Playbook

Disclaimer: This is my personal learning log and decision framework. Nothing here is financial advice. Style: long-term (months to years), no leverage, low-frequency execution, collaborative research with AI.

1. Scope & Objectives

Time horizon: months to years Primary objective: sustainable long-term compounding with explainable decisions Secondary objective: keep investing in the background (study/life first)

2. Non-Negotiables

• No leverage (no margin trading; leveraged ETFs avoided in current risk environment, conditionally revisitable if vectors change)
• No high-frequency trading
• No mechanical trend following for thesis formation — technical analysis is allowed only at execution stage for entry-point fine-tuning
• Do not change long-term structure due to short-term volatility
• Default to low action when expected return vs near-riskless is unfavorable or unclear
• Limit orders by default; market orders only in emergency exit
• Avoid complexity for the sake of being “smart”

3. World-view, Methodology & Decision Standard

3.1 World-view — Risk as Vectors

Risk does not exist as a single one-dimensional “regime level.” It decomposes into vectors, each hitting different sectors with different direction and magnitude:

• Interest-rate vector — affects highly leveraged + non-yielding assets
• Geopolitical vector — directionally opposite across sectors (defense may benefit; gold’s marginal-holder forced-selling may make it short-term negative)
• AI bubble vector — short-term bearish (best-case priced + story concentration; funding-leverage exposure compounds rate-vector hits); long-term bullish with a ceiling at the 2nd Industrial Revolution analogue; current state: AI is converting existing demand, not creating new
• Regulatory / political faction vector — government-customer dependency and faction alignment (e.g., defense vendors)
• Macro fiscal vector — sovereign debasement pressure (long-term gold thesis)

Any position judgment must land sector-specifically. There is no universal “should I be in stocks right now?” answer.

3.2 Methodology — Interconnections Framework

First-order (company financials) alone is insufficient. Must combine with second-order sector-specific ecosystem / dependency analysis to separate driver from symptom.

“When evaluating an industry, judgment must follow the industry’s own characteristics together with its linkages to certain other fields.”

Each sector has different second-order dimensions — no universal checklist. Build a library over time. Examples accumulated so far:

3.3 Decision Standard — vs Near-Riskless

For any potential position:

1. Apply world-view + methodology → derive risk-adjusted potential return
2. Compare with near-riskless yield (T-bill / money market / short-duration bonds / high-grade CDs — “T-bill” is shorthand for this category)
3. If < near-riskless → skip (opportunity cost insufficient)
4. If > near-riskless → proceed to sizing (Section 5)

Cash level is the cumulative downstream result of these decisions, not a policy target. A high cash share reflects that few opportunities currently pass the EV test, not a deliberate “elevated regime, hold cash” stance.

Monthly inflow provides an organic buffer — there is no fixed cash floor in absolute dollars.

4. AI Collaboration Protocol

The research flow is collaborative, not AI-led or solo:

1. User forms a rough understanding of the company (business, product, competitive position)
2. AI researches financials + Interconnections second-order analysis
3. User raises probing questions to confirm thesis (no yes-man acceptance)
4. Decision: deploy / don’t deploy / add to watchlist

AI is a tool, not a leader of risk decisions. If I cannot state my cushion (price + sizing + falsification + role) in one sentence, default to smaller size or no action.

4.1 KB Auto-Triggers

When financial discussion activates (any ticker / valuation / position / sector / buy-sell question), the AI side additionally consults the local KB:

• Ticker mentioned → silent grep the Watchlist; surface matching thesis stub + trigger conditions
• Sector discussed → check Circle of Competence (in-circle / edge / outside):
  • In-circle → full thesis analysis
  • Edge of circle → flag learning area + open questions
  • Out-of-circle → fast-skip, suggest stay out
• Substantive thesis stated → prompt for Confidence (10/25/50/75/90) + Deadline (YYYY-MM-DD) + Verification source → log to Calibration open predictions
• Build-up / trim / entry / exit decision → check against §5 sizing ladder + verify trigger criteria

Override: say “skip the check” to disable for the current turn. Conflict resolution: if KB says X and current discussion says Y → surface the conflict for explicit reconciliation, don’t silently overwrite.

5. Sizing — Dynamic Conviction Ladder

Not static categories. Every position has both upward and downward mobility; the tier reflects current conviction, not a permanent label.

5.1 Test-tier (entry)

• Used when thesis is generally bullish but not yet verified
• Entry conditions: low valuation + healthy financials + thesis directionally positive
• Typical size: ~2% NAV (reference: DUOL Tranche 1)
• Purpose: small position to observe and validate

5.2 Build-up

• Triggered when the thesis’s achievability has been verified by subsequent evidence
• Mechanism: staged adds with continued validation, not one-shot to target
• Requires an explicit deployment plan (price milestones / event triggers / time pacing)
• Minimum 5 trading days between steps (discipline, not a hard rule)
• Same-day or next-day adds count as the same step

5.3 Reduce / Exit

Multiple triggers, all valid:

• Same-sector swap to lower-risk peer — e.g., AMZN → MSFT inside Mag 7
• Thesis weakening + risk vector deterioration — e.g., TLN exit driven by interest-rate vector + leverage exposure
• Position too large + grinding lower + emotional drain — e.g., PGR at ~20% NAV in persistent decline
• Fundamental judgment of “no recovery” — e.g., PYPL post-FY2025 earnings + CEO change

5.4 Position assessment ruler

• Evaluate positions by relative-to-index performance (outperform / underperform the benchmark by X percentage points)
• NOT absolute % NAV loss threshold — in extreme regimes the broad market also drops, so absolute loss is uninformative

5.5 Cooling-off after large deposits

• Write the deployment plan first, then act
• All entries staged; no all-in deployment on freshly arrived capital
• Default to the gentlest options (near-riskless cash equivalents or staged adds to highest-conviction names)

6. Position Role Configuration

Holdings carry functional roles, not abstract category labels. Configuration considerations:

• Sector exposure needs (e.g., conscious decision to maintain some technology exposure)
• Sector-internal lowest-risk pick (e.g., among Mag 7, MSFT carries the cleanest AI exposure)
• Upstream/downstream pairing (e.g., V payment network + AXP card issuer)
• Macro hedge (e.g., GLDM for fiat debasement)

Dual gate for entry: prior preference (quality / defensiveness / low valuation / sector-exposure need / upstream-downstream pairing) + current pass of world-view & methodology test. Both must apply.

Current position roles snapshot:

7. Exception Clause

I may break a default rule only if:

• the action is small (does not alter overall structure)
• the reason is written in 1–2 sentences:
  • What hypothesis am I betting on?
  • If I’m wrong, how do I exit (time / price / size)?
• the exception is reviewed in the next monthly review

8. Review Cadence (and attention budget)

• Log: short entries close to the decision (3–10 minutes)
• Monthly review: structure, thesis changes, attention budget consumption
• Sector library review: quarterly (90 days) or trigger-based
• Lessons: mistake repeats twice → promote to explicit rule

Attention budget (hard constraint):

• Daily: 15–25 minutes max (check + log + orders)
• Weekly: one deep block (60–90 minutes) for reading + thesis maintenance
• If math / study suffers → reduce investing frequency, not “try harder”

Snapshots live in /invest/log/, not the Playbook.

9. Reference Materials

Operational files (working tools, frequently consulted):

• Watchlist — candidate pool + trigger conditions
• Circle of Competence — in-circle / edge / outside map
• Calibration — Prediction tracking + Tetlock-style calibration
• Calibration Methodology — How to derive a defensible confidence number

Historical record (retained, not currently-binding rules where they conflict with v2.0):

• Lessons — v1.x rules from the learning journey
• Log — Time-stamped notes close to each decision
• Monthly Reviews — Structure-drift and attention-cost checks

10. Change Log

• 2026-01-30 v1.0: initial version; defined Core ETFs as market tracking, stocks as defense/opportunities; established allocation bands and rebalancing rules.

• 2026-01-30 v1.1: removed baseline snapshots from Playbook (moved to Log); added deposit cooling-off rule, AI tool constraint, and attention budget.

• 2026-02-28 v1.2: Added valuation-based rotation clause, time-staged minimum gap, sector concentration cap, and band-violation protocol. Driven by February valuation rotation and SPGI entry speed.

• 2026-05-18 v2.0: major rewrite. Key changes:

  • Removed Core-ETF sleeve as a required structural component (concentrated quality compounders + dynamic cash position is the chosen configuration, not a deviation).
  • Removed fixed allocation bands (Core 35–60%, Stocks 20–50%, Gold 8–15%, Cash 5–30%) — replaced by EV-driven dynamic sizing and downstream cash level.
  • Removed sector 25% hard cap — consciously seek specific sector exposure (e.g., tech via MSFT, financial via AXP+V upstream/downstream pairing).
  • Removed rigid 5-day step gap as a hard rule — preserved as a discipline guideline within build-up sizing.
  • Removed “Sleep Test” absolute-loss framing — replaced by relative-to-index performance ruler.
  • Added §3 World-view (risk vectors) + Methodology (Interconnections framework) + Decision standard (vs near-riskless).
  • Added §5 Dynamic conviction ladder (test-tier → build-up → reduce/exit) replacing static sleeve allocation.
  • Added §6 Functional position-role configuration.
  • Added §4 Collaborative research protocol replacing “independent research” framing.

  Earlier (v1.x) lessons in /invest/lessons/ are retained as historical record of the learning journey, not as currently-binding rules. Where they conflict with v2.0 (e.g., Core-ETF margin-of-safety mechanics), v2.0 supersedes.

• 2026-05-18 v2.1: Added operational tooling layer — Watchlist, Circle of Competence, Calibration, and Calibration Methodology — described in §4.1 and §9.