---
type: methodology
status: active
version: 0.1
created: 2026-06-15
source: Process.pdf §4 (Risk framework) + §8 (Immutability) + §9 (Anonymity)
---

# Risk Framework

State it once. Follow it always. Publishing your drawdown protocol and visibly obeying it is worth more to a hiring PM than the Sharpe ratio.

## 1. Per-trade risk

- **Risk per trade: 25-75 bp of NAV at the stop.**
- This is the size of the loss if the stop is hit, expressed in basis points of current NAV.
- Every thesis in `02-trades/<id>/thesis.md` must state the bp risk and the level it implies.

## 2. Portfolio open-risk cap

- **Open-risk cap: 250 bp of NAV.**
- Sum of (per-trade bp risk × current open size) across all open positions must not exceed 250 bp.
- Recompute at every entry, every trim, every add.

## 3. Correlation buckets

Per-bucket caps prevent the "five trades, all USD" problem. Buckets and caps:

| Bucket | Instruments | Cap (bp of NAV) |
|---|---|---|
| USD-bloc | DXY-expressed: 6E, 6J, 6B, 6A, GC (USD-denominated) | Set in [[risk-framework]] §3 |
| Duration | ZT, ZN, ZB, Schatz, Bund, SOFR | TBD |
| Commodity | GC, SI, HG, CL (+ NG if traded) | TBD |
| Equity-beta | ES / MES, FESX | TBD |

(Per-bucket cap values to be set in next methodology pass. The point is they exist; the exact numbers are a calibration exercise.)

## 4. Margin and equity ceiling

- **30% margin / equity ceiling.** If initial margin × maintenance factor would push margin usage above 30% of equity, do not enter. Trim to fit.

## 5. Drawdown protocol

Peak-to-trough, measured against the high-water mark in the NAV series.

| DD level | Action |
|---|---|
| -5% | Halve all sizing. Continue trading. Document in the daily journal. |
| -10% | Flat all positions. **Written review required before resuming.** No exceptions. |
| -15% | Hard stop. Restart only after a full quarterly process audit + an explicit written sign-off in `90-archive/process-changelog.md`. |

The review at -10% must answer:
- What rule was violated, if any?
- Was the system itself broken (signal, sizing, regime filter), or was this a known-tail event?
- What is the specific change being made to the spec, with a date and version bump?

## 6. Trade-construction checks (every entry)

Before any entry, confirm in the thesis:
- [ ] Risk in bp is between 25 and 75
- [ ] Per-bucket cap not exceeded
- [ ] Portfolio open-risk after this trade ≤ 250 bp
- [ ] Margin usage after this trade ≤ 30% of equity
- [ ] Invalidation condition is a fact, not a price
- [ ] Add/trim rules pre-defined
- [ ] Event proximity: not inside the blackout window for a Tier 1 event (see [[process#3. Data and event coverage]])

## 7. Anonymity and publication hygiene

Operational rules for the pseudonym track (gated — see `README.md` §Gates and `TODO.md` §G):

- Publish under pseudonym on Substack + X; Darwinex alias for verified track record.
- Separate Proton email for the publication account. Do not cross-pollinate with employer-facing accounts.
- **No Swiss-specific tells in writing if you care about employer discovery.** Avoid: "CFA Society Switzerland," "Alpen Partners," fund names your employer holds, identifiable client-side references.
- **Do not comment on names your employer holds.**
- PGP keypair generated day one. Public fingerprint in Substack "About." Sign every published piece.
- "Identity commitment" hash (your real name + nonce) published in first post — the cryptographic proof that lets you later claim the body of work.

## 8. NAV calculation

Methodology: time-weighted, deposits/withdrawals separated. GIPS-inspired one-pager at `00-methodology/nav-methodology.md` (TODO §B — to be written).

Daily series: `04-performance/nav/NAV.csv`. Base 100 at inception.

## 9. Audit trail (immutability layers)

| Layer | Source | What it proves |
|---|---|---|
| IBKR Flex statements | `04-performance/flex-statements/` | Ground truth — what was actually traded, when, at what price |
| Darwinex / FundSeeder | third-party | Independently verified, risk-adjusted track record under alias |
| OpenTimestamps | hash of thesis / note PDFs | Document existed at timestamp T (Bitcoin-anchored) |
| Git signed commits | this repo | Research history with author identity |
| Substack / X | published posts | Public distribution |

For every published piece, the post footer must include the OpenTimestamps hash.