# Risk sheet — bands, margin, decay and the leverage choice

Generated 2026-09-11 by `pnpm --filter @lev/sim report` (`packages/sim/src/report.ts`). Re-run after any parameter change; do not hand-edit. Work item `docs/WEEK_PLAN.md` #18; the source of `docs/DECISIONS.md` D21, and the input to the leverage choice D22 defers to vault creation (#13).

Every simulated number comes from raw-integer arithmetic (AGENTS.md rule 1) over the committed daily OHLC in `packages/sim/data`, which is **planning data only** — see `packages/sim/data/SOURCES.md` for sources, the Stooq fallback and the commodity ETF proxies.

**The simulator runs the registry's policy as configured.** Margin bands are shares of notional (D21); `packages/registry` exports `MARGIN_OF_NOTIONAL` and every policy uses it. D7's old NAV-basis numbers are still selectable (`SimOptions.marginBasis = 'nav'` with `d7NavMargin()`) and appear only in §2c, which measures what the change bought.

---

## 0. Headline

**1. D21 does what it was meant to do.** Stating margin as a share of notional makes every margin and liquidation distance independent of leverage: 9.3% from the floor at 2x and 9.3% at 3x — the same number, because L cancels out of (m₀·L − Lx)/(L(1∓x)). Measured across 22 tokens at 3x, simulated liquidations fall from **57.1/yr under D7 as written to 9.4/yr** (§2c).

**2. The remaining buffer is 9.3% long / 8.7% short from the margin floor**, at every leverage, against a registry maintenance margin of 3.5%. Single names still gap through it — MSTR fell 27.4% over the 2024-08-05 weekend and COIN 24.9% on 2022-05-11 — so liquidation is a managed event, not an eliminated one. What D21 removed is liquidation on *ordinary* 6–9% gaps.

**3. A Phoenix liquidation is bounded, and that is itself a hazard.** Because only a bounded share of NAV is ever posted, the vault's worst case on any single gap is the posted collateral. On MSTR's −27.4% weekend a liquidated 3x token keeps **63% of NAV against the 18% an un-liquidated 3x position would have held** — the margin structure *beats* the honest mark in the deep tail, because Phoenix's backstop absorbs the negative equity. That is exactly the behaviour that gets a venue to intervene, and Phoenix's auto-deleverage would also close the winning side of the paired token. Safety net, never a strategy.

**4. Decay is the product, and per-week is the honest unit (D22).** Annualised drag tracks the textbook L(L−1)σ²/2, but holders sit for days to weeks. At 3x: uUS500 costs **0.22%/week**, uGOLD 0.21%/week, uTSLA 1.85%/week and uCOIN 4.25%/week. A fortnight in uUS500 costs 0.44% — about four times D8's 10 bps round trip, and small against a fortnight's move in a 3x index product. A fortnight in uCOIN costs 8.5%. §3b is the table for the docs page.

**5. Fees are irrelevant next to drag.** Fee plus slippage is 22 bps/yr on uUS500 at 3x against roughly 1142 bps/yr of total drag. Funding at ±10%/yr is a ±30%/yr swing at 3x — the only cost input that matters, and the one with no history (§5).

**6. Tracking error:** mean is well inside PRD §6's 25 bps/day (8.5 bps on uUS500 3x, 8.8 on uGOLD 3x). The **95th percentile** crosses 25 bps as soon as leverage passes 2x. If §6's target is a typical day it is met; if it is a bound, it is met only at 2x. Worth a ruling.

**7. Leverage, when #13 comes round:** §6.

---

## 1. What was simulated, and the assumptions Jonny should check

### Venue

| Assumption | Value | Source / status |
|---|---|---|
| Taker fee | 3.5 bps on every trade, maker never used | PLAN.md §0c. D12: band rebalances are momentum-following so they cannot rest. |
| Maintenance margin | **3.5% of notional**, from the registry's `Asset.maintenanceMarginBps` | Back-solves both buffers PLAN.md §0c publishes: (0.5 − x)/(1 − x) = 0.035 gives 48.2% at 2x, and (1/3 − x)/(1 − x) = 0.035 gives 30.9% at 3x. **Still an estimate** — the keeper must read the real value off the market account (OPEN_QUESTIONS Q13). The earlier 1/(2 × maxLeverage) derivation (2.5% on a 20x market) reproduces the 2x buffer but not the 3x one, so it is not used; it is more forgiving, so every liquidation count here is the conservative one. |
| Depth within ±0.5% | SPY $600K, TSLA $348K, NVDA $420K, GOLD $590K, SILVER $378K; **$300K assumed** for QQQ, AAPL, COIN, HOOD, MSTR, WTIOIL | PLAN.md §0c measured five markets on 11 Sep 2026. The rest are guesses at the thin end. |
| Half-spread | 2.1–2.5 bps measured, 2.5–3.0 bps assumed elsewhere | PLAN.md §0c |
| Price impact | linear in slice/depth; a slice equal to the whole ±0.5% book costs 50 bps at the margin, so D12's 20%-of-depth slice averages 5 bps | **Model, not measured.** Ignores resting size beyond ±0.5% and book refill between slices, so it is conservative. |
| Funding | hourly, configurable; **default 0**, scenarios ±10%/yr in §5 | PLAN.md §0c: "mean currently negative, longs paid; volatile". No usable history. |
| Liquidation | position closed at the mark, plus 100 bps of notional penalty on top of the close's fee and slippage; residual equity returns to idle, negative equity is absorbed by the venue backstop | **Assumption.** Phoenix's penalty is not documented in PLAN.md. Headline 3 does not depend on the number. |

### Product

| Assumption | Value | Source |
|---|---|---|
| Bands | D6 / registry: 2x → 2.0 / 1.8–2.3 / 2.6; 2.5x → 2.5 / 2.25–2.75 / 3.3; 3x → 3.0 / 2.7–3.3 / 4.0 | `packages/registry` |
| Rebalance rule | to target, only on leaving the band; expansion only while the external index is live; defensive and hard delever at any hour | D6 |
| Margin bands | **17.5 / 12.5 / 22.5% of notional**, the registry's `MARGIN_OF_NOTIONAL`, at every leverage | D21 |
| Margin ratio | venue **equity** (posted collateral ± uPnL) / notional | D21 and `docs/DATA_CONTRACT.md`. Equity rather than posted collateral matters: posted collateral alone does not fall when the position loses, so a floor on it would never fire. |
| Keeper loop order | liquidation check → margin → leverage band → margin again | The second margin pass is load-bearing under D21: a rebalance changes notional, so it changes the required equity. |
| Management fee | 100 bps/yr streamed off NAV | D8 |
| Issuance / redemption fees | **not modelled** | D8 charges them to the minter; they are not a holder-path cost. |
| Initial position build | free | It is the mint's execution, which D8's issuance fee exists to cover. Charging it would show up once per simulated year as fake rebalancing cost. |
| Flows | none — one deposit, held for the segment | Mint/redeem impact and the stale-NAV arb are out of scope (§8). |

### Path model

Each trading day becomes a close→open **gap** and then a **close**, and optionally the intraday **adverse** and **favourable** extremes in between. The keeper runs its full loop at every point, standing in for a 20–30 s crank.

- A **weekday overnight gap** is subdivided into four steps at `external`: Phoenix quotes equities on an external index 24/5 (Sun 20:00 – Fri 20:00 ET), so the keeper really does track it.
- A **weekend or holiday gap** lands as **one unattended jump** labelled `impact`, where D6 allows a defensive or hard delever but not an expansion. This is the tail that matters, and it is deliberately pessimistic: Phoenix's impact pricing actually bounds the weekend move to 1/maxLeverage (4–5%) from the last external print and only snaps at reopen, so the keeper gets a bite before the full gap lands.
- **Two intraday models are reported.** The headline tables use *close-to-close plus gaps*; the last two columns of every activity table use *open / low / high / close*, visiting the adverse extreme first. The intraday model is a hard upper bound, not a forecast: a straight line from the low to the high is the maximum-whipsaw path, and it roughly doubles turnover on high-vol names. Real paths are not V-shaped; the truth is between the two columns, nearer the close-only one.
- Commodities run CME hours rather than 24/5, so their weekday overnight gaps are also partly unattended. But the GLD/SLV/USO proxies only trade US hours, which overstates their gaps by more. Net: the commodity gap numbers are conservative.

### Segmentation

Each token is simulated in independent **252-trading-day segments**, each starting from a fresh $1,000,000 vault at target leverage, and the segments are pooled. A vault with continuous mint and redeem sits near a target size rather than compounding to a millionth of a cent over twenty years, and a NAV that decays that far turns every daily return into integer-rounding noise. "Token /yr" is therefore the *mean annual* log return of a repeatedly-rebased token, "worst yr" the worst single segment's decay, "worst-yr DD" the worst drawdown inside any one segment, and "wipeout yrs" the share of segments that lost more than 95%. Window: 2006-01-03 → 2026-09-11 for the indices, 21 segments; COIN and HOOD start at their IPOs.

---

## 2. Drawdown-to-trigger

Closed-form, from target leverage, as a % adverse move in the underlying, `long / short` — a short's notional *grows* as it loses, so it reaches every trigger sooner. Maintenance margin is 3.5% for every market in the registry, so one table covers all eleven assets. The last column is the one that matters: the buffer the keeper leaves the position sitting on at the margin floor.

### 2a. D21 as configured — margin as a share of notional

| Lev | Equity target | Posted, as share of NAV | → band_hi | → hard delever | → margin floor | → liquidation (from target) | → liquidation (from floor) |
|---|---|---|---|---|---|---|---|
| 2x | 17.5% of notional | 35.0% | 11.5% / 4.5% | 18.8% / 8.3% | 5.7% / 4.4% | 14.5% / 13.5% | **9.3%** / **8.7%** |
| 2.5x | 17.5% of notional | 43.8% | 5.7% / 2.7% | 13.9% / 7.4% | 5.7% / 4.4% | 14.5% / 13.5% | **9.3%** / **8.7%** |
| 3x | 17.5% of notional | 52.5% | 4.3% / 2.3% | 11.1% / 6.7% | 5.7% / 4.4% | 14.5% / 13.5% | **9.3%** / **8.7%** |

Every margin column is leverage-invariant. The cost is idle: at 3x the vault posts 52.5% of NAV, up to 67.5% at the ceiling, leaving 32.5%–47.5% idle for instant redemptions rather than the 55–65% D7 left. D21 records that trade-off. At 4x the policy would post 70% of NAV, which is where this approach stops working.

### 2b. D7 as it was written — margin as a share of NAV

| Lev | Equity target | Posted, as share of NAV | → band_hi | → hard delever | → margin floor | → liquidation (from target) | → liquidation (from floor) |
|---|---|---|---|---|---|---|---|
| 2x | 35.0% of NAV | 35.0% | 11.5% / 4.5% | 18.8% / 8.3% | 6.7% / 6.7% | 14.5% / 13.5% | **9.3%** / **8.7%** |
| 2.5x | 35.0% of NAV | 35.0% | 5.7% / 2.7% | 13.9% / 7.4% | 5.3% / 5.3% | 10.9% / 10.1% | **6.7%** / **6.3%** |
| 3x | 35.0% of NAV | 35.0% | 4.3% / 2.3% | 11.1% / 6.7% | 4.4% / 4.4% | 8.5% / 7.9% | **5.0%** / **4.7%** |

The 35%-of-NAV rule was sized for 2x, where it is identical to D21. At 3x it left **5.0% long / 4.7% short** from the margin floor. Every asset in `data/` has gapped further than that more than once.

### 2c. What the change was worth, measured

Simulated Phoenix liquidations per year at **3x**: same paths, same bands, only the margin basis differs. Not a closed form — these are counted events.

| Long token | D7 (NAV basis) | D21 (notional) | Short token | D7 (NAV basis) | D21 (notional) |
|---|---|---|---|---|---|
| uGOLD | 0.00 | 0.00 | dGOLD | 0.05 | 0.05 |
| uSILVER | 0.64 | 0.05 | dSILVER | 0.25 | 0.00 |
| uWTIOIL | 0.84 | 0.05 | dWTIOIL | 0.44 | 0.05 |
| uUS500 | 0.15 | 0.00 | dUS500 | 0.15 | 0.00 |
| uUS100 | 0.15 | 0.00 | dUS100 | 0.10 | 0.05 |
| uTSLA | 2.50 | 0.44 | dTSLA | 4.70 | 0.50 |
| uNVDA | 2.04 | 0.05 | dNVDA | 2.38 | 0.10 |
| uAAPL | 0.39 | 0.05 | dAAPL | 0.39 | 0.05 |
| uMSTR | 3.73 | 0.53 | dMSTR | 5.87 | 1.16 |
| uCOIN | 9.85 | 0.74 | dCOIN | 11.71 | 3.16 |
| uHOOD | 4.40 | 0.60 | dHOOD | 6.40 | 1.80 |

Across all 22 tokens at 3x: **57.1 liquidations/yr on the NAV basis, 9.4 on the notional basis** — a 83% reduction from one config line. The dev seed reproduces it from a separate code path: `pnpm --filter @lev/sim seed --margin nav` against `--margin notional`.

---

## 3. Decay, tracking and turnover

"Vol drag" is the structural cost of rebalancing (N × the underlying's annual log return minus a costless token). "Cost drag" is taker fee, slippage, funding and management on top. **Decay /wk** is the annual figure divided by 52 — a week of average volatility (D22); a volatile week costs several times it and a quiet week almost nothing. TE is |NAV daily return − N × underlying daily return| in bps/day on the costless run, PRD §6's "ex fees/funding" basis.

### Indices — US500, US100

**Long tokens**

| Token | Lev | From | UL vol | UL /yr | Token /yr | Vol drag | Cost drag | Decay /yr | **Decay /wk** | Worst yr | TE mean | TE p95 | Worst-yr DD | Wipeout yrs |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| uUS500 | 2x | 2006 | 19% | 8.7% | 13.2% | 3.3% | 0.95% | 4.3% | **0.08%** | 15% | 6.0 | 21.1 | 76% | 0% |
| uUS500 | 2.5x | 2006 | 19% | 8.7% | 14.3% | 6.3% | 1.12% | 7.5% | **0.14%** | 28% | 7.2 | 26.3 | 84% | 0% |
| uUS500 | 3x | 2006 | 19% | 8.7% | 14.7% | 9.8% | 1.65% | 11.4% | **0.22%** | 56% | 8.5 | 30.3 | 90% | 0% |
| uUS100 | 2x | 2006 | 22% | 13.8% | 22.3% | 4.2% | 1.05% | 5.2% | **0.10%** | 16% | 7.3 | 25.8 | 78% | 0% |
| uUS100 | 2.5x | 2006 | 22% | 13.8% | 24.9% | 8.3% | 1.30% | 9.6% | **0.18%** | 37% | 8.8 | 31.5 | 86% | 0% |
| uUS100 | 3x | 2006 | 22% | 13.8% | 26.4% | 13.0% | 1.97% | 15.0% | **0.29%** | 53% | 10.5 | 38.0 | 91% | 0% |

| Token | Lev | Turnover ×eq/yr | Rebalances/yr | Defensive | Top-ups/yr | Fee+slip bps/yr | Hard delevers/yr | Liquidations/yr | Turnover (intraday) | Hard delevers (intraday) |
|---|---|---|---|---|---|---|---|---|---|---|
| uUS500 | 2x | 0.3 | 2 | 43% | 3 | 3 | 0.00 | 0.00 | 0.4 | 0.00 |
| uUS500 | 2.5x | 1.2 | 6 | 47% | 3 | 12 | 0.00 | 0.00 | 1.5 | 0.00 |
| uUS500 | 3x | 2.3 | 9 | 48% | 2 | 22 | 0.05 | 0.00 | 3.1 | 0.10 |
| uUS100 | 2x | 0.5 | 3 | 32% | 4 | 6 | 0.00 | 0.00 | 0.8 | 0.05 |
| uUS100 | 2.5x | 1.7 | 7 | 45% | 4 | 18 | 0.00 | 0.00 | 2.4 | 0.00 |
| uUS100 | 3x | 3.4 | 13 | 46% | 3 | 37 | 0.00 | 0.00 | 4.5 | 0.10 |

**Short tokens**

| Token | Lev | From | UL vol | UL /yr | Token /yr | Vol drag | Cost drag | Decay /yr | **Decay /wk** | Worst yr | TE mean | TE p95 | Worst-yr DD | Wipeout yrs |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| dUS500 | 2x | 2006 | 19% | 8.7% | -29.0% | 10.0% | 1.60% | 11.6% | **0.22%** | 44% | 7.2 | 25.5 | 71% | 0% |
| dUS500 | 2.5x | 2006 | 19% | 8.7% | -38.6% | 14.7% | 2.08% | 16.8% | **0.32%** | 72% | 6.9 | 24.6 | 79% | 0% |
| dUS500 | 3x | 2006 | 19% | 8.7% | -48.5% | 20.1% | 2.25% | 22.3% | **0.43%** | 97% | 8.2 | 30.7 | 85% | 0% |
| dUS100 | 2x | 2006 | 22% | 13.8% | -42.5% | 13.3% | 1.61% | 14.9% | **0.29%** | 46% | 8.6 | 30.6 | 77% | 0% |
| dUS100 | 2.5x | 2006 | 22% | 13.8% | -56.1% | 19.3% | 2.35% | 21.6% | **0.42%** | 77% | 8.3 | 31.6 | 84% | 0% |
| dUS100 | 3x | 2006 | 22% | 13.8% | -70.9% | 26.5% | 3.06% | 29.6% | **0.57%** | 100% | 10.2 | 38.2 | 90% | 0% |

| Token | Lev | Turnover ×eq/yr | Rebalances/yr | Defensive | Top-ups/yr | Fee+slip bps/yr | Hard delevers/yr | Liquidations/yr | Turnover (intraday) | Hard delevers (intraday) |
|---|---|---|---|---|---|---|---|---|---|---|
| dUS500 | 2x | 5.0 | 14 | 55% | 7 | 49 | 0.10 | 0.00 | 9.2 | 0.34 |
| dUS500 | 2.5x | 9.9 | 26 | 61% | 6 | 98 | 0.19 | 0.00 | 16.4 | 0.58 |
| dUS500 | 3x | 15.2 | 31 | 61% | 5 | 153 | 0.19 | 0.00 | 25.0 | 0.82 |
| dUS100 | 2x | 6.5 | 19 | 56% | 9 | 69 | 0.05 | 0.00 | 11.6 | 0.29 |
| dUS100 | 2.5x | 12.9 | 34 | 62% | 7 | 162 | 0.10 | 0.05 | 22.6 | 0.44 |
| dUS100 | 3x | 20.0 | 41 | 61% | 7 | 247 | 0.15 | 0.05 | 34.7 | 0.92 |

### Commodities — GOLD, SILVER, WTIOIL

**Long tokens**

| Token | Lev | From | UL vol | UL /yr | Token /yr | Vol drag | Cost drag | Decay /yr | **Decay /wk** | Worst yr | TE mean | TE p95 | Worst-yr DD | Wipeout yrs |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| uGOLD | 2x | 2006 | 18% | 10.2% | 15.7% | 3.4% | 1.35% | 4.8% | **0.09%** | 14% | 6.9 | 25.3 | 56% | 0% |
| uGOLD | 2.5x | 2006 | 18% | 10.2% | 18.4% | 6.1% | 1.05% | 7.2% | **0.14%** | 20% | 7.1 | 25.8 | 62% | 0% |
| uGOLD | 3x | 2006 | 18% | 10.2% | 19.5% | 9.6% | 1.56% | 11.1% | **0.21%** | 31% | 8.8 | 30.8 | 72% | 0% |
| uSILVER | 2x | 2006 | 33% | 7.5% | 2.9% | 10.9% | 1.31% | 12.2% | **0.23%** | 41% | 11.6 | 44.1 | 82% | 0% |
| uSILVER | 2.5x | 2006 | 33% | 7.5% | -3.6% | 20.3% | 2.11% | 22.4% | **0.43%** | 75% | 12.4 | 45.4 | 90% | 0% |
| uSILVER | 3x | 2006 | 33% | 7.5% | -10.0% | 30.2% | 2.35% | 32.6% | **0.63%** | 85% | 15.4 | 55.1 | 94% | 0% |
| uWTIOIL | 2x | 2006 | 37% | -5.7% | -26.7% | 13.9% | 1.34% | 15.2% | **0.29%** | 48% | 14.0 | 50.7 | 97% | 0% |
| uWTIOIL | 2.5x | 2006 | 37% | -5.7% | -39.4% | 23.1% | 2.00% | 25.1% | **0.48%** | 69% | 14.8 | 49.9 | 99% | 10% |
| uWTIOIL | 3x | 2006 | 37% | -5.7% | -56.2% | 36.8% | 2.18% | 39.0% | **0.75%** | 120% | 17.6 | 59.0 | 100% | 10% |

| Token | Lev | Turnover ×eq/yr | Rebalances/yr | Defensive | Top-ups/yr | Fee+slip bps/yr | Hard delevers/yr | Liquidations/yr | Turnover (intraday) | Hard delevers (intraday) |
|---|---|---|---|---|---|---|---|---|---|---|
| uGOLD | 2x | 0.5 | 2 | 37% | 3 | 6 | 0.00 | 0.00 | 0.6 | 0.00 |
| uGOLD | 2.5x | 1.6 | 6 | 46% | 3 | 16 | 0.00 | 0.00 | 2.0 | 0.00 |
| uGOLD | 3x | 3.4 | 10 | 48% | 3 | 35 | 0.00 | 0.00 | 4.5 | 0.00 |
| uSILVER | 2x | 3.2 | 6 | 43% | 9 | 95 | 0.00 | 0.10 | 4.6 | 0.00 |
| uSILVER | 2.5x | 8.2 | 16 | 52% | 8 | 200 | 0.00 | 0.10 | 14.5 | 0.00 |
| uSILVER | 3x | 17.6 | 27 | 52% | 8 | 367 | 0.20 | 0.05 | 27.6 | 0.35 |
| uWTIOIL | 2x | 1.8 | 8 | 51% | 12 | 23 | 0.00 | 0.05 | 3.3 | 0.00 |
| uWTIOIL | 2.5x | 5.2 | 21 | 57% | 11 | 60 | 0.00 | 0.05 | 7.5 | 0.00 |
| uWTIOIL | 3x | 10.0 | 31 | 56% | 10 | 114 | 0.20 | 0.05 | 16.6 | 0.30 |

**Short tokens**

| Token | Lev | From | UL vol | UL /yr | Token /yr | Vol drag | Cost drag | Decay /yr | **Decay /wk** | Worst yr | TE mean | TE p95 | Worst-yr DD | Wipeout yrs |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| dGOLD | 2x | 2006 | 18% | 10.2% | -31.9% | 9.9% | 1.58% | 11.4% | **0.22%** | 42% | 7.0 | 25.5 | 71% | 0% |
| dGOLD | 2.5x | 2006 | 18% | 10.2% | -40.8% | 13.6% | 1.63% | 15.2% | **0.29%** | 46% | 6.8 | 24.5 | 79% | 0% |
| dGOLD | 3x | 2006 | 18% | 10.2% | -51.8% | 18.6% | 2.52% | 21.1% | **0.41%** | 73% | 8.2 | 29.2 | 85% | 0% |
| dSILVER | 2x | 2006 | 33% | 7.5% | -46.3% | 28.9% | 2.29% | 31.2% | **0.60%** | 84% | 12.7 | 45.1 | 95% | 0% |
| dSILVER | 2.5x | 2006 | 33% | 7.5% | -63.8% | 41.1% | 3.90% | 45.0% | **0.86%** | 122% | 13.9 | 48.6 | 97% | 5% |
| dSILVER | 3x | 2006 | 33% | 7.5% | -81.2% | 54.5% | 4.05% | 58.6% | **1.13%** | 150% | 17.6 | 63.1 | 99% | 10% |
| dWTIOIL | 2x | 2006 | 37% | -5.7% | -28.4% | 37.2% | 2.62% | 39.8% | **0.77%** | 114% | 14.7 | 49.6 | 90% | 0% |
| dWTIOIL | 2.5x | 2006 | 37% | -5.7% | -43.1% | 53.7% | 3.74% | 57.4% | **1.10%** | 165% | 15.5 | 52.1 | 94% | 0% |
| dWTIOIL | 3x | 2006 | 37% | -5.7% | -60.4% | 72.1% | 5.49% | 77.6% | **1.49%** | 216% | 19.6 | 67.4 | 97% | 5% |

| Token | Lev | Turnover ×eq/yr | Rebalances/yr | Defensive | Top-ups/yr | Fee+slip bps/yr | Hard delevers/yr | Liquidations/yr | Turnover (intraday) | Hard delevers (intraday) |
|---|---|---|---|---|---|---|---|---|---|---|
| dGOLD | 2x | 4.1 | 14 | 54% | 6 | 56 | 0.00 | 0.05 | 5.5 | 0.15 |
| dGOLD | 2.5x | 7.5 | 27 | 61% | 5 | 75 | 0.05 | 0.00 | 10.9 | 0.15 |
| dGOLD | 3x | 11.6 | 34 | 60% | 5 | 140 | 0.00 | 0.05 | 16.2 | 0.24 |
| dSILVER | 2x | 10.6 | 36 | 49% | 16 | 134 | 0.30 | 0.05 | 16.6 | 0.74 |
| dSILVER | 2.5x | 19.4 | 62 | 57% | 14 | 239 | 0.20 | 0.10 | 31.7 | 0.89 |
| dSILVER | 3x | 28.0 | 74 | 57% | 14 | 296 | 0.64 | 0.00 | 50.3 | 1.48 |
| dWTIOIL | 2x | 17.7 | 44 | 47% | 18 | 199 | 0.69 | 0.00 | 32.3 | 1.87 |
| dWTIOIL | 2.5x | 34.7 | 74 | 55% | 18 | 456 | 0.59 | 0.05 | 58.7 | 1.77 |
| dWTIOIL | 3x | 52.1 | 88 | 55% | 17 | 673 | 0.93 | 0.05 | 87.1 | 2.90 |

### Single names — TSLA, NVDA, AAPL, MSTR, COIN, HOOD

**Long tokens**

| Token | Lev | From | UL vol | UL /yr | Token /yr | Vol drag | Cost drag | Decay /yr | **Decay /wk** | Worst yr | TE mean | TE p95 | Worst-yr DD | Wipeout yrs |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| uTSLA | 2x | 2010 | 57% | 34.3% | 31.2% | 34.9% | 2.50% | 37.4% | **0.72%** | 79% | 21.5 | 74.3 | 90% | 0% |
| uTSLA | 2.5x | 2010 | 57% | 34.3% | 26.3% | 55.4% | 4.02% | 59.5% | **1.14%** | 107% | 22.7 | 82.7 | 95% | 0% |
| uTSLA | 3x | 2010 | 57% | 34.3% | 6.6% | 90.4% | 5.93% | 96.4% | **1.85%** | 179% | 28.3 | 101.0 | 97% | 0% |
| uNVDA | 2x | 2006 | 49% | 31.8% | 39.5% | 22.4% | 1.85% | 24.2% | **0.47%** | 72% | 17.9 | 65.2 | 98% | 5% |
| uNVDA | 2.5x | 2006 | 49% | 31.8% | 37.2% | 40.5% | 1.94% | 42.4% | **0.82%** | 113% | 20.3 | 74.0 | 99% | 5% |
| uNVDA | 3x | 2006 | 49% | 31.8% | 26.1% | 66.6% | 2.89% | 69.5% | **1.34%** | 204% | 23.7 | 83.7 | 100% | 5% |
| uAAPL | 2x | 2006 | 32% | 23.4% | 35.4% | 10.3% | 1.20% | 11.5% | **0.22%** | 44% | 11.4 | 40.6 | 87% | 0% |
| uAAPL | 2.5x | 2006 | 32% | 23.4% | 37.8% | 18.6% | 2.08% | 20.7% | **0.40%** | 62% | 12.8 | 44.7 | 92% | 0% |
| uAAPL | 3x | 2006 | 32% | 23.4% | 38.0% | 29.2% | 3.04% | 32.2% | **0.62%** | 103% | 15.3 | 53.8 | 96% | 5% |
| uMSTR | 2x | 2006 | 60% | 13.6% | -8.1% | 33.2% | 2.22% | 35.4% | **0.68%** | 150% | 20.5 | 76.1 | 98% | 5% |
| uMSTR | 2.5x | 2006 | 60% | 13.6% | -26.7% | 57.6% | 3.24% | 60.8% | **1.17%** | 215% | 22.3 | 82.2 | 100% | 5% |
| uMSTR | 3x | 2006 | 60% | 13.6% | -58.2% | 95.6% | 3.56% | 99.1% | **1.91%** | 377% | 27.7 | 99.9 | 100% | 14% |
| uCOIN | 2x | 2021 | 85% | -11.9% | -104.4% | 78.5% | 2.10% | 80.6% | **1.55%** | 169% | 36.1 | 117.6 | 99% | 17% |
| uCOIN | 2.5x | 2021 | 85% | -11.9% | -167.6% | 133.5% | 4.40% | 137.9% | **2.65%** | 270% | 37.9 | 130.8 | 100% | 17% |
| uCOIN | 3x | 2021 | 85% | -11.9% | -256.8% | 216.7% | 4.51% | 221.2% | **4.25%** | 417% | 47.9 | 163.0 | 100% | 50% |
| uHOOD | 2x | 2021 | 74% | 22.1% | -12.4% | 54.0% | 2.63% | 56.6% | **1.09%** | 106% | 29.9 | 95.5 | 100% | 20% |
| uHOOD | 2.5x | 2021 | 74% | 22.1% | -43.2% | 95.3% | 3.15% | 98.5% | **1.89%** | 163% | 33.4 | 107.7 | 100% | 20% |
| uHOOD | 3x | 2021 | 74% | 22.1% | -88.3% | 149.4% | 5.23% | 154.6% | **2.97%** | 231% | 42.1 | 135.6 | 100% | 20% |

| Token | Lev | Turnover ×eq/yr | Rebalances/yr | Defensive | Top-ups/yr | Fee+slip bps/yr | Hard delevers/yr | Liquidations/yr | Turnover (intraday) | Hard delevers (intraday) |
|---|---|---|---|---|---|---|---|---|---|---|
| uTSLA | 2x | 6.9 | 16 | 40% | 22 | 153 | 0.13 | 0.38 | 16.9 | 0.31 |
| uTSLA | 2.5x | 18.2 | 39 | 49% | 21 | 351 | 0.25 | 0.44 | 42.6 | 0.19 |
| uTSLA | 3x | 34.5 | 60 | 51% | 20 | 414 | 0.50 | 0.44 | 79.5 | 1.13 |
| uNVDA | 2x | 2.7 | 12 | 39% | 17 | 28 | 0.10 | 0.05 | 4.2 | 0.00 |
| uNVDA | 2.5x | 8.2 | 30 | 49% | 16 | 85 | 0.05 | 0.00 | 14.5 | 0.19 |
| uNVDA | 3x | 16.6 | 47 | 49% | 16 | 183 | 0.58 | 0.05 | 33.2 | 0.82 |
| uAAPL | 2x | 1.4 | 6 | 35% | 8 | 15 | 0.00 | 0.00 | 2.0 | 0.00 |
| uAAPL | 2.5x | 4.3 | 16 | 45% | 8 | 52 | 0.00 | 0.10 | 6.2 | 0.05 |
| uAAPL | 3x | 8.2 | 26 | 47% | 7 | 92 | 0.10 | 0.05 | 12.5 | 0.05 |
| uMSTR | 2x | 11.7 | 14 | 41% | 21 | 471 | 0.05 | 0.78 | 31.2 | 0.15 |
| uMSTR | 2.5x | 22.7 | 35 | 52% | 20 | 631 | 0.24 | 0.63 | 40.8 | 0.63 |
| uMSTR | 3x | 36.7 | 52 | 52% | 19 | 848 | 0.82 | 0.53 | 53.1 | 1.70 |
| uCOIN | 2x | 7.8 | 32 | 48% | 46 | 125 | 0.00 | 1.12 | 14.2 | 0.19 |
| uCOIN | 2.5x | 20.5 | 75 | 56% | 46 | 260 | 0.19 | 1.12 | 45.1 | 0.93 |
| uCOIN | 3x | 42.2 | 110 | 56% | 45 | 481 | 1.49 | 0.74 | 95.8 | 4.65 |
| uHOOD | 2x | 8.0 | 24 | 45% | 32 | 200 | 0.40 | 0.60 | 22.2 | 0.20 |
| uHOOD | 2.5x | 16.8 | 53 | 53% | 33 | 320 | 0.60 | 0.60 | 59.3 | 0.40 |
| uHOOD | 3x | 33.4 | 81 | 53% | 31 | 557 | 1.20 | 0.60 | 125.2 | 1.60 |

**Short tokens**

| Token | Lev | From | UL vol | UL /yr | Token /yr | Vol drag | Cost drag | Decay /yr | **Decay /wk** | Worst yr | TE mean | TE p95 | Worst-yr DD | Wipeout yrs |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| dTSLA | 2x | 2010 | 57% | 34.3% | -164.6% | 91.3% | 4.68% | 96.0% | **1.85%** | 180% | 23.9 | 91.7 | 99% | 19% |
| dTSLA | 2.5x | 2010 | 57% | 34.3% | -219.0% | 126.9% | 6.34% | 133.2% | **2.56%** | 225% | 27.2 | 102.8 | 100% | 25% |
| dTSLA | 3x | 2010 | 57% | 34.3% | -288.8% | 176.1% | 9.74% | 185.9% | **3.57%** | 321% | 34.6 | 133.0 | 100% | 31% |
| dNVDA | 2x | 2006 | 49% | 31.8% | -135.7% | 68.4% | 3.55% | 72.0% | **1.38%** | 195% | 19.7 | 70.5 | 97% | 5% |
| dNVDA | 2.5x | 2006 | 49% | 31.8% | -178.8% | 94.1% | 5.06% | 99.1% | **1.91%** | 288% | 21.6 | 75.9 | 99% | 24% |
| dNVDA | 3x | 2006 | 49% | 31.8% | -231.4% | 128.5% | 7.37% | 135.9% | **2.61%** | 394% | 28.2 | 103.6 | 99% | 38% |
| dAAPL | 2x | 2006 | 32% | 23.4% | -79.3% | 30.3% | 2.23% | 32.5% | **0.63%** | 103% | 12.5 | 44.4 | 91% | 0% |
| dAAPL | 2.5x | 2006 | 32% | 23.4% | -103.0% | 41.3% | 3.19% | 44.5% | **0.86%** | 159% | 12.8 | 48.4 | 95% | 0% |
| dAAPL | 3x | 2006 | 32% | 23.4% | -130.6% | 56.4% | 4.00% | 60.4% | **1.16%** | 221% | 15.6 | 56.1 | 97% | 10% |
| dMSTR | 2x | 2006 | 60% | 13.6% | -126.0% | 92.0% | 6.82% | 98.8% | **1.90%** | 367% | 24.5 | 87.3 | 100% | 14% |
| dMSTR | 2.5x | 2006 | 60% | 13.6% | -177.7% | 134.6% | 8.95% | 143.6% | **2.76%** | 518% | 29.7 | 105.8 | 100% | 19% |
| dMSTR | 3x | 2006 | 60% | 13.6% | -244.8% | 191.0% | 12.87% | 203.9% | **3.92%** | 732% | 38.0 | 139.6 | 100% | 29% |
| dCOIN | 2x | 2021 | 85% | -11.9% | -182.5% | 194.4% | 11.86% | 206.3% | **3.97%** | 407% | 40.3 | 171.9 | 99% | 17% |
| dCOIN | 2.5x | 2021 | 85% | -11.9% | -269.9% | 280.0% | 19.54% | 299.6% | **5.76%** | 615% | 54.3 | 205.1 | 100% | 33% |
| dCOIN | 3x | 2021 | 85% | -11.9% | -379.6% | 389.8% | 25.48% | 415.3% | **7.99%** | 837% | 74.7 | 301.4 | 100% | 67% |
| dHOOD | 2x | 2021 | 74% | 22.1% | -186.0% | 133.9% | 7.89% | 141.8% | **2.73%** | 212% | 37.8 | 118.7 | 100% | 20% |
| dHOOD | 2.5x | 2021 | 74% | 22.1% | -245.2% | 178.5% | 11.48% | 190.0% | **3.65%** | 279% | 47.5 | 149.9 | 100% | 20% |
| dHOOD | 3x | 2021 | 74% | 22.1% | -335.6% | 253.6% | 15.77% | 269.3% | **5.18%** | 422% | 61.1 | 195.4 | 100% | 60% |

| Token | Lev | Turnover ×eq/yr | Rebalances/yr | Defensive | Top-ups/yr | Fee+slip bps/yr | Hard delevers/yr | Liquidations/yr | Turnover (intraday) | Hard delevers (intraday) |
|---|---|---|---|---|---|---|---|---|---|---|
| dTSLA | 2x | 27.2 | 81 | 50% | 40 | 408 | 4.13 | 0.56 | 66.0 | 10.39 |
| dTSLA | 2.5x | 44.5 | 124 | 57% | 37 | 545 | 4.13 | 0.44 | 118.8 | 11.77 |
| dTSLA | 3x | 65.3 | 143 | 57% | 37 | 745 | 5.39 | 0.50 | 172.5 | 18.03 |
| dNVDA | 2x | 29.7 | 67 | 53% | 32 | 334 | 2.57 | 0.10 | 57.8 | 6.40 |
| dNVDA | 2.5x | 50.8 | 103 | 59% | 30 | 567 | 2.57 | 0.10 | 102.6 | 7.86 |
| dNVDA | 3x | 75.8 | 120 | 59% | 30 | 828 | 3.73 | 0.10 | 146.1 | 11.40 |
| dAAPL | 2x | 13.4 | 37 | 54% | 17 | 166 | 0.29 | 0.05 | 23.6 | 1.02 |
| dAAPL | 2.5x | 24.0 | 62 | 60% | 15 | 290 | 0.29 | 0.05 | 44.2 | 1.31 |
| dAAPL | 3x | 35.1 | 72 | 60% | 14 | 426 | 0.53 | 0.05 | 68.6 | 2.28 |
| dMSTR | 2x | 25.3 | 70 | 46% | 35 | 536 | 5.63 | 1.26 | 55.6 | 12.42 |
| dMSTR | 2.5x | 39.0 | 105 | 54% | 35 | 702 | 5.43 | 1.02 | 95.4 | 15.96 |
| dMSTR | 3x | 55.5 | 122 | 54% | 35 | 1101 | 6.74 | 1.16 | 133.4 | 23.04 |
| dCOIN | 2x | 81.0 | 142 | 43% | 72 | 2175 | 10.41 | 3.16 | 236.5 | 30.86 |
| dCOIN | 2.5x | 127.0 | 201 | 51% | 69 | 3218 | 11.15 | 3.35 | 357.5 | 39.03 |
| dCOIN | 3x | 183.7 | 226 | 51% | 70 | 4765 | 15.43 | 3.16 | 511.0 | 55.02 |
| dHOOD | 2x | 53.8 | 112 | 46% | 56 | 1414 | 6.00 | 1.80 | 120.7 | 16.99 |
| dHOOD | 2.5x | 99.2 | 161 | 53% | 53 | 3029 | 5.40 | 2.00 | 179.2 | 20.99 |
| dHOOD | 3x | 126.9 | 185 | 54% | 53 | 3500 | 7.60 | 1.80 | 235.3 | 33.19 |

### 3b. Per-week cost of holding, at 3x (D22)

The figure for the docs page. Holders' horizon is days to weeks, and the annualised number overstates what they pay: the same drag, spread across a year of average volatility.

**Long tokens**

| Token | UL vol | Decay 1 wk | 2 wks | 4 wks | 1 yr | Worst simulated year |
|---|---|---|---|---|---|---|
| uGOLD | 18% | 0.21% | 0.43% | 0.9% | 11.1% | 31% |
| uSILVER | 33% | 0.63% | 1.25% | 2.5% | 32.6% | 85% |
| uWTIOIL | 37% | 0.75% | 1.50% | 3.0% | 39.0% | 120% |
| uUS500 | 19% | 0.22% | 0.44% | 0.9% | 11.4% | 56% |
| uUS100 | 22% | 0.29% | 0.58% | 1.2% | 15.0% | 53% |
| uTSLA | 57% | 1.85% | 3.71% | 7.4% | 96.4% | 179% |
| uNVDA | 49% | 1.34% | 2.67% | 5.3% | 69.5% | 204% |
| uAAPL | 32% | 0.62% | 1.24% | 2.5% | 32.2% | 103% |
| uMSTR | 60% | 1.91% | 3.81% | 7.6% | 99.1% | 377% |
| uCOIN | 85% | 4.25% | 8.51% | 17.0% | 221.2% | 417% |
| uHOOD | 74% | 2.97% | 5.95% | 11.9% | 154.6% | 231% |

**Short tokens**

| Token | UL vol | Decay 1 wk | 2 wks | 4 wks | 1 yr | Worst simulated year |
|---|---|---|---|---|---|---|
| dGOLD | 18% | 0.41% | 0.81% | 1.6% | 21.1% | 73% |
| dSILVER | 33% | 1.13% | 2.25% | 4.5% | 58.6% | 150% |
| dWTIOIL | 37% | 1.49% | 2.98% | 6.0% | 77.6% | 216% |
| dUS500 | 19% | 0.43% | 0.86% | 1.7% | 22.3% | 97% |
| dUS100 | 22% | 0.57% | 1.14% | 2.3% | 29.6% | 100% |
| dTSLA | 57% | 3.57% | 7.15% | 14.3% | 185.9% | 321% |
| dNVDA | 49% | 2.61% | 5.23% | 10.5% | 135.9% | 394% |
| dAAPL | 32% | 1.16% | 2.32% | 4.6% | 60.4% | 221% |
| dMSTR | 60% | 3.92% | 7.84% | 15.7% | 203.9% | 732% |
| dCOIN | 85% | 7.99% | 15.97% | 31.9% | 415.3% | 837% |
| dHOOD | 74% | 5.18% | 10.36% | 20.7% | 269.3% | 422% |

Two things the docs page should say alongside it. First, this is an *average* week: realised cost scales with that week's variance, so the "worst simulated year" column is the honest companion — a bad year in uCOIN cost several times its average. Second, decay is not the dominant term over a week for anything except the high-vol single names; direction is. A holder losing money in uUS500 over a fortnight lost it to the index, not to the rebalancing.

---

## 4. Stress: the ten worst adverse gaps per asset

Close→open moves, ranked against a **long** token. "Health at floor" is venue equity over notional immediately after the gap, starting from the margin floor — the worst state the keeper leaves the position in. Under D21 it does not depend on leverage, so one column serves all three. `LIQ` marks it below Phoenix maintenance; what changes with leverage is how much the keeper has to unwind. The last column compares the NAV a liquidated 3x token keeps with what an un-liquidated 3x position would have held — where it reads higher, Phoenix's backstop has eaten the difference (headline 3).

**GOLD** (Gold, Phoenix GOLD, maintenance 3.5%)

| Date | Gap | Wknd | Health at floor | 2x → lev | 2.5x → lev | 3x → lev | NAV kept at 3x (actual vs no-liq) |
|---|---|---|---|---|---|---|---|
| 2026-01-30 | -6.0% |  | 6.9% | 2.14x in band | 2.76x delever | 3.44x delever | 82% vs 82% |
| 2013-04-15 | -5.5% | yes | 7.4% | 2.12x in band | 2.74x in band | 3.40x delever | 83% vs 83% |
| 2026-03-19 | -5.5% |  | 7.4% | 2.12x in band | 2.74x in band | 3.39x delever | 84% vs 84% |
| 2008-09-02 | -4.3% | yes | 8.5% | 2.10x in band | 2.68x in band | 3.30x in band | 87% vs 87% |
| 2025-10-21 | -4.1% |  | 8.7% | 2.09x in band | 2.67x in band | 3.28x in band | 88% vs 88% |
| 2013-06-20 | -4.1% |  | 8.7% | 2.09x in band | 2.67x in band | 3.28x in band | 88% vs 88% |
| 2020-03-16 | -4.0% | yes | 8.9% | 2.09x in band | 2.67x in band | 3.27x in band | 88% vs 88% |
| 2008-12-01 | -3.9% | yes | 8.9% | 2.09x in band | 2.66x in band | 3.27x in band | 88% vs 88% |
| 2020-11-09 | -3.7% | yes | 9.1% | 2.08x in band | 2.65x in band | 3.25x in band | 89% vs 89% |
| 2011-09-07 | -3.7% |  | 9.2% | 2.08x in band | 2.65x in band | 3.25x in band | 89% vs 89% |

**SILVER** (Silver, Phoenix SILVER, maintenance 3.5%)

| Date | Gap | Wknd | Health at floor | 2x → lev | 2.5x → lev | 3x → lev | NAV kept at 3x (actual vs no-liq) |
|---|---|---|---|---|---|---|---|
| 2026-01-30 | -15.4% |  | -3.4% | 2.44x **LIQ** | 3.44x **LIQ** | 4.71x **LIQ** | 63% vs 54% |
| 2026-02-05 | -13.8% |  | -1.5% | 2.38x **LIQ** | 3.29x **LIQ** | 4.42x **LIQ** | 63% vs 59% |
| 2020-03-16 | -13.6% | yes | -1.3% | 2.37x **LIQ** | 3.27x **LIQ** | 4.38x **LIQ** | 63% vs 59% |
| 2026-03-19 | -9.9% |  | 2.9% | 2.25x **LIQ** | 2.99x **LIQ** | 3.84x **LIQ** | 63% vs 70% |
| 2008-12-01 | -9.4% | yes | 3.4% | 2.23x **LIQ** | 2.96x **LIQ** | 3.78x **LIQ** | 63% vs 72% |
| 2013-04-15 | -9.1% | yes | 3.7% | 2.22x in band | 2.94x delever | 3.75x delever | 73% vs 73% |
| 2026-03-03 | -9.0% |  | 3.8% | 2.22x in band | 2.94x delever | 3.74x delever | 73% vs 73% |
| 2011-09-23 | -8.9% |  | 3.9% | 2.22x in band | 2.93x delever | 3.73x delever | 73% vs 73% |
| 2011-09-26 | -8.2% | yes | 4.6% | 2.20x in band | 2.89x delever | 3.66x delever | 75% vs 75% |
| 2025-12-29 | -7.7% | yes | 5.2% | 2.18x in band | 2.86x delever | 3.60x delever | 77% vs 77% |

**WTIOIL** (WTI Crude, Phoenix WTIOIL, maintenance 3.5%)

| Date | Gap | Wknd | Health at floor | 2x → lev | 2.5x → lev | 3x → lev | NAV kept at 3x (actual vs no-liq) |
|---|---|---|---|---|---|---|---|
| 2020-03-09 | -21.9% | yes | -12.0% | 2.78x **LIQ** | 4.31x **LIQ** | 6.82x **LIQ** | 63% vs 34% |
| 2020-04-21 | -14.9% |  | -2.9% | 2.43x **LIQ** | 3.39x **LIQ** | 4.62x **LIQ** | 63% vs 55% |
| 2026-04-08 | -13.8% |  | -1.5% | 2.38x **LIQ** | 3.29x **LIQ** | 4.41x **LIQ** | 63% vs 59% |
| 2020-03-16 | -12.5% | yes | 0.0% | 2.33x **LIQ** | 3.18x **LIQ** | 4.20x **LIQ** | 63% vs 63% |
| 2020-03-18 | -11.4% |  | 1.3% | 2.29x **LIQ** | 3.10x **LIQ** | 4.04x **LIQ** | 63% vs 66% |
| 2020-04-20 | -10.7% | yes | 2.0% | 2.27x **LIQ** | 3.05x **LIQ** | 3.94x **LIQ** | 63% vs 68% |
| 2020-04-27 | -10.5% | yes | 2.2% | 2.27x **LIQ** | 3.03x **LIQ** | 3.92x **LIQ** | 63% vs 68% |
| 2008-10-24 | -9.2% |  | 3.6% | 2.23x in band | 2.95x delever | 3.77x delever | 72% vs 72% |
| 2026-04-17 | -8.8% |  | 4.0% | 2.21x in band | 2.92x delever | 3.72x delever | 74% vs 74% |
| 2025-04-04 | -7.4% |  | 5.5% | 2.17x in band | 2.84x delever | 3.57x delever | 78% vs 78% |

**US500** (US 500, Phoenix SPY, maintenance 3.5%)

| Date | Gap | Wknd | Health at floor | 2x → lev | 2.5x → lev | 3x → lev | NAV kept at 3x (actual vs no-liq) |
|---|---|---|---|---|---|---|---|
| 2020-03-16 | -10.4% | yes | 2.3% | 2.26x **LIQ** | 3.03x **LIQ** | 3.91x **LIQ** | 63% vs 69% |
| 2008-10-24 | -8.3% |  | 4.6% | 2.20x in band | 2.89x delever | 3.67x delever | 75% vs 75% |
| 2020-03-09 | -7.4% | yes | 5.5% | 2.18x in band | 2.84x delever | 3.58x delever | 78% vs 78% |
| 2020-03-12 | -6.7% |  | 6.2% | 2.15x in band | 2.80x delever | 3.50x delever | 80% vs 80% |
| 2020-03-18 | -6.5% |  | 6.4% | 2.15x in band | 2.79x delever | 3.49x delever | 80% vs 80% |
| 2015-08-24 | -5.1% | yes | 7.8% | 2.11x in band | 2.72x in band | 3.36x delever | 85% vs 85% |
| 2008-10-10 | -4.3% |  | 8.5% | 2.10x in band | 2.68x in band | 3.30x in band | 87% vs 87% |
| 2024-08-05 | -4.0% | yes | 8.9% | 2.09x in band | 2.67x in band | 3.27x in band | 88% vs 88% |
| 2020-04-01 | -3.8% |  | 9.1% | 2.08x in band | 2.66x in band | 3.26x in band | 89% vs 89% |
| 2008-01-22 | -3.7% | yes | 9.2% | 2.08x in band | 2.65x in band | 3.25x in band | 89% vs 89% |

**US100** (US Tech 100, Phoenix QQQ, maintenance 3.5%)

| Date | Gap | Wknd | Health at floor | 2x → lev | 2.5x → lev | 3x → lev | NAV kept at 3x (actual vs no-liq) |
|---|---|---|---|---|---|---|---|
| 2020-03-16 | -9.5% | yes | 3.4% | 2.23x **LIQ** | 2.96x **LIQ** | 3.79x **LIQ** | 63% vs 72% |
| 2015-08-24 | -8.0% | yes | 4.9% | 2.19x in band | 2.87x delever | 3.63x delever | 76% vs 76% |
| 2008-10-24 | -7.7% |  | 5.2% | 2.18x in band | 2.86x delever | 3.60x delever | 77% vs 77% |
| 2020-03-09 | -7.0% | yes | 5.9% | 2.16x in band | 2.82x delever | 3.53x delever | 79% vs 79% |
| 2020-03-12 | -6.7% |  | 6.2% | 2.16x in band | 2.80x delever | 3.51x delever | 80% vs 80% |
| 2020-03-18 | -5.8% |  | 7.1% | 2.13x in band | 2.75x delever | 3.42x delever | 83% vs 83% |
| 2008-01-22 | -5.6% | yes | 7.4% | 2.13x in band | 2.74x in band | 3.40x delever | 83% vs 83% |
| 2024-08-05 | -5.4% | yes | 7.5% | 2.12x in band | 2.73x in band | 3.38x delever | 84% vs 84% |
| 2025-04-03 | -4.1% |  | 8.7% | 2.09x in band | 2.67x in band | 3.28x in band | 88% vs 88% |
| 2020-09-08 | -4.0% | yes | 8.9% | 2.09x in band | 2.67x in band | 3.27x in band | 88% vs 88% |

**TSLA** (Tesla, Phoenix TSLA, maintenance 3.5%)

| Date | Gap | Wknd | Health at floor | 2x → lev | 2.5x → lev | 3x → lev | NAV kept at 3x (actual vs no-liq) |
|---|---|---|---|---|---|---|---|
| 2020-09-08 | -14.9% | yes | -2.8% | 2.42x **LIQ** | 3.39x **LIQ** | 4.62x **LIQ** | 63% vs 55% |
| 2020-03-16 | -14.1% | yes | -1.9% | 2.39x **LIQ** | 3.32x **LIQ** | 4.47x **LIQ** | 63% vs 58% |
| 2020-03-09 | -13.9% | yes | -1.7% | 2.39x **LIQ** | 3.30x **LIQ** | 4.44x **LIQ** | 63% vs 58% |
| 2013-11-06 | -12.4% |  | 0.1% | 2.33x **LIQ** | 3.18x **LIQ** | 4.19x **LIQ** | 63% vs 63% |
| 2015-08-24 | -12.1% | yes | 0.4% | 2.32x **LIQ** | 3.15x **LIQ** | 4.14x **LIQ** | 63% vs 64% |
| 2018-09-28 | -12.1% |  | 0.4% | 2.32x **LIQ** | 3.15x **LIQ** | 4.14x **LIQ** | 63% vs 64% |
| 2019-07-25 | -11.8% |  | 0.7% | 2.31x **LIQ** | 3.13x **LIQ** | 4.10x **LIQ** | 63% vs 64% |
| 2024-08-05 | -10.8% | yes | 1.9% | 2.28x **LIQ** | 3.06x **LIQ** | 3.96x **LIQ** | 63% vs 68% |
| 2019-04-04 | -10.3% |  | 2.5% | 2.26x **LIQ** | 3.02x **LIQ** | 3.89x **LIQ** | 63% vs 69% |
| 2011-12-08 | -9.8% |  | 3.0% | 2.24x **LIQ** | 2.99x **LIQ** | 3.83x **LIQ** | 63% vs 71% |

**NVDA** (Nvidia, Phoenix NVDA, maintenance 3.5%)

| Date | Gap | Wknd | Health at floor | 2x → lev | 2.5x → lev | 3x → lev | NAV kept at 3x (actual vs no-liq) |
|---|---|---|---|---|---|---|---|
| 2008-07-03 | -28.0% |  | -21.6% | 3.27x **LIQ** | 6.01x **LIQ** | 13.54x **LIQ** | 63% vs 16% |
| 2018-11-16 | -19.3% |  | -8.4% | 2.63x **LIQ** | 3.90x **LIQ** | 5.75x **LIQ** | 63% vs 42% |
| 2019-01-28 | -14.7% | yes | -2.6% | 2.42x **LIQ** | 3.37x **LIQ** | 4.58x **LIQ** | 63% vs 56% |
| 2024-08-05 | -14.2% | yes | -2.0% | 2.40x **LIQ** | 3.32x **LIQ** | 4.48x **LIQ** | 63% vs 57% |
| 2025-01-27 | -12.5% | yes | 0.0% | 2.33x **LIQ** | 3.18x **LIQ** | 4.20x **LIQ** | 63% vs 63% |
| 2020-03-16 | -11.8% | yes | 0.8% | 2.31x **LIQ** | 3.13x **LIQ** | 4.10x **LIQ** | 63% vs 65% |
| 2020-03-09 | -9.8% | yes | 3.0% | 2.24x **LIQ** | 2.99x **LIQ** | 3.84x **LIQ** | 63% vs 71% |
| 2008-01-22 | -9.5% | yes | 3.3% | 2.24x **LIQ** | 2.97x **LIQ** | 3.80x **LIQ** | 63% vs 71% |
| 2008-10-10 | -8.7% |  | 4.2% | 2.21x in band | 2.92x delever | 3.70x delever | 74% vs 74% |
| 2020-03-12 | -8.5% |  | 4.4% | 2.21x in band | 2.91x delever | 3.69x delever | 74% vs 74% |

**AAPL** (Apple, Phoenix AAPL, maintenance 3.5%)

| Date | Gap | Wknd | Health at floor | 2x → lev | 2.5x → lev | 3x → lev | NAV kept at 3x (actual vs no-liq) |
|---|---|---|---|---|---|---|---|
| 2020-03-16 | -13.0% | yes | -0.5% | 2.35x **LIQ** | 3.22x **LIQ** | 4.27x **LIQ** | 63% vs 61% |
| 2008-01-23 | -12.5% |  | -0.0% | 2.33x **LIQ** | 3.18x **LIQ** | 4.20x **LIQ** | 63% vs 62% |
| 2013-01-24 | -10.5% |  | 2.2% | 2.27x **LIQ** | 3.03x **LIQ** | 3.92x **LIQ** | 63% vs 68% |
| 2008-07-22 | -10.4% |  | 2.3% | 2.26x **LIQ** | 3.03x **LIQ** | 3.91x **LIQ** | 63% vs 69% |
| 2015-08-24 | -10.3% | yes | 2.5% | 2.26x **LIQ** | 3.02x **LIQ** | 3.89x **LIQ** | 63% vs 69% |
| 2024-08-05 | -9.4% | yes | 3.4% | 2.23x **LIQ** | 2.96x **LIQ** | 3.79x **LIQ** | 63% vs 72% |
| 2019-01-03 | -8.8% |  | 4.0% | 2.21x in band | 2.92x delever | 3.72x delever | 74% vs 74% |
| 2020-03-09 | -8.7% | yes | 4.1% | 2.21x in band | 2.92x delever | 3.71x delever | 74% vs 74% |
| 2026-07-31 | -8.6% |  | 4.3% | 2.21x in band | 2.91x delever | 3.69x delever | 74% vs 74% |
| 2008-01-22 | -8.2% | yes | 4.6% | 2.20x in band | 2.89x delever | 3.66x delever | 75% vs 75% |

**MSTR** (Strategy, Phoenix MSTR, maintenance 3.5%)

| Date | Gap | Wknd | Health at floor | 2x → lev | 2.5x → lev | 3x → lev | NAV kept at 3x (actual vs no-liq) |
|---|---|---|---|---|---|---|---|
| 2024-08-05 | -27.4% | yes | -20.5% | 3.21x **LIQ** | 5.75x **LIQ** | 12.18x **LIQ** | 63% vs 18% |
| 2022-06-13 | -26.6% | yes | -19.2% | 3.13x **LIQ** | 5.47x **LIQ** | 10.87x **LIQ** | 63% vs 20% |
| 2008-05-02 | -15.5% |  | -3.6% | 2.45x **LIQ** | 3.45x **LIQ** | 4.74x **LIQ** | 63% vs 53% |
| 2007-05-08 | -15.0% |  | -2.9% | 2.43x **LIQ** | 3.40x **LIQ** | 4.63x **LIQ** | 63% vs 55% |
| 2022-01-24 | -14.9% | yes | -2.8% | 2.42x **LIQ** | 3.39x **LIQ** | 4.61x **LIQ** | 63% vs 55% |
| 2010-11-02 | -14.6% |  | -2.5% | 2.41x **LIQ** | 3.37x **LIQ** | 4.57x **LIQ** | 63% vs 56% |
| 2012-10-31 | -13.7% | yes | -1.4% | 2.38x **LIQ** | 3.28x **LIQ** | 4.40x **LIQ** | 63% vs 59% |
| 2021-05-19 | -13.2% |  | -0.8% | 2.36x **LIQ** | 3.24x **LIQ** | 4.31x **LIQ** | 63% vs 60% |
| 2010-02-04 | -13.1% |  | -0.7% | 2.35x **LIQ** | 3.23x **LIQ** | 4.29x **LIQ** | 63% vs 61% |
| 2024-03-19 | -11.8% |  | 0.8% | 2.31x **LIQ** | 3.12x **LIQ** | 4.09x **LIQ** | 63% vs 65% |

**COIN** (Coinbase, Phoenix COIN, maintenance 3.5%)

| Date | Gap | Wknd | Health at floor | 2x → lev | 2.5x → lev | 3x → lev | NAV kept at 3x (actual vs no-liq) |
|---|---|---|---|---|---|---|---|
| 2022-05-11 | -24.9% |  | -16.4% | 2.99x **LIQ** | 4.96x **LIQ** | 8.87x **LIQ** | 63% vs 25% |
| 2022-06-13 | -21.3% | yes | -11.2% | 2.74x **LIQ** | 4.22x **LIQ** | 6.56x **LIQ** | 63% vs 36% |
| 2024-08-05 | -20.8% | yes | -10.4% | 2.71x **LIQ** | 4.12x **LIQ** | 6.30x **LIQ** | 63% vs 38% |
| 2023-03-23 | -19.8% |  | -9.1% | 2.66x **LIQ** | 3.97x **LIQ** | 5.93x **LIQ** | 63% vs 41% |
| 2023-06-06 | -19.8% |  | -9.1% | 2.65x **LIQ** | 3.97x **LIQ** | 5.91x **LIQ** | 63% vs 41% |
| 2025-08-01 | -11.3% |  | 1.4% | 2.29x **LIQ** | 3.09x **LIQ** | 4.02x **LIQ** | 63% vs 66% |
| 2021-05-19 | -10.9% |  | 1.8% | 2.28x **LIQ** | 3.06x **LIQ** | 3.97x **LIQ** | 63% vs 67% |
| 2022-01-24 | -9.8% | yes | 3.0% | 2.24x **LIQ** | 2.99x **LIQ** | 3.83x **LIQ** | 63% vs 71% |
| 2022-05-12 | -9.6% |  | 3.2% | 2.24x **LIQ** | 2.98x **LIQ** | 3.81x **LIQ** | 63% vs 71% |
| 2023-02-03 | -9.5% |  | 3.3% | 2.23x **LIQ** | 2.97x **LIQ** | 3.80x **LIQ** | 63% vs 71% |

**HOOD** (Robinhood, Phoenix HOOD, maintenance 3.5%)

| Date | Gap | Wknd | Health at floor | 2x → lev | 2.5x → lev | 3x → lev | NAV kept at 3x (actual vs no-liq) |
|---|---|---|---|---|---|---|---|
| 2024-08-05 | -17.8% | yes | -6.4% | 2.55x **LIQ** | 3.70x **LIQ** | 5.29x **LIQ** | 63% vs 47% |
| 2022-01-28 | -14.1% |  | -1.9% | 2.39x **LIQ** | 3.32x **LIQ** | 4.47x **LIQ** | 63% vs 58% |
| 2026-04-29 | -11.9% |  | 0.7% | 2.31x **LIQ** | 3.14x **LIQ** | 4.11x **LIQ** | 63% vs 64% |
| 2025-04-07 | -11.1% | yes | 1.6% | 2.29x **LIQ** | 3.08x **LIQ** | 4.00x **LIQ** | 63% vs 67% |
| 2025-04-03 | -11.0% |  | 1.7% | 2.28x **LIQ** | 3.07x **LIQ** | 3.99x **LIQ** | 63% vs 67% |
| 2021-08-05 | -10.6% |  | 2.1% | 2.27x **LIQ** | 3.04x **LIQ** | 3.94x **LIQ** | 63% vs 68% |
| 2022-04-29 | -10.3% |  | 2.4% | 2.26x **LIQ** | 3.02x **LIQ** | 3.90x **LIQ** | 63% vs 69% |
| 2023-11-08 | -10.0% |  | 2.8% | 2.25x **LIQ** | 3.00x **LIQ** | 3.86x **LIQ** | 63% vs 70% |
| 2021-08-19 | -9.5% |  | 3.3% | 2.24x **LIQ** | 2.97x **LIQ** | 3.80x **LIQ** | 63% vs 71% |
| 2021-10-27 | -9.1% |  | 3.8% | 2.22x in band | 2.94x delever | 3.75x delever | 73% vs 73% |

For contrast, US500 under **D7's NAV basis** — the rows that read "delever" above turn red here:

| Date | Gap | Wknd | Health at floor | 2x → lev | 2.5x → lev | 3x → lev | NAV kept at 3x (actual vs no-liq) |
|---|---|---|---|---|---|---|---|
| 2020-03-16 | -10.4% | yes | 2.3% / -0.5% / -2.4% | 2.26x **LIQ** | 3.03x **LIQ** | 3.91x **LIQ** | 75% vs 69% |
| 2008-10-24 | -8.3% |  | 4.6% / 1.8% / 0.0% | 2.20x in band | 2.89x **LIQ** | 3.67x **LIQ** | 75% vs 75% |
| 2020-03-09 | -7.4% | yes | 5.5% / 2.8% / 1.0% | 2.18x in band | 2.84x **LIQ** | 3.58x **LIQ** | 75% vs 78% |
| 2020-03-12 | -6.7% |  | 6.2% / 3.5% / 1.8% | 2.15x in band | 2.80x delever | 3.50x **LIQ** | 75% vs 80% |
| 2020-03-18 | -6.5% |  | 6.4% / 3.7% / 1.9% | 2.15x in band | 2.79x delever | 3.49x **LIQ** | 75% vs 80% |
| 2015-08-24 | -5.1% | yes | 7.8% / 5.1% / 3.4% | 2.11x in band | 2.72x in band | 3.36x **LIQ** | 75% vs 85% |
| 2008-10-10 | -4.3% |  | 8.5% / 5.9% / 4.2% | 2.10x in band | 2.68x in band | 3.30x in band | 87% vs 87% |
| 2024-08-05 | -4.0% | yes | 8.9% / 6.3% / 4.5% | 2.09x in band | 2.67x in band | 3.27x in band | 88% vs 88% |
| 2020-04-01 | -3.8% |  | 9.1% / 6.5% / 4.7% | 2.08x in band | 2.66x in band | 3.26x in band | 89% vs 89% |
| 2008-01-22 | -3.7% | yes | 9.2% / 6.6% / 4.8% | 2.08x in band | 2.65x in band | 3.25x in band | 89% vs 89% |

---

## 5. Funding scenarios

Funding is the one input with no history. Cost drag at 3x under a flat ±10%/yr rate, against the 0% default. Longs pay when the rate is positive.

| Token | Cost drag @ 0% | @ +10%/yr | @ −10%/yr | Swing | Swing /wk |
|---|---|---|---|---|---|
| uGOLD | 1.56% | 31.25% | -29.08% | 60.33% | 1.16% |
| dGOLD | 2.52% | -27.69% | 32.40% | 60.09% | 1.16% |
| uSILVER | 2.35% | 33.81% | -26.76% | 60.58% | 1.16% |
| dSILVER | 4.05% | -25.94% | 34.48% | 60.42% | 1.16% |
| uWTIOIL | 2.18% | 34.63% | -28.46% | 63.08% | 1.21% |
| dWTIOIL | 5.49% | -25.34% | 36.17% | 61.51% | 1.18% |
| uUS500 | 1.65% | 31.90% | -27.36% | 59.26% | 1.14% |
| dUS500 | 2.25% | -28.86% | 32.72% | 61.58% | 1.18% |
| uUS100 | 1.97% | 31.77% | -26.61% | 58.37% | 1.12% |
| dUS100 | 3.06% | -26.69% | 33.78% | 60.46% | 1.16% |
| uTSLA | 5.93% | 37.49% | -22.25% | 59.74% | 1.15% |
| dTSLA | 9.74% | -20.70% | 39.53% | 60.22% | 1.16% |
| uNVDA | 2.89% | 32.29% | -25.00% | 57.29% | 1.10% |
| dNVDA | 7.37% | -22.55% | 36.86% | 59.42% | 1.14% |
| uAAPL | 3.04% | 31.82% | -27.29% | 59.11% | 1.14% |
| dAAPL | 4.00% | -26.83% | 33.67% | 60.50% | 1.16% |
| uMSTR | 3.56% | 36.51% | -25.18% | 61.68% | 1.19% |
| dMSTR | 12.87% | -17.33% | 43.29% | 60.62% | 1.17% |
| uCOIN | 4.51% | 36.35% | -25.48% | 61.82% | 1.19% |
| dCOIN | 25.48% | -3.54% | 55.14% | 58.68% | 1.13% |
| uHOOD | 5.23% | 34.69% | -25.48% | 60.17% | 1.16% |
| dHOOD | 15.77% | -21.35% | 43.65% | 65.00% | 1.25% |

A 3x token carries 3× notional, so a 10%/yr funding rate costs the paying side ~30%/yr — larger than every fee in D8 combined by an order of magnitude, and comparable to the volatility drag itself. The "Swing" column is the full ±10% range, so ~60%/yr and ~1.16%/week; one-sided that is ~0.58%/week, **more than uUS500's 0.22%/week of decay at the same leverage.** At an index holder's horizon funding, not decay, is the dominant cost. Instrument it from day one (`snapshots.funding_rate_bps_hourly` is already in the schema) and revisit after two weeks of live data.

---

## 6. Leverage, for the choice D22 defers to #13

D22 keeps the registry at 3x for every token and moves the tiering decision to vault creation. These are the numbers for it. Since D21 makes solvency leverage-invariant, the rule is about decay, forced trading and tracking: drop to 2.5x if, at 3x, the long token decays more than 50%/yr, loses >95% in more than 5% of simulated years, shows more than 1.5 hard delevers/yr on the intraday model, breaks 50 bps/day tracking p95, or the asset's history gaps past the 3x hard-delever distance more than once every two years.

| Asset | Suggested | Why | Short side |
|---|---|---|---|
| GOLD | **3x** | decay 11.1%/yr (0.21%/wk), tracking p95 31 bps/day, 0.0 hard delevers/yr, 0% wipeout years, worst gap 6.0% against an 11.1% hard delever | tracks acceptably at the same level (p95 29 bps, 0% wipeout years) |
| SILVER | **2.5x** | tracking p95 55 bps/day at 3x (45 at 2.5x) against PRD §6's 25 | tracks acceptably at the same level (p95 63 bps, 10% wipeout years) |
| WTIOIL | **2.5x** | 10% of simulated years lose >95% at 3x (10% at 2.5x); tracking p95 59 bps/day at 3x (50 at 2.5x) against PRD §6's 25 | tracks acceptably at the same level (p95 67 bps, 5% wipeout years) |
| US500 | **3x** | decay 11.4%/yr (0.22%/wk), tracking p95 30 bps/day, 0.1 hard delevers/yr, 0% wipeout years, worst gap 10.4% against an 11.1% hard delever | tracks acceptably at the same level (p95 31 bps, 0% wipeout years) |
| US100 | **3x** | decay 15.0%/yr (0.29%/wk), tracking p95 38 bps/day, 0.1 hard delevers/yr, 0% wipeout years, worst gap 9.5% against an 11.1% hard delever | tracks acceptably at the same level (p95 38 bps, 0% wipeout years) |
| TSLA | **2.5x** | decays 96%/yr — 1.85%/wk — against 59% and 1.14%/wk at 2.5x; tracking p95 101 bps/day at 3x (83 at 2.5x) against PRD §6's 25 | materially worse (31% wipeout years, p95 133 bps): one notch lower, or do not list it |
| NVDA | **2.5x** | decays 69%/yr — 1.34%/wk — against 42% and 0.82%/wk at 2.5x; tracking p95 84 bps/day at 3x (74 at 2.5x) against PRD §6's 25 | materially worse (38% wipeout years, p95 104 bps): one notch lower, or do not list it |
| AAPL | **2.5x** | tracking p95 54 bps/day at 3x (45 at 2.5x) against PRD §6's 25 | tracks acceptably at the same level (p95 56 bps, 10% wipeout years) |
| MSTR | **2.5x** | decays 99%/yr — 1.91%/wk — against 61% and 1.17%/wk at 2.5x; 14% of simulated years lose >95% at 3x (5% at 2.5x); 1.7 hard delevers/yr at 3x on the intraday model; tracking p95 100 bps/day at 3x (82 at 2.5x) against PRD §6's 25; 14 gaps ≥ 11.1% in 21y (0.68/yr) blow straight through the 3x hard delever | materially worse (29% wipeout years, p95 140 bps): one notch lower, or do not list it |
| COIN | **2.5x** | decays 221%/yr — 4.25%/wk — against 138% and 2.65%/wk at 2.5x; 50% of simulated years lose >95% at 3x (17% at 2.5x); 4.6 hard delevers/yr at 3x on the intraday model; tracking p95 163 bps/day at 3x (131 at 2.5x) against PRD §6's 25; 6 gaps ≥ 11.1% in 5y (1.11/yr) blow straight through the 3x hard delever | materially worse (67% wipeout years, p95 301 bps): one notch lower, or do not list it |
| HOOD | **2.5x** | decays 155%/yr — 2.97%/wk — against 98% and 1.89%/wk at 2.5x; 20% of simulated years lose >95% at 3x (20% at 2.5x); 1.6 hard delevers/yr at 3x on the intraday model; tracking p95 136 bps/day at 3x (108 at 2.5x) against PRD §6's 25; 4 gaps ≥ 11.1% in 5y (0.78/yr) blow straight through the 3x hard delever | materially worse (60% wipeout years, p95 195 bps): one notch lower, or do not list it |

**Summary: 3x on GOLD, US500, US100; 2.5x on SILVER, WTIOIL, TSLA, NVDA, AAPL, MSTR, COIN, HOOD.**

Where the call is close:

- **SILVER** — borderline: the only trigger is tracking p95 55 bps, just over the 50 bps line. 3x is defensible if a 3x product is allowed ~2× PRD §6's tracking target.
- **TSLA** — even 2.5x decays 59%/yr (1.14%/wk), and 2x 37%/yr — a day-trade instrument at any level; 2x is the honest listing.
- **AAPL** — borderline: the only trigger is tracking p95 54 bps, just over the 50 bps line. 3x is defensible if a 3x product is allowed ~2× PRD §6's tracking target.
- **MSTR** — even 2.5x decays 61%/yr (1.17%/wk), and 2x 35%/yr — a day-trade instrument at any level; 2x is the honest listing.
- **COIN** — even 2.5x decays 138%/yr (2.65%/wk), and 2x 81%/yr — a day-trade instrument at any level; 2x is the honest listing.
- **HOOD** — even 2.5x decays 98%/yr (1.89%/wk), and 2x 57%/yr — a day-trade instrument at any level; 2x is the honest listing.

D5 launches US500 first and it has the most headroom either way, so nothing about the launch order depends on this. Leverage is a per-token field (`TokenDef.policy`), so applying it at #13 is a registry edit — but D22 is right that it cannot change on a live token, so it has to be right at vault creation.

---

## 7. Parameter status

| Parameter | Status | Note |
|---|---|---|
| Leverage bands | **D6, unchanged** | Band width sets turnover, and turnover costs tens of bps/yr at every level tested — nothing next to the drag. Tightening buys nothing; widening raises tracking error without lowering risk, because the hard delever and the margin floor bound the tail. |
| Margin bands | **D21, adopted** — 17.5 / 12.5 / 22.5% of notional | §2c measures what it bought. Identical to D7 at 2x. |
| Hard delever | **D6, unchanged**; express the on-chain backstop as a *price* | At 3x the hard delever needs an 11.1% adverse move and liquidation 9.3% from the floor — the keeper acts first only if it is awake. Set D7's Phoenix conditional stop at the **margin-floor price**, refreshed on every rebalance; the dev seed already writes that price into `snapshots.stop_price` so the keeper has a reference implementation. |
| Expansion gating | **D6, unchanged** | Costs a little tracking error on Monday mornings; prevents buying an impact-pricing print that has not snapped yet. |
| Per-token leverage | **D22, deferred to #13** | §6 has the numbers. |
| Short tokens | **open** | A short's drag is L(L+1)σ²/2 rather than L(L−1)σ²/2, and it is short a positive drift. §3 shows every single-name short losing the lot inside a year at 3x. Recommend launching shorts only on the index and metal underlyings, one notch lower. |
| Maintenance margin | **estimate, Q13 open** | 3.5% from the registry. Every liquidation number in this sheet moves with it. |

---

## 8. What this sheet does not cover

- **Flows.** One deposit per segment, held. Mint/redeem impact, the stale-NAV arb in D8 and the cap logic are out of scope; D12 already says mints, not rebalances, are the impact risk.
- **OI caps.** PLAN.md §0c puts them near $3M per market, so a 3x token caps near $1M TVL against a 2x token's $1.5M. A business constraint on the leverage choice this sheet does not price.
- **Phoenix ADL / backstop.** Modelled as "the venue absorbs negative equity". A real auto-deleverage would also close the *winning* side of a token.
- **The impact-pricing bound.** Weekend gaps are applied in full at the reopen; the real 4–5% cap plus snap is more forgiving.
- **Commodity levels.** WTIOIL is USO, a futures-roll fund: its volatility and gap shape are usable, its decay-versus-buy-and-hold is not a statement about spot crude.
- **Pre-2006 history**, trimmed out of `data/`. MSTR fell 62% in a day in March 2000 and AAPL 52% in September 2000; nothing at any leverage survives those, and the 2026 order book is not the 2000 one.
- **Correlated venue failure.** A Phoenix outage during a gap is the scenario the on-chain conditional stop exists for, and it is not simulated here.