Entries are emitted only by the
producer daemon; exits are owned only by the runtime DSL.

Versions

• v2.1 (current) — two fixes from the first live run (HYPE chop):
  (1) 1d-alignment chop filter on core + dip — they only enter when the daily
  trend confirms the 4h, so they stop buying the top of a range (in chop the 1d is
  NEUTRAL → stand down); dip also requires a deeper RSI pullback (dipRsiMax 42).
  (2) Opt-in hybrid hedge (hedgeIncludesThesis) — for an idiosyncratic coin
  (HYPE), the hedge will ALSO short the thesis coin when IT is the one breaking down,
  cushioning a coin-specific move the cross-asset blend misses. ETH/SOL stay pure
  cross-asset (false). NOTE: the loose core DSL is a runtime file — recreate
  the core runtime on upgrade or it keeps the tighter prior DSL (the v2.0→host gap).
• v2.0 — single-coin conviction fund: core/dip = long-term hold on a
  catastrophic-only DSL; the hedge became a cross-asset blend short (vol-parity,
  thesis-stress-scaled, never the thesis coin).
• v1.0 — single-coin portfolio fund: three regime-gated heads with a same-asset
  stress short hedge and a tighter ratcheting core DSL.

Planned v2.2: Smart-Money confirmation; the "follow them out" exit isn't applicable
here (it's a single-coin thesis, not a copy book).