Ranger OSO Vault

§01Thesis

A yield vault with a small, disciplined directional bet.

Ninety percent of capital earns USDC lending yield on Kamino. The remaining ten percent runs a signal-driven BTC perp strategy derived from Deribit options positioning data. When trades profit, gains sweep back to lending; the bucket resets to 10% at each +20% realized gain. We don't predict price; we detect imbalance in options positioning and fade it.

Backtest APY: ~21%
Stress APY: ~12%
Worst vault DD: -0.25%
Alpha vs BTC: +40.8%

Backtest window Apr 2025 — Mar 2026 · 11 months · 20,000 Monte-Carlo paths. Strategy version v2 · April 2026.

§02How it works

Seven deterministic steps.

01Deribit BTC options data (hourly)
02Feature engineering (positioning metrics)
033 signal strategies (A / B / C)
04Regime filter — trade only when conditions are favorable
05Execute on BTC perp (1× leverage)
06Rebalance to 90/10 at +20% realized gain
07Trading bucket capped at 10% of NAV

Parameters

Base asset: USDC
Lending: 90% — overcollateralized lending on Solana
Trading: 10% — BTC perp via Jupiter Perps or Flash, 1× leverage
Sweep: Rebalance to 90/10 at +20% realized trading gain
Cost: ~7.5 bps per round-trip trade

§03The edge

Four positioning metrics, threshold-crossing rules.

When options traders hedge aggressively with puts, it signals fear; BTC historically bounces. When call speculation dominates, it signals greed; BTC pulls back. Four metrics engineered from hourly Deribit data capture this imbalance.

Signal metrics

Metric	Formula	Captures
`tail_skew`	Deep OTM put vol vs deep OTM call vol (smoothed)	Hedging demand
`iv_skew`	IV of puts vs calls across strike surface (smoothed)	Fear premium
`skew_composite`	−tail_skew + iv_skew	Unified skew — combines both
`net_dw`	Delta-weighted net options flow (calls − puts)	Smart-money direction

Strategy logic

§A · Long · dual confirmation

ENTER LONG when tail_skew ≤ T1 AND iv_skew ≥ T2
EXIT when tail_skew ≥ T3 OR iv_skew ≤ T4

Both metrics confirm bullish positioning to enter. Either reverting triggers exit.

§B · Short · contrarian

ENTER SHORT when tail_skew ≥ T5
EXIT when tail_skew ≤ T6

Single-metric rule, triggered when upside speculation is extreme.

§C · Flip · state machine

ENTER LONG when net_dw ≤ T7 AND skew_composite ≥ T8
EXIT LONG when net_dw ≥ T9 OR skew_composite ≤ T10
ENTER SHORT when net_dw ≥ T11 AND skew_composite ≤ T12
EXIT SHORT when net_dw ≤ T13 OR skew_composite ≥ T14

FLAT → LONG → FLAT → SHORT. AND to enter, OR to exit.

Thresholds T1–T14 are determined by robust plateau grid search and frozen in production. Values withheld.

§04Quantified edge

Winning more often, winning bigger.

Portfolio trades: 217
Win rate: 78%
Avg W/L ratio: 2.2×
Cost / trade: 7.5 bps

Average win is 2.2× the average loss. The edge isn't in winning occasionally — it's in winning 78% of the time AND winning bigger than we lose.

BTC buy & hold returned -21.4% in the same window; the vault returned +19.4%. That's +40.8% alpha over passive holding.

§05Three strategies

One per category — best after stress testing.

§A · long30% of 10%

Long · dual confirmation

Enters long when two independent positioning metrics both confirm bullish imbalance. Fewer but higher-quality entries.

Trades: 94
Win rate: 79%
Return: +342%
W/L: 1.8×
Avg hold: 1.7d
Per month: 9
Best trade: +13.3%
Worst trade: -8.4%
Strat max DD: -15.9%
Vault impact: -0.25%

§B · short30% of 10%

Short · contrarian

Enters short when positioning shows extreme upside speculation. Simplest rule, best win/loss ratio — fades the crowd.

Trades: 83
Win rate: 76%
Return: +320%
W/L: 2.8×
Avg hold: 1d
Per month: 8
Best trade: +11.5%
Worst trade: -7.3%
Strat max DD: -10.4%
Vault impact: -0.22%

§C · flip40% of 10%

Flip · state machine

Bidirectional FLAT → LONG → FLAT → SHORT. Both signals must agree to enter (AND); either can trigger exit (OR). Most selective, highest win rate.

Trades: 40
Win rate: 82%
Return: +81%
W/L: 2.5×
Avg hold: 0.8d
Per month: 4
Best trade: +4.0%
Worst trade: -1.2%
Strat max DD: -1.6%
Vault impact: -0.05%

No single strategy controls more than 4% of total vault NAV (10% trading bucket × 40% max sub-allocation).

§06Risk management

Three rules + a regime filter.

Rule 101

Capped exposure

Only 10% of vault capital is ever at risk. A total wipeout of the trading bucket costs 10% of NAV — nothing more.

Rule 202

Profit locking

When the trading bucket realizes +20% gain, rebalance to 90/10. Realized profits move to lending, safe from future trades. Trading bucket resets.

Rule 303

Rebalance cap

Trading bucket cannot silently grow beyond 10% of NAV. Excess is swept to lending automatically.

Regime filter

Trend filter (moving average) + volatility filter (rolling realized vol) gate entries, blocking trades during extreme volatility or choppy markets. Tested across 5 gate combinations and 12 strategies to confirm the detection is stable, not fragile.

§07Performance

Baseline vs stress, and why 90/10.

Baseline vs stress · 11-mo vault

Metric	Baseline	Stress (alpha halved + 5% lending)
Vault return	+19.4%	+9.6%
APY	~21%	~11%
Total trades	217	—
Portfolio win rate	78%	—
Worst trade vault impact	-0.25%	—

Per-strategy breakdown

Strategy	Trades	Win rate	Return	Max DD	Vault impact
§A	94	79%	+342%	-15.9%	-0.25%
§B	83	76%	+320%	-10.4%	-0.22%
§C	40	82%	+81%	-1.6%	-0.05%

Why 90/10? — allocation sweep

We tested 70/30 through 95/5 on the same 3-strategy portfolio. 90/10 clears the 10% stress-APY floor while capping drawdown under −0.4%. Going wider doubles drawdown; going narrower kills stress yield.

70/30

APY: ~83%
Stress APY: ~25%
Max DD (p5): -1.19%

80/20

APY: ~58%
Stress APY: ~18%
Max DD (p5): -0.80%

90/10chosen

APY: ~21%
Stress APY: ~11%
Max DD (p5): -0.39%

95/5

APY: ~20%
Stress APY: ~8%
Max DD (p5): -0.18%

§08Stress

Five scenarios, one projected-drawdown matrix.

Stress tests applied

Stress test	What it simulates	Still profitable?
Higher costs (15 bps)	Slippage during volatile markets	Yes
Execution delay (1-bar)	Can't always execute instantly	Yes
Signal edge weakens 50%	Markets adapt over time	Yes (~11% APY)
Split-sample (half vs half)	Is the edge stable or a fluke?	Yes (nearly identical)
Combined (all above)	Everything degrades at once	Yes (~11% APY)

Projected max drawdown · Monte Carlo · 20,000 paths

Scenario	Median	p5 (worst 5%)	p1 (worst 1%)
Baseline (8% lending, full signal)	-0.17%	-0.31%	-0.41%
Stress (5% lending, signal halved)	-0.23%	-0.39%	-0.49%

§09Breaking point

We tested each dimension until the vault broke.

Headroom per dimension

Dimension	We test at	Breaks at	Headroom
Signal decay	50%	Never (still +0.4% APY at 95%)	45%+
Trading cost	15 bps	Never (still +15.8% APY at 50 bps)	35+ bps
Lending APY	5%	Never (still +28.2% APY at 2%)	3%+

Combined stress — when it actually breaks

Signal decay	Trading cost	Result
50%	50 bps (6.7×)	+15.8% — still fine
70%	30 bps (4×)	+9.0% — still fine
75%	50 bps (6.7×)	−1.3% — first break

No single stress dimension breaks the vault. It only breaks when extreme scenarios combine simultaneously — at roughly 75% signal decay combined with 50 bps cost (6.7× baseline friction).