# Listing grades

Generated by `pnpm --filter @lev/sim grade` (`packages/sim/src/grade.ts`) — do not edit by hand.
Venue parameters read 2026-09-12 from `https://perp-api.phoenix.trade/v1/view/markets`; price history from the committed CSVs under `packages/sim/data`.

**Since D26 the tier below is a risk grade, not a gate.** What we list, and at what leverage, is an operator decision: 3x on the two indexes, 2x on everything else, including names this rule would refuse outright. The grade is what gets shown on the token page and drives the disclosure next to the mint button; *Cost of holding, last year* is the number that disclosure quotes. Where the grade and the listing disagree, the disagreement is the product, and it is printed rather than resolved.

**Planning only.** Per `AGENTS.md` rule 5 the keeper reads live market state from the venue; nothing in this file or in `packages/sim/data` may reach a signing path.

## In plain English

What we sell each market at, what the risk grade says about it, and what holding it has actually cost. "A typical 30-day hold" is the middle outcome across the rolling months of the last year, simulated under our own keeper and fees. "Paid to the leverage itself" is the part of the loss that is not simply being twice or three times as exposed.

- **SPY: 3x.** The risk grade agrees. Steady enough for our highest leverage. A typical 30-day hold at 3x over the last year has made 4%, paid 0.5% to the leverage itself (1.6% in the worst tenth of months), and no holder in the last year lost half.
- **QQQ: 3x.** The risk grade is 2.5x, below what we sell it at. Too volatile for 3x; it swings 23% a year against the S&P's 19%. A typical 30-day hold at 3x over the last year has broken even, paid 0.2% to the leverage itself (2.1% in the worst tenth of months), and no holder in the last year lost half.
- **GOLD: 2x.** The risk grade agrees. Too volatile for 2.5x; it swings 29% a year against the S&P's 19%. A typical 30-day hold at 2x over the last year has made 6%, paid 0.8% to the leverage itself (3.3% in the worst tenth of months), and no holder in the last year lost half.
- **SILVER: 2x.** The risk grade says no leverage at all is safe here, and we sell it anyway because people want it — far too volatile to lever at all: it swings 62% a year. A typical 30-day hold at 2x over the last year has made 4%, paid 0.2% to the leverage itself (9.5% in the worst tenth of months), and one in 45 holders lost half.
- **WTIOIL: 2x.** The risk grade is 1.5x, below what we sell it at. Too volatile for 2x; it swings 49% a year against the S&P's 19%. A typical 30-day hold at 2x over the last year has made 10%, paid 1.1% to the leverage itself (7.2% in the worst tenth of months), and one in 45 holders lost half.
- **TSLA: 2x.** The risk grade says no leverage at all is safe here, and we sell it anyway because people want it — far too volatile to lever at all: it swings 60% a year. A typical 30-day hold at 2x over the last year has lost 11%, paid 0.5% to the leverage itself (2.7% in the worst tenth of months), and no holder in the last year lost half.
- **NVDA: 2x.** The risk grade says no leverage at all is safe here, and we sell it anyway because people want it — it jumps too far overnight — 26% in one gap — for us to protect anyone at any leverage. A typical 30-day hold at 2x over the last year has made 2%, paid 0.5% to the leverage itself (1.9% in the worst tenth of months), and no holder in the last year lost half.
- **AAPL: 2x.** The risk grade agrees. Too volatile for 2.5x; it swings 28% a year against the S&P's 19%. A typical 30-day hold at 2x over the last year has made 7%, paid nothing to the leverage (0.7% in the worst tenth of months), and no holder in the last year lost half.
- **MSTR: 2x.** The risk grade says no leverage at all is safe here, and we sell it anyway because people want it — far too volatile to lever at all: it swings 91% a year. A typical 30-day hold at 2x over the last year has lost 30%, gained 1.3% from the leverage because the move trended (6.7% in the worst tenth of months), and one in 4 holders lost half.
- **COIN: 2x.** The risk grade says no leverage at all is safe here, and we sell it anyway because people want it — far too volatile to lever at all: it swings 86% a year. A typical 30-day hold at 2x over the last year has lost 21%, paid 3.0% to the leverage itself (9.2% in the worst tenth of months), and one in 9 holders lost half.
- **HOOD: 2x.** The risk grade says no leverage at all is safe here, and we sell it anyway because people want it — far too volatile to lever at all: it swings 72% a year. A typical 30-day hold at 2x over the last year has lost 13%, paid 1.2% to the leverage itself (5.7% in the worst tenth of months), and one in 9 holders lost half.
- **LLY: 2x.** The risk grade is 1.5x, below what we sell it at. Too volatile for 2x; it swings 36% a year against the S&P's 19%. A typical 30-day hold at 2x over the last year has made 10%, gained 0.4% from the leverage because the move trended (0.9% in the worst tenth of months), and no holder in the last year lost half.
- **MRNA: 2x.** The risk grade says no leverage at all is safe here, and we sell it anyway because people want it — far too volatile to lever at all: it swings 192% a year. A typical 30-day hold at 2x over the last year has made 17%, gained 5.3% from the leverage because the move trended (12.8% in the worst tenth of months), and no holder in the last year lost half. **It is unusually wild right now, so minting is paused.**
- **SPCX: 2x.** The risk grade says no leverage at all is safe here, and we sell it anyway because people want it — far too volatile to lever at all: it swings 82% a year. A typical 30-day hold at 2x over the last year has lost 16%, gained 2.1% from the leverage because the move trended (3.9% in the worst tenth of months), and one in 4 holders lost half.
- **MU: 2x.** The risk grade says no leverage at all is safe here, and we sell it anyway because people want it — far too volatile to lever at all: it swings 82% a year. A typical 30-day hold at 2x over the last year has made 37%, gained 6.2% from the leverage because the move trended (10.3% in the worst tenth of months), and one in 45 holders lost half.
- **SNDK: 2x.** The risk grade says no leverage at all is safe here, and we sell it anyway because people want it — far too volatile to lever at all: it swings 116% a year. A typical 30-day hold at 2x over the last year has made 103%, gained 16.4% from the leverage because the move trended (22.7% in the worst tenth of months), and one in 15 holders lost half.
- **PLTR: 2x.** The risk grade says no leverage at all is safe here, and we sell it anyway because people want it — far too volatile to lever at all: it swings 67% a year. A typical 30-day hold at 2x over the last year has lost 14%, paid 1.0% to the leverage itself (4.7% in the worst tenth of months), and one in 45 holders lost half.

### What changed since the first run, and why

The first version of this file listed one market at 3x and refused 38 of 43. That was mostly an artefact of how the budget was set, not a finding about the markets. Three things were wrong with it.

**It measured the S&P over the wrong year.** The budget for how much a leveraged product is allowed to cost was calibrated on the S&P's last twelve months, which were unusually calm — 12.8% against 19.3% over the full twenty years — and the calm was concentrated in a handful of large AI names while the rest of the market was choppier than the index suggested. Using one quiet year as the yardstick made everything else look reckless by comparison. The budget is now set on the full twenty-year history, which also matches what the risk sheet already measures for US500 at 3x: about **0.21% a week**.

**It judged each market on one year too.** A quiet twelve months could flatter a name that had been wild for four; a wild twelve months could condemn one that is normally calm. Each market is now graded on whichever is worse, its one-year or its five-year volatility, so neither a lucky year nor an unlucky one decides the answer on its own.

**It assumed we post the same collateral at every leverage.** We do not have to. Posting more collateral at lower leverage is what makes a lower tier actually safer against an overnight jump — under the old flat assumption, dropping a market from 3x to 1.5x bought no extra protection at all, which is plainly the wrong shape for a leverage ladder. The rule now assumes collateral scales with the tier, which is a proposed change to D21 set out in full below.

The result is a ladder that spreads across the rungs instead of refusing almost everything: 1 at 3x, 1 at 2.5x, 5 at 2x, 10 at 1.5x, and 26 not listed. The markets still refused are refused for reasons none of these choices moves: they swing 44% to 192% a year, or they jump too far overnight for any rung to absorb, or Phoenix itself will not carry them.

One thing did **not** change, and is worth saying plainly. Every market the registry ships that this rule refuses outright — `COIN`, `HOOD`, `MRNA`, `MSTR`, `MU`, `NVDA`, `PLTR`, `SILVER`, `SNDK`, `SPCX`, `TSLA` — was refused on the old calibration too. The first run's overall strictness was an artefact; these particular refusals are not, and loosening the budget further will not reach them.

## Headline

- Of **43** Phoenix equity and commodity markets, the rule lists **1** at 3x, **1** at 2.5x, **5** at 2x, **10** at 1.5x, and refuses **26**.
- **40 refusals and downgrades are the decay budget**, 2 are the gap cap, 0 are Phoenix's own ceiling and 0 are too little history. The binding constraint on this book is the cost of rebalancing a volatile underlying, not the risk of a gap jumping the keeper.
- The budget is calibrated on the full ~20-year file of US500, σ 19.3%, giving 11.12%/yr. On the trailing year alone it would be 4.95% — see *How sensitive is this to the reference window?* for what that does.
- Markets still refused are refused for reasons no calibration choice moves: their own σ runs to 192%, or Phoenix itself caps them.
- **27 markets could not reach 3x on the venue cap alone**, whatever their statistics said: Phoenix gives them 10x and a fifth of that is the ceiling.
- **1 market clears 3x on the rule**: `SPY`.
- **D26 lists 17 markets**, `QQQ` and `SPY` at 3x and the rest at 2x. **11 of them are graded unlistable by the rule** — `COIN`, `HOOD`, `MRNA`, `MSTR`, `MU`, `NVDA`, `PLTR`, `SILVER`, `SNDK`, `SPCX`, `TSLA` — and 3 are listed above grade. That is the D26 bet: disclosure rather than refusal.
- Cap 2 assumes a **proposed amendment to D21** — collateral scaling with the tier — without which the lower rungs are no safer against a gap than 2.5x. See *Proposed D21 amendment* under the rule.

## The rule

A market is listed at the largest rung of the ladder **3x / 2.5x / 2x / 1.5x** that satisfies all four caps. A market that fails every rung is not listed.

**1. Decay budget.** The structural cost of holding constant leverage is `N(N−1)σ²/2` per year. A listing may not cost more than **US500 at 3x** already costs, because that is the one leveraged product the risk sheet has argued is acceptable. Measured over **the full ~20-year file** (`--budget-window 20y`), US500's σ is **19.3%**, so the budget is **11.12%/yr — 0.21%/wk**.

That closed form is worth one check against the simulator: `docs/RISK_SHEET.md` §3 measures uUS500 at 3x with an underlying vol of 19% and a *volatility drag* of 11.5%/yr, against the 11.12% this formula gives. The approximation is good to under half a point on the market it is calibrated on, which is why a closed form is usable as a published rule at all.

**σ for grading a market is the larger of its trailing 1-year and 5-year realised vol.** One window alone is wrong in both directions: a quiet year flatters a name that has been wild for four, and a five-year average buries one that has just come apart. The 60-day figure never enters the rule — it drives the mint-pause flag only. In vol terms each rung tolerates:

| Rung | `N(N−1)/2` | Max σ | Markets awarded |
|---|---|---|---|
| 3x | 3.000 | 19.3% | 1 |
| 2.5x | 1.875 | 24.4% | 1 |
| 2x | 1.000 | 33.4% | 5 |
| 1.5x | 0.375 | 54.5% | 10 |
| not listed | — | above 54.5% | 26 |

**2. Gap.** The keeper cannot act inside an unattended gap, so a single gap must not travel far enough to do something the keeper cannot undo. That distance is the nearer of two lines: the **hard-delever** line, past which the position sits outside the band with no trade in between, and **liquidation**, past which there is no position. The worst close→open gap the market has printed in five years must clear it with a margin of **1.25×**:

| Rung | Target → hard delever | Hard-delever distance | Liquidation from margin target | **Unattended distance** | Largest 5y gap it admits |
|---|---|---|---|---|---|
| 3x | 3.0x → 4.00x | 11.11% | 14.51% | **11.11%** | 8.89% |
| 2.5x | 2.5x → 3.30x | 13.91% | 17.10% | **13.91%** | 11.13% |
| 2x | 2.0x → 2.60x | 18.75% | 22.28% | **18.75%** | 15.00% |
| 1.5x | 1.5x → 1.95x | 31.58% | 30.91% | **30.91%** | 24.73% |

The liquidation column assumes **no top-up on the way**, which is the right assumption inside a gap: D21's floor-and-ceiling mechanism refills margin from idle, but it needs a block to do it in and a gap does not provide one.

Both directions count: a market is listed as a u- and a d-token, so the gap tested is the worst in *either* direction, not the worst adverse one for a long.

### Proposed D21 amendment: collateral scales with the tier

**This cap assumes a change to D21 that has not been adopted.** D21 as written posts 17.5% of notional at every leverage. Because margin is a share of *notional*, that makes the distance from the margin target to liquidation a constant 14.51% whatever the tier, while the distance to the hard-delever line widens as leverage falls. The consequence is perverse: at 2x and below it is the posted margin, not the leverage band, that an unattended gap runs into, so **lowering a market's leverage buys it no gap protection at all** — which defeats the point of having a ladder.

The amendment posts more collateral at lower leverage, so the two protections widen together:

| Rung | Margin target (proposed) | × N = share of NAV | Liquidation from target | Unattended distance | On D21 as written |
|---|---|---|---|---|---|
| 3x | 17.5% *(D21 unchanged)* | 52.5% | 14.51% | **11.11%** | 11.11% |
| 2.5x | 20.0% | 50.0% | 17.10% | **13.91%** | 13.91% |
| 2x | 25.0% | 50.0% | 22.28% | **18.75%** | 14.51% |
| 1.5x | 33.3% | 50.0% | 30.91% | **30.91%** | 14.51% |

The 3x row is D21 exactly as it stands, so **nothing shipping today changes**. The 2.5x rung is interpolated on the same `0.5 / N` rule the 2x and 1.5x rows follow; only 3x, 2x and 1.5x were specified. Floor and ceiling keep D21's proportions to the target (floor = target × 0.714, ceiling = target × 1.286), so the top-up and withdraw bands stay the same fraction of the target they are at 3x.

With it, the largest gap the rule admits runs 8.9% → 11.1% → 15.0% → 24.7% down the ladder; on flat collateral it runs 8.9% → 11.1% → 11.6% → 11.6% and stops moving. The cost is capital efficiency: a 1.5x listing ties up 33.3% of notional at the venue instead of 17.5%, which is 50% of NAV rather than 26%. That is the trade to decide in `docs/DECISIONS.md`.

Everything else in this file is graded **on the amendment**. The holding-period simulations below run it too. If it is rejected, re-read the last column above: the ladder collapses to "2x and 1.5x are the same product" and several markets graded 1.5x here would have to be refused outright.


**3. Venue.** `N ≤ Phoenix maxLeverage / 5`. Phoenix sets maintenance margin as a fraction of `1/maxLeverage`, so a fifth of the venue ceiling leaves the keeper the same proportional room on a 10x market as on a 25x one. This is the cap that puts every 10x name at 2x whatever its vol says.

**4. History.** A vol estimate from a few months of prints is not a vol estimate. Under one year of data caps the rung at 2x; under six months, at 1.5x. Nothing is refused a listing for being young — the other three caps still have to pass on the data that exists.

### What the rule does not do

- It does not use the simulator. The decay term is the closed-form continuous-rebalancing approximation, not measured drag, because a rule has to be checkable by hand from two published numbers. `docs/RISK_SHEET.md` remains the measured number of record, and the *Holding-period outcomes* table below is the simulator's answer for the same markets.
- It does not consider liquidity. Phoenix's OI cap and book depth are listed in *Venue parameters* for context but no cap reads them; sizing the vault against depth is `docs/RISK_SHEET.md` §1 and PRD §4.1, a separate question from what leverage the product carries.
- It does not look at the business. Whether a market is worth listing at all — demand, licensing, the name on the ticker — is not a risk question and is not decided here.

## Grades

| Market | Name | Class | **Listed (D26)** | **Grade** | Bound by | **σ graded** | σ 1y | σ 5y | Regime | Decay at grade | Worst gap 5y | Unattended distance | Venue max | Decay cap | Gap cap | Venue cap | Hist cap |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| `SPY` | US 500 | index | **3x** | **3x** | none | **17%** | 13% | 17% | normal | 8.9% | 4.0% | 11.1% | 20x | 3x | 3x | 3x | 3x |
| `QQQ` | US Tech 100 | index | **3x** | **2.5x** | decay | **23%** | 20% | 23% | normal | 9.9% | 5.4% | 13.9% | 20x | 2.5x | 3x | 3x | 3x |
| `AAPL` | Apple | equity | **2x** | **2x** | decay | **28%** | 25% | 28% | watch | 7.9% | 9.4% | 18.8% | 20x | 2x | 2.5x | 3x | 3x |
| `COPPER` | Copper | commodity | **not listed** | **2x** | decay | **29%** | 29% | 27% | normal | 8.3% | 9.7% | 18.8% | 20x | 2x | 2.5x | 3x | 3x |
| `GOLD` | Gold | commodity | **2x** | **2x** | decay | **29%** | 29% | 19% | normal | 8.5% | 6.0% | 18.8% | 25x | 2x | 3x | 3x | 3x |
| `GOOGL` | Google | equity | **not listed** | **2x** | decay | **32%** | 32% | 32% | normal | 10.3% | 11.8% | 18.8% | 20x | 2x | 2x | 3x | 3x |
| `MSFT` | Microsoft | equity | **not listed** | **2x** | decay | **32%** | 32% | 28% | watch | 10.5% | 12.1% | 18.8% | 20x | 2x | 2x | 3x | 3x |
| `LLY` | Eli Lilly and Company | equity | **2x** | **1.5x** | decay | **36%** | 36% | 33% | normal | 4.8% | 14.4% | 29.8% | 10x | 1.5x | 2x | 2x | 3x |
| `AMZN` | Amazon | equity | **not listed** | **1.5x** | decay | **36%** | 34% | 36% | watch | 5.0% | 12.5% | 30.9% | 20x | 1.5x | 2x | 3x | 3x |
| `TSM` | TSMC | equity | **not listed** | **1.5x** | decay | **40%** | 40% | 38% | normal | 6.0% | 12.5% | 29.8% | 10x | 1.5x | 2x | 2x | 3x |
| `META` | Meta | equity | **not listed** | **1.5x** | decay | **45%** | 39% | 45% | normal | 7.6% | 24.5% | 30.9% | 20x | 1.5x | 1.5x | 3x | 3x |
| `AVGO` | Broadcom | equity | **not listed** | **1.5x** | decay | **46%** | 46% | 44% | normal | 8.0% | 18.4% | 29.8% | 10x | 1.5x | 1.5x | 2x | 3x |
| `ASML` | ASML | equity | **not listed** | **1.5x** | decay | **46%** | 46% | 43% | normal | 8.1% | 8.8% | 29.8% | 10x | 1.5x | 3x | 2x | 3x |
| `WTIOIL` | West Texas Intermediate | commodity | **2x** | **1.5x** | decay | **49%** | 49% | 37% | normal | 8.9% | 13.8% | 30.9% | 20x | 1.5x | 2x | 3x | 3x |
| `BABA` | Alibaba | equity | **not listed** | **1.5x** | decay | **52%** | 43% | 52% | normal | 10.0% | 19.9% | 29.8% | 10x | 1.5x | 1.5x | 2x | 3x |
| `QCOM` | Qualcomm | equity | **not listed** | **1.5x** | decay | **53%** | 53% | 42% | normal | 10.4% | 12.9% | 29.8% | 10x | 1.5x | 2x | 2x | 3x |
| `CRWD` | CrowdStrike Holdings | equity | **not listed** | **1.5x** | decay | **54%** | 54% | 52% | watch | 10.9% | 21.0% | 29.8% | 10x | 1.5x | 1.5x | 2x | 3x |
| `NFLX` | Netflix | equity | **not listed** | **—** | gap | **44%** | 36% | 44% | normal | — | 29.7% | — | 10x | 1.5x | — | 2x | 3x |
| `NVDA` | NVIDIA | equity | **2x** | **—** | gap | **52%** | 38% | 52% | normal | — | 26.1% | — | 20x | 1.5x | — | 3x | 3x |
| `ORCL` | Oracle Corporation | equity | **not listed** | **—** | decay | **57%** | 57% | 44% | normal | — | 32.2% | — | 10x | — | — | 2x | 3x |
| `AMAT` | Applied Materials | equity | **not listed** | **—** | decay | **60%** | 60% | 47% | watch | — | 13.6% | — | 10x | — | 2x | 2x | 3x |
| `TSLA` | Tesla | equity | **2x** | **—** | decay | **60%** | 47% | 60% | normal | — | 14.5% | — | 20x | — | 2x | 3x | 3x |
| `SILVER` | Silver | commodity | **2x** | **—** | decay | **62%** | 62% | 37% | calm | — | 15.4% | — | 25x | — | 1.5x | 3x | 3x |
| `PLTR` | Palantir | equity | **2x** | **—** | decay | **67%** | 61% | 67% | watch | — | 22.8% | — | 10x | — | 1.5x | 2x | 3x |
| `NET` | Cloudflare | equity | **not listed** | **—** | decay | **69%** | 64% | 69% | normal | — | 25.5% | — | 10x | — | — | 2x | 3x |
| `AMD` | AMD | equity | **not listed** | **—** | decay | **72%** | 72% | 57% | normal | — | 37.5% | — | 10x | — | — | 2x | 3x |
| `HOOD` | Robinhood | equity | **2x** | **—** | decay | **72%** | 72% | 68% | normal | — | 23.1% | — | 10x | — | 1.5x | 2x | 3x |
| `DELL` | Dell Technologies | equity | **not listed** | **—** | decay | **75%** | 75% | 54% | normal | — | 31.8% | — | 10x | — | — | 2x | 3x |
| `ARM` | ARM | equity | **not listed** | **—** | decay | **77%** | 76% | 77% | normal | — | 22.6% | — | 10x | — | 1.5x | 2x | 3x |
| `MRVL` | Marvell Technology, Inc. | equity | **not listed** | **—** | decay | **79%** | 79% | 65% | normal | — | 20.9% | — | 10x | — | 1.5x | 2x | 3x |
| `INTC` | Intel | equity | **not listed** | **—** | decay | **80%** | 80% | 55% | normal | — | 27.6% | — | 10x | — | — | 2x | 3x |
| `MU` | Micron | equity | **2x** | **—** | decay | **82%** | 82% | 57% | normal | — | 18.1% | — | 15x | — | 1.5x | 3x | 3x |
| `SPCX` | SpaceX | equity | **2x** | **—** | decay | **82%** | 82% | 82% | normal | — | 10.3% | — | 15x | — | 2.5x | 3x | 1.5x |
| `COIN` | Coinbase | equity | **2x** | **—** | decay | **86%** | 71% | 86% | normal | — | 31.3% | — | 10x | — | — | 2x | 3x |
| `MSTR` | Strategy | equity | **2x** | **—** | decay | **91%** | 80% | 91% | normal | — | 27.4% | — | 10x | — | — | 2x | 3x |
| `CRCL` | Circle Internet Group | equity | **not listed** | **—** | decay | **108%** | 99% | 108% | normal | — | 23.2% | — | 10x | — | 1.5x | 2x | 3x |
| `NBIS` | Nebius | equity | **not listed** | **—** | decay | **109%** | 109% | 80% | watch | — | 51.7% | — | 10x | — | — | 2x | 3x |
| `CRWV` | CoreWeave | equity | **not listed** | **—** | decay | **113%** | 95% | 113% | normal | — | 20.3% | — | 10x | — | 1.5x | 2x | 3x |
| `SNDK` | Sandisk | equity | **2x** | **—** | decay | **116%** | 116% | 106% | normal | — | 20.8% | — | 15x | — | 1.5x | 3x | 3x |
| `IREN` | IREN Limited | equity | **not listed** | **—** | decay | **118%** | 110% | 118% | normal | — | 27.5% | — | 10x | — | — | 2x | 3x |
| `SKHY` | SK Hynix | equity | **not listed** | **—** | decay | **119%** | 119% | 119% | normal | — | 10.3% | — | 10x | — | 2.5x | 2x | 1.5x |
| `CBRS` | Cerebras | equity | **not listed** | **—** | decay | **124%** | 124% | 124% | normal | — | 16.0% | — | 10x | — | 1.5x | 2x | 1.5x |
| `MRNA` | Moderna | equity | **2x** | **—** | decay | **192%** | 192% | 104% | pause | — | 84.3% | — | 10x | — | — | 2x | 3x |

Sorted by grade, then by σ within a grade. **Listed (D26)** is what we actually sell the token at and **Grade** is what this rule says about the risk; since D26 those are separate decisions. **σ graded** is the larger of the 1-year and 5-year columns beside it, and it is the only one the rule reads. *Bound by* names the cap that produced the tier; the last four columns show what each cap allowed on its own, so a row bound by `venue` with a `decay cap` of 3x is a market whose volatility would carry more leverage than Phoenix will give us room for.

## Vol regime now vs 1 year

A tier is set on the trailing-year vol and should not move every time a name gets noisy. But minting into a name whose 60-day vol has run far above its own annual number adds exposure at the worst moment, so the grader flags the regime instead of changing the tier: **pause** at σ60 ≥ 1.5 × σ1y, **watch** at ≥ 1.25×.

Currently flagged **pause mint**: `MRNA` (375% vs 192%, 1.95×).

| Market | σ 60d | σ 1y | σ60 / σ1y | Regime | Tier | Flags |
|---|---|---|---|---|---|---|
| `MRNA` | 375% | 192% | 1.95 | pause | — | **vol spike — pause mint**; one print (2026-08-19) drives σ |
| `MSFT` | 44% | 32% | 1.35 | watch | 2x | vol elevated |
| `PLTR` | 82% | 61% | 1.35 | watch | — | vol elevated |
| `AMAT` | 79% | 60% | 1.33 | watch | — | vol elevated |
| `NBIS` | 141% | 109% | 1.30 | watch | — | vol elevated |
| `AMZN` | 44% | 34% | 1.29 | watch | 1.5x | vol elevated |
| `AAPL` | 32% | 25% | 1.29 | watch | 2x | vol elevated |
| `CRWD` | 68% | 54% | 1.27 | watch | 1.5x | vol elevated |
| `TSLA` | 57% | 47% | 1.21 | normal | — | — |
| `META` | 47% | 39% | 1.19 | normal | 1.5x | — |
| `SNDK` | 136% | 116% | 1.17 | normal | — | — |
| `GOOGL` | 36% | 32% | 1.16 | normal | 2x | — |
| `MU` | 93% | 82% | 1.14 | normal | — | — |
| `CRWV` | 107% | 95% | 1.12 | normal | — | — |
| `MRVL` | 88% | 79% | 1.12 | normal | — | — |
| `MSTR` | 89% | 80% | 1.11 | normal | — | — |
| `QQQ` | 22% | 20% | 1.11 | normal | 2.5x | — |
| `TSM` | 44% | 40% | 1.10 | normal | 1.5x | — |
| `DELL` | 82% | 75% | 1.10 | normal | — | — |
| `IREN` | 118% | 110% | 1.07 | normal | — | — |
| `NVDA` | 40% | 38% | 1.06 | normal | — | — |
| `WTIOIL` | 52% | 49% | 1.06 | normal | 1.5x | ETF proxy |
| `ASML` | 49% | 46% | 1.05 | normal | 1.5x | — |
| `NFLX` | 38% | 36% | 1.05 | normal | — | — |
| `ARM` | 79% | 76% | 1.04 | normal | — | — |
| `HOOD` | 75% | 72% | 1.04 | normal | — | — |
| `BABA` | 45% | 43% | 1.03 | normal | 1.5x | — |
| `CBRS` | 126% | 124% | 1.02 | normal | — | **< 6 months of data** |
| `COIN` | 72% | 71% | 1.01 | normal | — | — |
| `SPCX` | 82% | 82% | 1.00 | normal | — | **< 6 months of data** |
| `SKHY` | 119% | 119% | 1.00 | normal | — | **< 6 months of data** |
| `ORCL` | 56% | 57% | 0.98 | normal | — | — |
| `AMD` | 69% | 72% | 0.96 | normal | — | — |
| `INTC` | 76% | 80% | 0.95 | normal | — | — |
| `LLY` | 34% | 36% | 0.94 | normal | 1.5x | — |
| `SPY` | 12% | 13% | 0.92 | normal | 3x | — |
| `AVGO` | 42% | 46% | 0.92 | normal | 1.5x | — |
| `CRCL` | 90% | 99% | 0.91 | normal | — | — |
| `QCOM` | 47% | 53% | 0.90 | normal | 1.5x | — |
| `NET` | 56% | 64% | 0.89 | normal | — | — |
| `GOLD` | 25% | 29% | 0.86 | normal | 2x | ETF proxy |
| `COPPER` | 24% | 29% | 0.84 | normal | 2x | ETF proxy |
| `SILVER` | 42% | 62% | 0.67 | calm | — | ETF proxy |

## Risk inputs

Realised vol is the annualised stdev of daily close-to-close returns over the trailing window. Worst day and worst gap are the largest |move| in *either* direction. The last two columns count days past the hard-delever distance at 3x (11.11%) and at 2x (18.75%) — a day past that line is a day the keeper had to make a forced trade, and a *gap* past it is a day it could not.

| Market | Bars | From | σ 60d | σ 1y | σ 5y | σ all | Worst day 1y | Worst day 5y | Worst gap 1y | Worst gap 5y | On | Days >11.1%/yr | Days >18.8%/yr |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| `SPY` | 5205 | 2006-01-03 | 12% | 13% | 17% | 19% | 2.9% | 10.5% | 2.6% | 4.0% | 2024-08-05 | 0.10 | 0.00 |
| `QQQ` | 5205 | 2006-01-03 | 22% | 20% | 23% | 22% | 4.8% | 12.0% | 3.4% | 5.4% | 2024-08-05 | 0.15 | 0.00 |
| `AAPL` | 5205 | 2006-01-03 | 32% | 25% | 28% | 32% | 7.4% | 15.3% | 8.6% | 9.4% | 2024-08-05 | 0.39 | 0.00 |
| `COPPER` | 3397 | 2011-11-16 | 24% | 29% | 27% | 25% | 5.9% | 19.3% | 9.7% | 9.7% | 2026-01-29 | 0.07 | 0.07 |
| `GOLD` | 5205 | 2006-01-03 | 25% | 29% | 19% | 18% | 10.3% | 10.3% | 6.0% | 6.0% | 2026-01-30 | 0.05 | 0.00 |
| `GOOGL` | 5205 | 2006-01-03 | 36% | 32% | 32% | 30% | 10.0% | 10.2% | 6.9% | 11.8% | 2024-04-26 | 0.44 | 0.05 |
| `MSFT` | 5205 | 2006-01-03 | 44% | 32% | 28% | 28% | 15.5% | 15.5% | 12.1% | 12.1% | 2026-07-30 | 0.39 | 0.00 |
| `LLY` | 5205 | 2006-01-03 | 34% | 36% | 33% | 27% | 10.3% | 14.9% | 7.1% | 14.4% | 2025-04-17 | 0.44 | 0.00 |
| `AMZN` | 5205 | 2006-01-03 | 44% | 34% | 36% | 38% | 15.3% | 15.3% | 12.5% | 12.5% | 2026-07-31 | 1.07 | 0.19 |
| `TSM` | 5205 | 2006-01-03 | 44% | 40% | 38% | 34% | 7.9% | 13.3% | 7.2% | 12.5% | 2022-11-15 | 0.34 | 0.05 |
| `META` | 3598 | 2012-05-21 | 47% | 39% | 45% | 40% | 11.3% | 26.4% | 11.0% | 24.5% | 2022-10-27 | 1.19 | 0.49 |
| `AVGO` | 4300 | 2009-08-07 | 42% | 46% | 44% | 38% | 12.6% | 24.4% | 14.7% | 18.4% | 2024-12-13 | 0.94 | 0.12 |
| `ASML` | 5204 | 2006-01-03 | 49% | 46% | 43% | 37% | 9.5% | 16.3% | 8.8% | 8.8% | 2026-04-08 | 0.44 | 0.00 |
| `WTIOIL` | 5137 | 2006-04-11 | 52% | 49% | 37% | 37% | 12.9% | 12.9% | 13.8% | 13.8% | 2026-04-08 | 0.64 | 0.10 |
| `BABA` | 3011 | 2014-09-22 | 45% | 43% | 52% | 42% | 11.0% | 36.8% | 9.6% | 19.9% | 2022-03-16 | 1.00 | 0.08 |
| `QCOM` | 5205 | 2006-01-03 | 47% | 53% | 42% | 36% | 15.1% | 15.2% | 11.2% | 12.9% | 2021-11-04 | 0.97 | 0.05 |
| `CRWD` | 1822 | 2019-06-13 | 68% | 54% | 52% | 56% | 20.5% | 20.5% | 10.1% | 21.0% | 2024-03-06 | 3.18 | 0.14 |
| `NFLX` | 5205 | 2006-01-03 | 38% | 36% | 44% | 49% | 13.8% | 35.1% | 11.9% | 29.7% | 2022-04-20 | 2.42 | 0.63 |
| `NVDA` | 5205 | 2006-01-03 | 40% | 38% | 52% | 49% | 8.7% | 24.4% | 6.3% | 26.1% | 2023-05-25 | 1.69 | 0.19 |
| `ORCL` | 5205 | 2006-01-03 | 56% | 57% | 44% | 32% | 12.7% | 35.9% | 14.5% | 32.2% | 2025-09-10 | 0.77 | 0.10 |
| `AMAT` | 5205 | 2006-01-03 | 79% | 60% | 47% | 39% | 15.0% | 16.1% | 11.1% | 13.6% | 2025-08-15 | 0.73 | 0.05 |
| `TSLA` | 4075 | 2010-06-30 | 57% | 47% | 60% | 57% | 14.5% | 22.7% | 8.8% | 14.5% | 2024-10-24 | 4.02 | 0.49 |
| `SILVER` | 5124 | 2006-05-01 | 42% | 62% | 37% | 33% | 28.5% | 28.5% | 15.4% | 15.4% | 2026-01-30 | 0.59 | 0.05 |
| `PLTR` | 1493 | 2020-10-01 | 82% | 61% | 67% | 70% | 29.5% | 30.8% | 15.5% | 22.8% | 2025-02-04 | 6.08 | 2.03 |
| `NET` | 1757 | 2019-09-16 | 56% | 64% | 69% | 67% | 23.6% | 27.1% | 15.4% | 25.5% | 2023-04-28 | 5.88 | 0.86 |
| `AMD` | 5205 | 2006-01-03 | 69% | 72% | 57% | 59% | 23.7% | 23.8% | 37.5% | 37.5% | 2025-10-06 | 4.16 | 0.44 |
| `HOOD` | 1285 | 2021-07-30 | 75% | 72% | 68% | 74% | 16.6% | 24.9% | 11.9% | 23.1% | 2022-05-13 | 6.28 | 1.77 |
| `DELL` | 1939 | 2018-12-24 | 82% | 75% | 54% | 49% | 32.8% | 32.8% | 31.8% | 31.8% | 2026-05-29 | 2.34 | 0.65 |
| `ARM` | 750 | 2023-09-15 | 79% | 76% | 77% | 77% | 17.4% | 47.9% | 14.7% | 22.6% | 2024-02-08 | 7.07 | 1.35 |
| `MRVL` | 5205 | 2006-01-03 | 88% | 79% | 65% | 49% | 32.5% | 32.5% | 15.5% | 20.9% | 2021-12-03 | 1.99 | 0.39 |
| `INTC` | 5205 | 2006-01-03 | 76% | 80% | 55% | 38% | 23.6% | 26.1% | 27.6% | 27.6% | 2025-09-18 | 1.31 | 0.24 |
| `MU` | 5205 | 2006-01-03 | 93% | 82% | 57% | 54% | 19.3% | 19.3% | 17.6% | 18.1% | 2024-09-26 | 3.00 | 0.24 |
| `SPCX` | 62 | 2026-06-15 | 82% | 82% | 82% | 82% | 16.4% | 16.4% | 10.3% | 10.3% | 2026-08-05 | 12.39 | 0.00 |
| `COIN` | 1359 | 2021-04-15 | 72% | 71% | 86% | 85% | 16.5% | 31.1% | 8.9% | 31.3% | 2022-08-04 | 11.32 | 1.67 |
| `MSTR` | 5205 | 2006-01-03 | 89% | 80% | 91% | 60% | 26.1% | 26.1% | 8.6% | 27.4% | 2024-08-05 | 5.23 | 1.07 |
| `CRCL` | 318 | 2025-06-06 | 90% | 99% | 108% | 108% | 35.5% | 35.5% | 20.2% | 23.2% | 2025-06-09 | 18.28 | 5.56 |
| `NBIS` | 3182 | 2011-05-25 | 141% | 109% | 80% | 62% | 34.1% | 49.4% | 17.0% | 51.7% | 2025-09-09 | 4.99 | 0.55 |
| `CRWV` | 365 | 2025-03-31 | 107% | 95% | 113% | 113% | 22.6% | 41.8% | 20.3% | 20.3% | 2026-08-12 | 23.54 | 7.62 |
| `SNDK` | 395 | 2025-02-14 | 136% | 116% | 106% | 106% | 27.6% | 28.6% | 20.8% | 20.8% | 2026-01-30 | 27.50 | 4.48 |
| `IREN` | 1207 | 2021-11-18 | 118% | 110% | 118% | 118% | 30.5% | 67.0% | 20.3% | 27.5% | 2024-08-05 | 26.75 | 5.22 |
| `SKHY` | 44 | 2026-07-13 | 119% | 119% | 119% | 119% | 27.3% | 27.3% | 10.3% | 10.3% | 2026-07-14 | 23.44 | 5.86 |
| `CBRS` | 82 | 2026-05-15 | 126% | 124% | 124% | 124% | 19.9% | 19.9% | 16.0% | 16.0% | 2026-08-13 | 31.11 | 9.33 |
| `MRNA` | 1949 | 2018-12-10 | 375% | 192% | 104% | 97% | 177.0% | 177.0% | 84.3% | 84.3% | 2026-08-19 | 7.24 | 1.29 |

## Holding-period outcomes

What the simulator says a holder actually got, at the keeper cadence `DEFAULT_PATH_OPTIONS` sets (about every 15 minutes). Rolling 30- and 90-trading-day NAV returns of the long token, pooled across the independent one-year segments the risk sheet uses, all costs on. `loss>50%` is the share of windows in which the holder lost more than half.

These are outcomes, not inputs: the rule above never reads them. They are here so the tier can be sanity-checked against what it implies for someone who buys and sits. A market whose history is shorter than one window shows `—` rather than a zero.

Read the medians together with the `loss>50%` shares, not instead of them. A median is a statement about the middle of a rising sample: `SNDK` shows a +64% median 30-day return at 3x *and* a 21% chance of losing more than half, because it is a young name that has mostly gone up very fast. The pairing is the point — the rule refuses both halves of that trade.

| Market | 2x 30d med | 2x 30d loss>50% | 2x 90d med | 2x 90d loss>50% | 3x 30d med | 3x 30d loss>50% | 3x 90d med | 3x 90d loss>50% |
|---|---|---|---|---|---|---|---|---|
| `SPY` | +3.4% | 0.3% | +8.6% | 1.8% | +4.9% | 1.2% | +12.3% | 2.1% |
| `QQQ` | +4.5% | 0.3% | +13.3% | 1.9% | +6.5% | 1.3% | +18.3% | 2.8% |
| `AAPL` | +6.7% | 0.8% | +24.9% | 2.5% | +8.9% | 2.7% | +32.9% | 6.2% |
| `COPPER` | +1.1% | 0.0% | -1.2% | 0.3% | +0.4% | 1.0% | -5.1% | 4.4% |
| `GOLD` | +0.6% | 0.0% | +3.5% | 0.0% | +0.4% | 0.1% | +4.0% | 2.5% |
| `GOOGL` | +4.4% | 0.2% | +17.5% | 2.2% | +5.4% | 2.5% | +21.9% | 6.4% |
| `MSFT` | +4.3% | 0.0% | +14.1% | 0.7% | +5.4% | 1.2% | +17.4% | 3.3% |
| `LLY` | +2.2% | 0.5% | +7.9% | 1.4% | +1.7% | 1.2% | +8.4% | 2.7% |
| `AMZN` | +6.1% | 1.2% | +17.9% | 3.8% | +7.2% | 3.7% | +20.7% | 8.0% |
| `TSM` | +4.2% | 0.0% | +5.9% | 3.0% | +4.3% | 2.1% | +3.0% | 5.7% |
| `META` | +5.8% | 2.4% | +10.7% | 8.2% | +5.9% | 4.4% | +10.0% | 13.2% |
| `AVGO` | +7.8% | 0.5% | +24.7% | 0.5% | +9.2% | 2.0% | +27.8% | 3.7% |
| `ASML` | +4.0% | 0.4% | +14.6% | 4.2% | +3.4% | 2.9% | +13.8% | 11.1% |
| `WTIOIL` | +0.0% | 3.0% | -4.1% | 11.6% | -2.1% | 6.7% | -11.1% | 21.0% |
| `BABA` | -1.4% | 0.9% | -7.1% | 6.0% | -5.1% | 7.7% | -16.5% | 19.8% |
| `QCOM` | +1.2% | 0.7% | -1.6% | 3.3% | -0.0% | 3.9% | -7.6% | 10.3% |
| `CRWD` | +6.4% | 6.5% | +11.2% | 17.4% | +3.1% | 15.0% | -4.3% | 34.0% |
| `NFLX` | +5.8% | 3.2% | +15.7% | 9.9% | +5.2% | 8.1% | +10.1% | 20.5% |
| `NVDA` | +7.5% | 3.4% | +21.2% | 12.0% | +6.9% | 9.2% | +16.7% | 18.7% |
| `ORCL` | +3.7% | 0.7% | +8.2% | 1.1% | +4.3% | 2.8% | +9.1% | 5.2% |
| `AMAT` | +1.5% | 1.2% | +9.6% | 5.9% | -0.4% | 4.2% | +5.7% | 17.7% |
| `TSLA` | +1.1% | 3.9% | +0.2% | 10.3% | -4.0% | 11.9% | -19.7% | 26.8% |
| `SILVER` | -0.2% | 0.8% | +2.1% | 3.1% | -2.2% | 3.4% | -1.3% | 8.0% |
| `PLTR` | +0.1% | 5.8% | +13.9% | 19.9% | -11.3% | 17.5% | -11.3% | 38.8% |
| `NET` | +9.9% | 5.8% | +14.7% | 14.3% | +2.4% | 14.6% | -14.2% | 30.0% |
| `AMD` | -1.5% | 5.3% | -7.5% | 17.8% | -7.4% | 15.5% | -25.3% | 34.0% |
| `HOOD` | -4.8% | 10.7% | -6.8% | 31.6% | -9.8% | 20.5% | -29.1% | 40.8% |
| `DELL` | +10.3% | 2.1% | +20.8% | 2.7% | +12.4% | 6.9% | +22.3% | 19.4% |
| `ARM` | +2.0% | 8.2% | +35.3% | 8.3% | -5.5% | 23.9% | +6.4% | 29.6% |
| `MRVL` | -0.3% | 2.8% | +0.7% | 14.4% | -5.1% | 7.9% | -9.9% | 27.4% |
| `INTC` | +1.7% | 1.3% | +0.8% | 7.1% | +0.3% | 5.1% | -4.3% | 16.6% |
| `MU` | +0.5% | 2.8% | +0.3% | 16.2% | -4.3% | 12.3% | -13.3% | 30.2% |
| `SPCX` | -15.8% | 15.6% | — | — | -38.1% | 34.4% | — | — |
| `COIN` | -12.0% | 13.6% | -29.5% | 34.8% | -29.0% | 33.4% | -53.9% | 51.2% |
| `MSTR` | +0.2% | 6.6% | -4.9% | 15.6% | -3.1% | 12.9% | -17.2% | 29.9% |
| `CRCL` | -23.7% | 26.4% | -63.4% | 57.4% | -50.9% | 50.4% | -88.1% | 74.1% |
| `NBIS` | +1.8% | 5.8% | +4.5% | 15.1% | -2.7% | 14.6% | -8.7% | 26.7% |
| `CRWV` | -18.2% | 21.6% | -54.7% | 60.0% | -43.5% | 42.6% | -83.5% | 77.3% |
| `SNDK` | +48.3% | 12.2% | +420.8% | 6.5% | +63.7% | 20.9% | +894.3% | 16.3% |
| `IREN` | -14.0% | 28.3% | -28.0% | 38.9% | -35.5% | 43.7% | -72.6% | 63.1% |
| `SKHY` | +1.8% | 0.0% | — | — | -17.5% | 14.3% | — | — |
| `CBRS` | -33.1% | 17.3% | — | — | -59.9% | 59.6% | — | — |
| `MRNA` | -6.5% | 9.7% | -12.7% | 27.1% | -20.2% | 26.3% | -45.2% | 47.7% |

## Venue parameters

`riskFactors.maintenance` is in 1/10,000 of the initial margin requirement, and initial margin is `1/maxLeverage` — so a 5000 risk factor on a 20x market is 250 bps of notional and on a 10x market is 500 bps. The product registry instead assumes a flat 350 bps for every market (`Asset.maintenanceMarginBps`, back-solved from Phoenix's published 2x and 3x buffers). `OPEN_QUESTIONS` Q13 is still open on which is real, so the grader uses **the larger of the two per market**, which is the registry's 350 on the 20x and 25x markets and Phoenix's 500 on the 10x ones.

| Market | Phoenix maxLev | Maintenance risk factor | Phoenix maint (bps of notional) | Registry maint | Graded on | OI cap (units) | Calendar | History source | Note |
|---|---|---|---|---|---|---|---|---|---|
| `SPY` | 20x | 5000 | 250 | 350 | 350 | 3,920 | `us_equities_extended` | `e/SPY` | the Phoenix SPY market underlying |
| `QQQ` | 20x | 5000 | 250 | 350 | 350 | 4,180 | `us_equities_extended` | `e/QQQ` | the Phoenix QQQ market underlying |
| `AAPL` | 20x | 5000 | 250 | 350 | 350 | 10,320 | `us_equities_extended` | `s/AAPL` | — |
| `COPPER` | 20x | 5000 | 250 | 350 | 350 | 488,647 | `cme_commodities` | `e/CPER` | ETF proxy for copper (United States Copper Index Fund); thin and roll-based - shape only |
| `GOLD` | 25x | 5000 | 200 | 350 | 350 | 696.379 | `cme_commodities` | `e/GLD` | ETF proxy for spot gold (SPDR Gold Shares); US hours only, so its gaps bundle the whole overnight move |
| `GOOGL` | 20x | 5000 | 250 | 350 | 350 | 8,210 | `us_equities_extended` | `s/GOOGL` | — |
| `MSFT` | 20x | 5000 | 250 | 350 | 350 | 7,990 | `us_equities_extended` | `s/MSFT` | — |
| `LLY` | 10x | 5000 | 500 | 350 | 500 | 2,400 | `us_equities_extended` | `s/LLY` | — |
| `AMZN` | 20x | 5000 | 250 | 350 | 350 | 12,670 | `us_equities_extended` | `s/AMZN` | — |
| `TSM` | 10x | 5000 | 500 | 350 | 500 | 7,770 | `us_equities_extended` | `s/TSM` | — |
| `META` | 20x | 5000 | 250 | 350 | 350 | 5,260 | `us_equities_extended` | `s/META` | — |
| `AVGO` | 10x | 5000 | 500 | 350 | 500 | 7,930 | `us_equities_extended` | `s/AVGO` | — |
| `ASML` | 10x | 5000 | 500 | 350 | 500 | 1,920 | `us_equities_extended` | `s/ASML` | — |
| `WTIOIL` | 20x | 5000 | 250 | 350 | 350 | 32,633.5 | `cme_commodities` | `e/USO` | ETF proxy for WTI (United States Oil Fund); roll decay and a 1:8 reverse split Apr 2020 - shape only |
| `BABA` | 10x | 5000 | 500 | 350 | 500 | 23,560 | `us_equities_extended` | `s/BABA` | — |
| `QCOM` | 10x | 5000 | 500 | 350 | 500 | 18,160 | `us_equities_extended` | `s/QCOM` | — |
| `CRWD` | 10x | 5000 | 500 | 350 | 500 | 15,640 | `us_equities_extended` | `s/CRWD` | — |
| `NFLX` | 10x | 5000 | 500 | 350 | 500 | 36,970 | `us_equities_extended` | `s/NFLX` | — |
| `NVDA` | 20x | 5000 | 250 | 350 | 350 | 14,870 | `us_equities_extended` | `s/NVDA` | — |
| `ORCL` | 10x | 5000 | 500 | 350 | 500 | 20,750 | `us_equities_extended` | `s/ORCL` | — |
| `AMAT` | 10x | 5000 | 500 | 350 | 500 | 5,660 | `us_equities_extended` | `s/AMAT` | — |
| `TSLA` | 20x | 5000 | 250 | 350 | 350 | 7,520 | `us_equities_extended` | `s/TSLA` | — |
| `SILVER` | 25x | 5000 | 200 | 350 | 350 | 44,220.4 | `cme_commodities` | `e/SLV` | ETF proxy for spot silver (iShares Silver Trust); US hours only |
| `PLTR` | 10x | 5000 | 500 | 350 | 500 | 17,030 | `us_equities_extended` | `s/PLTR` | direct listing Sep 2020 |
| `NET` | 10x | 5000 | 500 | 350 | 500 | 10,660 | `us_equities_extended` | `s/NET` | — |
| `AMD` | 10x | 5000 | 500 | 350 | 500 | 5,730 | `us_equities_extended` | `s/AMD` | — |
| `HOOD` | 10x | 5000 | 500 | 350 | 500 | 26,230 | `us_equities_extended` | `s/HOOD` | listed Jul 2021 |
| `DELL` | 10x | 5000 | 500 | 350 | 500 | 6,760 | `us_equities_extended` | `s/DELL` | — |
| `ARM` | 10x | 5000 | 500 | 350 | 500 | 12,380 | `us_equities_extended` | `s/ARM` | relisted Sep 2023 |
| `MRVL` | 10x | 5000 | 500 | 350 | 500 | 12,260 | `us_equities_extended` | `s/MRVL` | — |
| `INTC` | 10x | 5000 | 500 | 350 | 500 | 21,820 | `us_equities_extended` | `s/INTC` | — |
| `MU` | 15x | 5000 | 333 | 350 | 350 | 5,562.97 | `us_equities_extended` | `s/MU` | — |
| `SPCX` | 15x | 5000 | 333 | 350 | 350 | 50,602.2 | `us_equities_extended` | `s/SPCX` | listed Jun 2026 |
| `COIN` | 10x | 5000 | 500 | 350 | 500 | 18,150 | `us_equities_extended` | `s/COIN` | listed Apr 2021 |
| `MSTR` | 10x | 5000 | 500 | 350 | 500 | 30,240 | `us_equities_extended` | `s/MSTR` | — |
| `CRCL` | 10x | 5000 | 500 | 350 | 500 | 45,470 | `us_equities_extended` | `s/CRCL` | listed Jun 2025 |
| `NBIS` | 10x | 5000 | 500 | 350 | 500 | 15,759.6 | `us_equities_extended` | `s/NBIS` | ticker carries Yandex N.V. history before Oct 2024; pre-2024 bars are a different company |
| `CRWV` | 10x | 5000 | 500 | 350 | 500 | 27,140 | `us_equities_extended` | `s/CRWV` | listed Mar 2025 |
| `SNDK` | 15x | 5000 | 333 | 350 | 350 | 3,832.89 | `us_equities_extended` | `s/SNDK` | spun out of Western Digital Feb 2025 |
| `IREN` | 10x | 5000 | 500 | 350 | 500 | 72,640 | `us_equities_extended` | `s/IREN` | listed Nov 2021 |
| `SKHY` | 10x | 5000 | 500 | 350 | 500 | 19,290 | `us_equities_extended` | `s/SKHY` | US-listed SK Hynix tracker, listed Jul 2026 |
| `CBRS` | 10x | 5000 | 500 | 350 | 500 | 17,160 | `us_equities_extended` | `s/CBRS` | listed May 2026 |
| `MRNA` | 10x | 5000 | 500 | 350 | 500 | 21,640 | `us_equities_extended` | `s/MRNA` | listed Dec 2018 |

## Caveats

- **3 markets have under a year of history**: `CBRS` (0.33y), `SKHY` (0.16y), `SPCX` (0.24y). Their σ is a short-window estimate annualised, and their five-year gap is really an all-history gap. The history cap holds them down but does not make the estimate good.
- **The commodity markets are graded through US-listed ETF proxies**, which trade US equity hours only. Phoenix quotes GOLD, SILVER, COPPER and WTIOIL on `cme_commodities`, many more hours per week, so the real unattended gap is *smaller* than these files show. Commodity gap numbers here are conservative. `USO` and `CPER` are futures-roll funds: use them for vol and gap shape, never for a level or a decay-versus-spot claim.
- **`NBIS` carries Yandex N.V. history before October 2024.** Its all-history σ and its 5-year gap describe a Russian internet company that was delisted and restructured; only the trailing year describes Nebius. The rule reads σ over the trailing year, so the tier is sound, but the 5-year gap feeding the gap cap is not this company's.
- **1 market has a σ driven by one print**, not by volatility: `MRNA` (192% → 76% without 2026-08-19). The rule refuses them, and it is probably the right answer — a name that can gap like that once can do it again — but it is reached by treating a corporate event as though it were ordinary variance. Do not read their σ as a forecast.
- **σ is backward-looking by construction.** Every cap here is a statement about what has happened. The regime column is the only forward-looking thing in the file and it only ever says "stop minting", never "raise the tier".

## How sensitive is this to the reference window?

The budget is `3 × 2 × σ²/2` on US500's realised vol, and which window that is measured over matters: 12.8% over the trailing year, 17.2% over five, 19.3% over the full file. The rule uses **the full ~20-year file**, on the grounds that a product ladder which re-tiers the whole book whenever one index has a quiet year is a weathervane rather than a rule. Re-running everything on each window:

| US500 σ measured over | σ | Budget /yr | 3x | 2.5x | 2x | 1.5x | not listed | bound by decay | by gap | by venue |
|---|---|---|---|---|---|---|---|---|---|---|
| trailing 1y | 12.8% | 4.95% | 0 | 0 | 1 | 7 | 35 | 43 | 0 | 0 |
| trailing 5y | 17.2% | 8.89% | 1 | 0 | 4 | 8 | 30 | 41 | 1 | 0 |
| full file, ~20y **(the rule)** | 19.3% | 11.12% | 1 | 1 | 5 | 10 | 26 | 40 | 2 | 0 |

The trailing-year row is the rule's first version, kept here as the comparison. It is the outlier, and it is the outlier in one direction: the reference year was calm and the book was not, so everything else looked reckless against it. `--budget-window 1y` reproduces it exactly.

The case for the long window is that leverage tiers are a product decision, not a trading signal. The case against is that decay is paid at today's volatility, not at a twenty-year average, and a holder in a turbulent year really does pay more than this budget assumes. That objection is answered by the mint-pause flag rather than by the tier: the regime column below stops new exposure going into a name that has come apart, while leaving the tier where the long-run evidence put it. Whether that division of labour is the right one belongs in `docs/DECISIONS.md`.

## Cost of holding, last year

The number D26 puts next to the mint button. For every listed token, at **the leverage it is actually sold at**, the measured cost of a 24-hour, 7-day and 30-day hold over the last year: one continuous simulation under the keeper cadence `DEFAULT_PATH_OPTIONS` sets, every fee and funding charge on, rolling windows, **medians not means** — the distribution is skewed enough that a mean is dominated by a handful of windows, and a holder wants to know what usually happens.

`decay` is `token return − N × stock return` per window: what the leverage itself took, over and above simply being N times as exposed. **The bold 24h / 7d / 30d columns are means**, and they are the numbers the disclosure quotes.

**Why the mean and not the median.** The median decay sits near zero and hides the cost. Trending windows compound in the token's favour and choppy ones against it, and over a year of rolling windows those roughly cancel at the midpoint — the *30d typical* column shows how close to nothing that midpoint is for most tokens. The drag is real, but it is paid as a small loss most of the time and a large one occasionally, so it lives in the average. The *worst tenth* column is the 10th percentile of the same per-window distribution: what a bad month actually cost. A disclosure that quoted only the median would be true and misleading; quoting the mean with the worst decile beside it is the honest pair.

All of these are computed per window and then aggregated. None is the difference of the two return columns beside it — those describe different windows and cannot be subtracted.

| Token | Lev | **24h** | **7d** | **30d** | 30d worst tenth | 30d typical | 30d median return | 30d stock | 30d loss>50% | Windows (1d/7d/30d) |
|---|---|---|---|---|---|---|---|---|---|---|
| `uGOLD` | 2x | -0.019% | -0.14% | -0.81% | -3.3% | -0.24% | +5.7% | +3.3% | 0.0% | 251 / 49 / 45 |
| `dGOLD` | 2x | +0.001% | -0.05% | -0.50% | -3.8% | -0.59% | -7.3% | +3.3% | 0.0% | 251 / 49 / 45 |
| `uSILVER` | 2x | -0.029% | -0.39% | -0.22% | -9.5% | -2.55% | +4.5% | +4.0% | 2.2% | 251 / 49 / 45 |
| `dSILVER` | 2x | +0.102% | +0.29% | +1.69% | -17.2% | -2.24% | -17.6% | +4.0% | 15.6% | 251 / 49 / 45 |
| `uWTIOIL` | 2x | -0.027% | -0.38% | -1.05% | -7.2% | -0.27% | +10.3% | +6.7% | 2.2% | 251 / 49 / 45 |
| `dWTIOIL` | 2x | -0.057% | +0.07% | +2.54% | -9.8% | -2.26% | -19.7% | +6.7% | 13.3% | 251 / 49 / 45 |
| `uUS500` | 3x | -0.013% | -0.09% | -0.48% | -1.6% | -0.46% | +4.3% | +1.5% | 0.0% | 251 / 49 / 45 |
| `dUS500` | 3x | -0.006% | -0.04% | -0.21% | -2.1% | -0.45% | -5.9% | +1.5% | 0.0% | 251 / 49 / 45 |
| `uUS100` | 3x | -0.009% | -0.08% | -0.22% | -2.1% | -0.90% | +0.1% | +0.6% | 0.0% | 251 / 49 / 45 |
| `dUS100` | 3x | -0.026% | -0.26% | -0.77% | -5.2% | -2.59% | -5.8% | +0.6% | 2.2% | 251 / 49 / 45 |
| `uTSLA` | 2x | +0.025% | -0.01% | -0.54% | -2.7% | -0.67% | -10.8% | -4.9% | 0.0% | 251 / 49 / 45 |
| `dTSLA` | 2x | -0.069% | -0.76% | -6.13% | -14.3% | -6.18% | -0.4% | -4.9% | 0.0% | 251 / 49 / 45 |
| `uNVDA` | 2x | +0.017% | +0.02% | -0.54% | -1.9% | -0.29% | +1.6% | +1.0% | 0.0% | 251 / 49 / 45 |
| `dNVDA` | 2x | -0.032% | -0.39% | -3.84% | -7.6% | -5.03% | -8.0% | +1.0% | 0.0% | 251 / 49 / 45 |
| `uAAPL` | 2x | +0.005% | +0.03% | -0.01% | -0.7% | -0.11% | +6.5% | +3.8% | 0.0% | 251 / 49 / 45 |
| `dAAPL` | 2x | -0.010% | -0.04% | -0.61% | -3.8% | -1.15% | -8.4% | +3.8% | 0.0% | 251 / 49 / 45 |
| `uMSTR` | 2x | +0.025% | -0.07% | +1.34% | -6.7% | -1.61% | -30.0% | -13.3% | 26.7% | 251 / 49 / 45 |
| `dMSTR` | 2x | -0.103% | -1.33% | -2.70% | -30.6% | -10.07% | -0.2% | -13.3% | 15.6% | 251 / 49 / 45 |
| `uCOIN` | 2x | -0.009% | -0.48% | -2.96% | -9.2% | -3.27% | -20.6% | -7.4% | 11.1% | 251 / 49 / 45 |
| `dCOIN` | 2x | -0.149% | -2.23% | -14.26% | -27.5% | -17.95% | -10.7% | -7.4% | 2.2% | 251 / 49 / 45 |
| `uHOOD` | 2x | +0.037% | -0.21% | -1.18% | -5.7% | -3.17% | -13.3% | -3.4% | 11.1% | 251 / 49 / 45 |
| `dHOOD` | 2x | -0.091% | -1.66% | -8.35% | -24.7% | -13.97% | -14.4% | -3.4% | 15.6% | 251 / 49 / 45 |
| `uSPCX` | 2x | +0.001% | +0.86% | +2.13% | -3.9% | +2.15% | -16.0% | -5.8% | 28.6% | 61 / 11 / 7 |
| `dSPCX` | 2x | -0.239% | -0.39% | -11.72% | -36.3% | -9.37% | -24.1% | -5.8% | 28.6% | 61 / 11 / 7 |
| `uSNDK` | 2x | -0.019% | -0.87% | +16.37% | -22.7% | +1.98% | +103.4% | +54.8% | 6.7% | 251 / 49 / 45 |
| `dSNDK` | 2x | -0.113% | -1.05% | +81.46% | -39.6% | +73.89% | -83.0% | +78.4% | 94.1% | 83 / 17 / 17 |
| `uMU` | 2x | -0.015% | +0.03% | +6.24% | -10.3% | -2.39% | +36.8% | +22.5% | 2.2% | 251 / 49 / 45 |
| `dMU` | 2x | -0.152% | +0.70% | +26.40% | -19.6% | +7.90% | -51.2% | +29.0% | 54.3% | 171 / 35 / 35 |
| `uPLTR` | 2x | +0.018% | -0.04% | -0.95% | -4.7% | -1.41% | -13.6% | -4.6% | 2.2% | 251 / 49 / 45 |
| `dPLTR` | 2x | -0.017% | -0.26% | -5.19% | -15.7% | -8.01% | -6.8% | -4.6% | 13.3% | 251 / 49 / 45 |
| `uLLY` | 2x | +0.015% | +0.06% | +0.41% | -0.9% | -0.13% | +10.3% | +5.5% | 0.0% | 251 / 49 / 45 |
| `dLLY` | 2x | -0.021% | -0.23% | -0.12% | -4.6% | -2.91% | -14.5% | +5.5% | 0.0% | 251 / 49 / 45 |
| `uMRNA` | 2x | +0.851% | +1.21% | +5.27% | -12.8% | -5.62% | +16.6% | +12.2% | 0.0% | 251 / 49 / 45 |
| `dMRNA` | 2x | +0.825% | +1.94% | +12.72% | -25.0% | -14.50% | -41.5% | +12.2% | 44.4% | 235 / 47 / 45 |

Windows are sampled every trading day at 24h and every fifth day at 7d and 30d, so a 30-day row's windows do not overlap in 29 days out of 30. `holdingCostTable()` in `@lev/sim` produces this for any asset, policy and side; the web reads it from `packages/sim/data/grades.json` under `holdingCost`, where `headlineField` names the one to show. A phrasing the web can use directly, taking `uTSLA` as the example: *"holding this for a month has typically cost 0.5% in decay, and 2.7% in the worst tenth of months"*.

Three things to read carefully.

- **The 24h figures are small** — a fraction of a basis point on the indexes — and that is the honest answer for a day. Decay is a function of how long you hold, and it compounds; quoting only the 30-day number would overstate the cost to someone trading in and out, and quoting only the 24h number would understate it to everyone else. That is why D26 asks for all three.
- **A negative 30-day median with a positive stock column** is the case the disclosure exists for: the underlying went up and the holder still lost, because the path was choppy.
- **Some rows show a *positive* decay, and that is not an error.** A constant-leverage product cuts its exposure as it loses and adds as it wins, so in a sustained trend it does better than N times the move. `dSNDK` is the extreme case: SanDisk rallied hard over the window, and a 2x short that delevered all the way down lost far less than 2× the rally would imply. The closed-form `N(N−1)σ²/2` the listing rule uses is an unconditional expectation; a measured figure over one particular trending year can land either side of it. **Read a positive number as "this year trended", never as "this token is free".** Anything built on these figures has to say so, because a holder who reads "+2%" as a property of the product will be wrong the moment the trend breaks.

### Why these numbers are smaller than the textbook figure

A daily-rebalanced leveraged fund pays the full `N(N−1)σ²/2`. **Ours does not rebalance daily** — D6 gives it a ±10% band and it only trades when realised leverage leaves that band. A saw-tooth is the clean demonstration, and `holding.test.ts` pins it: a 6% daily oscillation costs a 2x token **nothing beyond fees**, because leverage never leaves the band; the same path at 3x also costs nothing; a 10% oscillation costs 3x about 44% over 30 days while 2x still pays only fees; at 15% both are paying heavily. The band is doing real work, and it is most of why the measured 24h and 7d numbers above are as small as they are.

The consequence for the disclosure is worth being precise about: the cost is not linear in volatility, it is closer to a threshold. Ordinary chop is free. It is the moves large enough to force a rebalance — and the days when those come repeatedly — that the holder pays for, which is the same tail the *worst tenth* column is measuring.

## Listing against grade

D26 decides what is listed and at what leverage; this file grades the risk. They are answers to different questions and they disagree on several markets by design:

| Market | Listed at (D26) | Rule grades | Gap | Bound by | σ graded | Worst gap 5y | Venue max |
|---|---|---|---|---|---|---|---|
| `AAPL` (AAPL) | 2x | 2x | agree | decay | 28% | 9.4% | 20x |
| `COIN` (COIN) | 2x | — | **graded unlistable** | decay | 86% | 31.3% | 10x |
| `GOLD` (GOLD) | 2x | 2x | agree | decay | 29% | 6.0% | 25x |
| `HOOD` (HOOD) | 2x | — | **graded unlistable** | decay | 72% | 23.1% | 10x |
| `LLY` (LLY) | 2x | 1.5x | **listed 0.5x above grade** | decay | 36% | 14.4% | 10x |
| `MRNA` (MRNA) | 2x | — | **graded unlistable** | decay | 192% | 84.3% | 10x |
| `MSTR` (MSTR) | 2x | — | **graded unlistable** | decay | 91% | 27.4% | 10x |
| `MU` (MU) | 2x | — | **graded unlistable** | decay | 82% | 18.1% | 15x |
| `NVDA` (NVDA) | 2x | — | **graded unlistable** | gap | 52% | 26.1% | 20x |
| `PLTR` (PLTR) | 2x | — | **graded unlistable** | decay | 67% | 22.8% | 10x |
| `QQQ` (US100) | 3x | 2.5x | **listed 0.5x above grade** | decay | 23% | 5.4% | 20x |
| `SILVER` (SILVER) | 2x | — | **graded unlistable** | decay | 62% | 15.4% | 25x |
| `SNDK` (SNDK) | 2x | — | **graded unlistable** | decay | 116% | 20.8% | 15x |
| `SPCX` (SPCX) | 2x | — | **graded unlistable** | decay | 82% | 10.3% | 15x |
| `SPY` (US500) | 3x | 3x | agree | none | 17% | 4.0% | 20x |
| `TSLA` (TSLA) | 2x | — | **graded unlistable** | decay | 60% | 14.5% | 20x |
| `WTIOIL` (WTIOIL) | 2x | 1.5x | **listed 0.5x above grade** | decay | 49% | 13.8% | 20x |

A row marked **listed above grade** is a deliberate D26 choice, not an oversight: demand is concentrated in exactly the volatile names the rule refuses, and the operator decision was that the answer to decay is disclosure at the point of trade. What this file owes that decision is an honest number, which is the section above. The one thing the grade should still be allowed to veto is a market whose *gap* the keeper cannot survive at the listed leverage — that is a solvency question rather than a cost one, and it is the column to watch in the table above.