--- type: methodology status: draft version: 0.2 created: 2026-06-15 updated: 2026-07-01 source: Process.pdf §4 — "technical trigger (define it precisely once in a standing methodology doc, then reference it)" change: v0.1 → v0.2 (2026-07-01). Replaced single accelerating-momentum signal definition with a trigger-type taxonomy. Trigger types: positioning, microstructure, momentum, event. Each type follows the same template; a trade's §2 Entry trigger references one or more types. --- # Signal Specification — Entry Triggers **Status: DRAFT.** The PDF says define this once, then reference it from every thesis. This is a placeholder for the spec you write. Do not commit a real trade on this version — finish the spec first. ## What this doc is The exact, reproducible definition of the entry trigger(s) you reference from every pre-trade thesis (`02-trades//thesis.md` §Entry trigger). One definition per trigger type, versioned here, referenced everywhere. If a definition changes, change it here once and bump the version; don't quietly drift. ## Why a trigger taxonomy (not a single signal) Not every trade is triggered by the same market condition. A positioning-extreme trade (Block 005 two-trigger rule) doesn't have a momentum signal — it has a CFTC reading. An event-driven trade (FOMC, CPI) doesn't have momentum — it has a print or a presser. A microstructure trade (curve regime shift, breakeven roll) doesn't have CFTC — it has yield-curve dynamics. Forcing all triggers through a single "accelerating-momentum" template made sense when the program was one PDF; with Phase 1/2/3/4 material teaching four distinct trigger mechanisms, the spec has to acknowledge them. A trade's §2 Entry trigger names *which type(s)* apply and references the matching definition below. A trade can have multiple trigger types (the two-trigger rule uses two). ## Trigger types ### Type A — Positioning trigger The market is positioned wrong. The trigger is a *reading* (CFTC COT, ETF flow, CTA signal, fund survey, dealer positioning) hitting a pre-defined extreme. The trade is the unwind. **Typical use:** crowded-trade reversals. Block 005 Taper Tantrum was a positioning trigger on duration (spec positioning long, secular stagnation narrative anchored). **Litmus:** if the trigger reading moved back to neutral, would you cover? If yes, it's a positioning trigger. If the position can be neutral and you'd still be long, it's not a positioning trigger. ### Type B — Microstructure trigger The market's internal mechanics (curve shape, term-structure, basis, breakeven, vol surface) are starting to shift. The trigger is a *regime change in market plumbing*. **Typical use:** curve trades (PC1/PC2/PC3-emergence per Block 003/004/005/007), cross-currency basis shifts (Block 014), vol-surface regime shifts (Block 010). **Litmus:** is the change in how the market is priced, not in what people think? A yield curve regime shift is microstructure. A new Fed chair being more hawkish is macro driver, not microstructure. ### Type C — Momentum trigger Price/yield is accelerating in the direction of the view. The trigger is a *rate-of-change* (price vs MA, slope of MA, second derivative, RSI, MACD, ATR percentile). **Typical use:** trend-following expressions of a macro view. Often the "and the market is starting to agree" half of a two-trigger setup. **Litmus:** does removing the price chart make the trigger disappear? If yes, it's a momentum trigger. If the trigger survives without the chart (CFTC reading, breakeven regime shift), it's not. ### Type D — Event trigger A scheduled or unscheduled event will resolve the view. The trigger is the *event itself* (FOMC, CPI, NFP, ECB, OPEC meeting, refunding announcement, election, geopolitical flashpoint). **Typical use:** event-driven positioning. The trade is constructed around the event, not held through it (or held through it with explicit acknowledgment). **Litmus:** is there a specific calendar date / time when the view resolves or re-prices? If yes, it's an event trigger. If the view plays out over weeks/months without a single resolution point, it's not. ## Required sections per trigger type (template — fill in for each type you use) For each of Type A/B/C/D that you activate, fill in this template. The first version only needs to be filled for the type you trade most; the rest are placeholders. ### Type A — Positioning #### 1. Universe Which instruments does this trigger apply to? (Per Coverage Universe — likely all liquid futures + the monitored set where positioning data is available. Some instruments excluded by construction, e.g. NG.) #### 2. Source CFTC COT (Friday release, lagged), ETF flows (daily for gold, weekly for others), CTA estimates (weekly), fund surveys (BoJ TFX, IMM currency survey), dealer positioning (primary dealer data, when available). #### 3. Threshold definition Define the "extreme" reading. Two approaches: - **Percentile-based:** top/bottom decile of the 3-year distribution - **Z-score-based:** |z| > 1.5 or 2.0 vs trailing window State the threshold, the window, and the source series. Concrete numbers, not "extreme." #### 4. Trigger condition Positioning trigger fires when: . #### 5. Filters and disqualifiers - Event proximity (no new positioning triggers inside X days of Tier 1 event — positioning readings don't move on the event) - Liquidity (some positioning series thin in Aug / Dec) - Revision risk (COT revises prior weeks; large revisions invalidate) ### Type B — Microstructure #### 1. Universe Per Coverage Universe. Often specific to curve trades, basis trades, or vol-surface trades. #### 2. Source Curve data (CME, BrokerTec), breakeven series (TIPS, FRED), vol surface (CME, B3, Eurex for swaptions), cross-currency basis (Fed H.4.1, ECB SDW). #### 3. Threshold definition Regime shift, not percentile. Examples: "PC3 emerges (wings opposite body)," "2s10s steepens >X bp in Y days," "10y real yield breaks out of [a, b] range," "10y breakeven prints Z for the first time in 6mo." #### 4. Trigger condition Microstructure trigger fires when: . #### 5. Filters and disqualifiers - Event proximity (curve regime shifts cluster around Tier 1 events — may be event-driven, not pure microstructure) - Data revisions - Cross-asset confirmation (regime shift in one instrument, no confirmation in correlated bucket = noise) ### Type C — Momentum #### 1. Universe All liquid futures in Coverage Universe. #### 2. Timeframes Primary, confirmation, and noise timeframes. State the bar size and session definition (RTH vs 24h). #### 3. Core components The signal must be decomposable into discrete, testable components. Typical building blocks: - Trend (e.g. price vs N-period MA) - Slope (rate of change of the trend) - Acceleration (second derivative) - Volatility regime filter (e.g. ATR percentile) - Cross-asset confirmation (correlation-bucket aligned, e.g. for a USD-bloc FX trade, is DXY confirming?) For each component: name, formula, parameters, default values. #### 4. Directional rules Concrete, not vibes: - Long condition: - Short condition: - Neutral / no-trade band: #### 5. Filters and disqualifiers - Event proximity (no new momentum entries inside X minutes of Tier 1 event) - Vol regime (e.g. no signal in lowest vol decile) - Crowdedness (CFTC COT extreme, ETF flow extreme) - Correlation bucket cap (see risk-framework) ### Type D — Event #### 1. Universe Events on the Tier 1 / Tier 2 list (process.md §3) and any unscheduled events the trade is structured around. #### 2. Source Event calendar (CME, Bloomberg ECO, central bank calendars). Position the trade around the event; resolve or cover on/after the event. #### 3. Threshold definition Asymmetry of the trade. The trigger is the event itself, but the *trade structure* depends on: - What's priced (OIS, fed funds futures, consensus survey) - Skew vs consensus (your view vs the median) - Asymmetry: is your view "more hawkish" or "less hawkish" than consensus, and by how much #### 4. Trigger condition Event trigger fires when: AND . #### 5. Filters and disqualifiers - Stale events (don't trade around an event that's already happened) - Tier classification (Tier 1 events get full previews; Tier 2 get weekly comment; unscheduled events ad hoc) - Slippage budget (event trades pay slippage; budget for it in §5 risk calc) ## Worked examples (template — fill in) 3-5 historical examples per active trigger type. Chart + trigger value at the time + outcome. For each: did the trigger fire at the right time? What did it miss? ## Failure modes (per active trigger type) Known failure patterns and how the spec handles them: - **A (positioning):** COT revises after the trigger fires; crowded trade stays crowded - **B (microstructure):** regime shift is event-driven, not structural; reverses in 48h - **C (momentum):** whipsaw in low-vol regimes; news-driven gaps; crowded-trade reversal - **D (event):** event priced-in (no reaction); event surprise direction wrong; liquidity withdrawal ## Versioning - v0.1 (2026-06-15) — placeholder; single accelerating-momentum signal; no real trades - v0.2 (2026-07-01) — trigger-type taxonomy (positioning / microstructure / momentum / event); each type has its own template; trade §2 references one or more types - v1.0 — locked spec for at least one trigger type, audited - v1.x+ — change only via documented changelog in `90-archive/signal-changelog.md` ## Anti-drift rules 1. The signal spec is referenced from every thesis. If a thesis would require a different definition, you update the spec first, not the thesis. 2. The spec is locked 24h before any trade. No "I'll just use the close-of-today version." 3. Any change to parameters bumps the version, gets dated, and gets a one-paragraph justification in the changelog. 4. Quarterly process audit checks: did any of the last 90 days of theses quietly use a non-current version? 5. Trigger type for a trade must be explicit in the thesis. "I had a signal" is not acceptable — name the type (A/B/C/D) and the spec section it references.