🐂 OX v1.0 — Risk-Parity / All-Weather Hedge Fund

Ox is the fund line's core holding — the steady thing you hold while betting
with the others. Its edge isn't a view; it's risk balancing. Each sleeve is
sized by inverse realized volatility (w_i = (1/vol_i) / Σ(1/vol_j)), so a
low-vol sleeve (gold, indices) carries more notional than a high-vol one (a
crypto alt) and no single asset class dominates portfolio risk — true risk
parity. It is always invested, low leverage, low turnover. One producer script
(ox-producer.py) serves both books; the OX_LEG env var selects which.

Book	Style	Wallet env	Runtime	Scanner
core	Vol-balanced all-weather basket (LONG)	OX_CORE_WALLET	runtime-core.yaml	ox_core_signals
ballast	Defensives, scaled up on risk-off (LONG)	OX_BALLAST_WALLET	runtime-ballast.yaml	ox_ballast_signals

What makes Ox different

• vs. every other fund: all of them use a flat margin_pct per position. Ox
  sizes each sleeve by inverse volatility — that is the product.
• vs. Wolf (regime rotation): Wolf rotates fully and can go to cash. Ox is
  always invested in a balanced basket and never shorts.
• vs. Rhino (tail-risk): Rhino is asymmetric (bleeds in calm, pays in shocks).
  Ox is the low-drawdown core — it doesn't try to time anything.
• vs. Elephant (macro): Elephant trades each macro asset on its own trend.
  Ox holds a diversified, risk-balanced basket and rebalances, no trend call.

The all-weather sleeve basket (core)

Spans risk sleeves and genuinely-diversifying defensive sleeves, intersected
each tick with the live instrument board:

• Crypto: BTC, ETH, SOL
• Equity indices (XYZ): SP500, XYZ100
• Metals (XYZ): GOLD, COPPER
• Energy (XYZ): BRENTOIL
• FX / defensive (XYZ): DXY (dollar), JPY (yen)

The defensive sleeves (gold, dollar, yen) are what make the core all-weather —
they tend to rise when risk assets fall, so the vol-balanced basket has a built-in
cushion before the BALLAST book even engages.

CORE book — risk-parity sizing (LONG only)

Each tick:

1. Estimate realized vol (stdev of 1h returns over volBars 30) for the full basket.
2. Compute inverse-vol weights over the whole basket (so a re-entering sleeve gets its correct fractional weight, never the full budget).
3. Size each un-held sleeve: marginUsd = portfolioBudgetPct (0.60) × equity × weight, capped at maxWeightPct (22%) of equity.
4. Knife guard: decline to add a sleeve in a hard 4h downtrend (≥0.8 strength); existing sleeves are held through normal drawdowns by the DSL.

Holds up to 10 slots (the whole basket), 3x leverage, 600s tick (slow rebalance).

BALLAST book — defensive overlay (LONG only)

Always-on. Sizes the defensive basket (gold/silver/dollar/yen) inverse-vol to a
small base budget (portfolioBudgetPct 0.18), and scales it up ×riskOffMultiplier
(2.0) when a light risk-off lean confirms — equities 4h bearish + gold/dollar
4h bullish, riskOffThreshold 1 vote. Never shorts, never fully to cash; it's
the active downside cushion on top of the core's static balance.

Execution & exit

• both books: LOW leverage clamp (3x), then each asset's HL venue max; FEE_OPTIMIZED_LIMIT entries + exits
• CORE slots 10, portfolioBudgetPct 60% — DSL wide let-it-hold (a core is rebalanced, not stopped on noise): phase1 max_loss 16%, all time-cuts OFF, hard_timeout 14d; phase2 12%→0 / 25%→45 / 50%→65 / 90%→80
• BALLAST slots 4, portfolioBudgetPct 18% (×2 on risk-off) — DSL wide: phase1 max_loss 14%, time-cuts OFF, hard_timeout 10d; phase2 10%→0 / 22%→45 / 45%→65 / 80%→80

Sizing dependency: Ox emits a different marginUsd per sleeve (the
inverse-vol weight). Its risk parity depends on the runtime honoring
per-signal signal.data.marginUsd rather than collapsing every position to a
flat strategy.margin_pct. The YAML margin_pct (6 core / 5 ballast) is only a
fallback cap.

XYZ handling

Most sleeves are XYZ — candle fetches route with dex="xyz" (the producer keys
off the xyz: prefix). XYZ trades 24/7 (no market-hours gating). The
main/xyz clearinghouse sections are two VIEWS of ONE cross-margined wallet,
so get_positions() takes accountValue ONCE via max() — never sums.

Risk gates (risk.guard_rails)

Gate	core	ballast
daily_loss_limit_pct	8	8
max_entries_per_day	12	6
max_consecutive_losses	5	4
cooldown_minutes	60	90
drawdown_halt_pct	18	16
per_asset_cooldown_minutes	360	360
data_retention_hours	168	168
drawdown_reset_on_day_rollover	true	true

Files

File	Purpose
runtime-core.yaml	Core-book runtime spec (wide let-it-hold DSL)
runtime-ballast.yaml	Ballast-book runtime spec (wide, defensive)
scripts/ox-producer.py	Book-aware producer daemon (realized-vol + inverse-vol weighting + both books)
scripts/ox_config.py	Leg resolution + SenpiClient wrapper + helpers
config/ox-core-config.json	Core tunables (sleeves, budget, vol window)
config/ox-ballast-config.json	Ballast tunables (defensives, risk-off scaler)

Operator install

See README.md — the two books are two daemons (OX_LEG=core and
OX_LEG=ballast) on two wallets (default 70/30 core/ballast funding), each with
its own runtime YAML.

Hard rule for user-conversation Claude sessions

User-conversation Claude sessions MUST NOT call any of:
create_position, close_position, edit_position,
ratchet_stop_add, ratchet_stop_edit, ratchet_stop_delete,
cancel_order, strategy_close, strategy_close_positions.

These tools are reserved for the producer daemon (entry path) and the
DSL ratchet engine (exit path). User-conversation sessions are
read-only. Each producer daemon handles real signals on its next tick.

License

Apache-2.0 — Copyright 2026 Senpi (https://senpi.ai)