P06-QUANT / INVESTMENT FLIGHT RECORDER

A return curve is not an investment process.

This cockpit exposes the bridge the first external reviewer correctly asked us to prove: evidence → scenario priors → layer economics → risk → shadow weights. Every number is reproducible; every assumption is labelled; no chart authorizes capital.

CALCULATION BASISData through 2026-10-13Source files: portfolio.csv · tracked market/*.json

CAPITAL BOUNDARYShadow research only · 0 production weights changed · 0 trades authorized · 0 alpha claims

01 / RISK COCKPIT

Risk cockpit

Gross daily-reset exposure replay from 2020. It describes the risk of today’s basket projected backward—not what an investor could have selected then.

Annualized volatility28.6%OBSERVED

Maximum drawdown-37.8%OBSERVED

2022 drawdown-37.1%OBSERVED

Beta to QQQ1.09OBSERVED

Correlation to QQQ0.95OBSERVED

Daily ES 95%-4.1%OBSERVED

Effective independent bets2.7DIAGNOSTIC, NOT TRACK RECORD

Weighted pair correlation0.32DIAGNOSTIC, NOT TRACK RECORD

DIAGNOSTIC, NOT TRACK RECORD

Daily adjusted-close returns; current target weights reset daily from 2020; targets without observable history remain in SHV; 252-session annualization; QQQ benchmark.

Who actually carries the risk?

Marginal contribution to variance over the latest common 252-session window. Twenty-four eligible equities collapse to roughly three independent risk bets.

Semicap + memory26.9%

Models + cloud23.2%

Compute21.7%

Power + grid17.6%

Networking7.8%

EDA + IP2.4%

Data centers0.4%

DR² = (Σwᵢσᵢ)² / (w′Σw) · 252 common sessions · SPCX excluded: 58 < 200 observations

02 / CAUSAL STRESS

Causal stress laboratory

These are disclosed layer shocks, not forecasts. Unlike one uniform equity shock, each scenario specifies where the break begins and how it propagates.

ASSUMPTIONCapability plateau / regulation

capability adoption slows→ hyperscaler capex is repriced→ semicap and compute reset first→ power backlog follows later

ASSUMPTIONAI capex digestion

cloud utilization trails capacity→ orders and accelerator pricing weaken→ foundry, HBM and networks absorb inventory→ power projects are delayed

ASSUMPTIONPower delivery delay

interconnection lead times extend→ commissioning slips→ compute monetizes later→ data-center and power reprice

03 / SCENARIO → WEIGHT

Make 20 / 55 / 25 earn the weights

Move the priors. The shadow layer mix changes mechanically as the probability-weighted average of three disclosed scenario allocations.

Fast takeoff20%Industrial scaling55%Plateau / regulation25%

Input total100%Calculations normalize the three inputs to 100%.

CurrentShadow mixDelta

Models + cloud38.0%37.2%−0.85 pp

Compute17.0%16.8%−0.25 pp

Semicap + memory18.0%16.5%−1.50 pp

Power + grid14.0%14.6%+0.55 pp

Networking5.0%4.7%−0.30 pp

EDA + IP3.0%3.2%+0.15 pp

Data centers1.5%1.5%+0.03 pp

Reserve3.5%5.7%+2.18 pp

This is a transparent sensitivity surface, not a recommendation. Production weights remain unchanged.

04 / REBALANCE SEMANTICS

One threshold no longer does two jobs

Risk review and execution economics are now separate. A 1.00% target is reviewed outside 0.75–1.25%; a separate 0.10 pp de-minimis filter can suppress a tiny trade but cannot widen that risk band.

LEGACY REPLAY

The historical replay still used max(±25% of target, ±0.75 pp). We disclose it instead of silently rewriting history.

FORWARD POLICY

Position risk band ±25% of target; layer band ±20%; execution de-minimis 0.10 pp, applied only after a breach.

05 / POINT-IN-TIME

The honest backtest boundary

A point-in-time backtest is not yet available. We will not call the current 2015 replay investable history until the dated eligible universe and vintage decision ledger exist.

BLOCKEDRequired before release

Dated eligible universes and index constituents
Vintage prices, fundamentals and classifications
Versioned decisions made only with information available then

Current replay may be used for risk diagnostics only.

06 / PROBABILISTIC VALUE

Probabilistic underwriting sandbox

The quick EPS × multiple calculator remains useful for orientation. This separate model adds scenario probabilities, dilution and a hurdle rate, then discounts each terminal value before aggregation.

Entry priceOwner earnings / shareYearsHurdle rate

Fast takeoff

ProbabilityGrowthDilutionExit multiple

Industrial scaling

ProbabilityGrowthDilutionExit multiple

Plateau / regulation

ProbabilityGrowthDilutionExit multiple

Probability-weighted value —Margin vs entry —

Generic sandbox · issuer-specific inputs must come from dated workpapers before any valuation is approved.

07 / FORWARD SCOREBOARD

Forecast outcome ledger

The asset compounds only when forecasts resolve. Until then, calibration is unavailable and the alpha-claim counter stays at zero.

Admission receipts9

Admitted hypotheses0

Resolved outcomes0

Alpha claims0

CalibrationNOT YET AVAILABLE

08 / EXIT MAP

A finite six-step track

P06-QUANT ends at .5. The open data dependency is visible instead of becoming P06-QUANT.1700.

P06-QUANT.0COMPLETE

Publish the external-review receipt and freeze exact calculation definitions.

P06-QUANT.1COMPLETE

Render risk and concentration metrics from reproducible files.

P06-QUANT.2COMPLETE

Publish causal layer stress assumptions and first-order losses without calling them forecasts.

P06-QUANT.3COMPLETE

Make scenario priors mechanically produce a shadow layer mix; no production weights may change.

P06-QUANT.4BLOCKED ON DATA

Replace selection-biased history with a point-in-time eligible universe and publish the vintage ledger.

P06-QUANT.5IN PROGRESS

Connect issuer underwriting to resolved forecasts and publish calibration after sufficient outcomes exist.

Inspect the strategy ↗

Risk cockpit

Who actually carries the risk?

Causal stress laboratory

Make 20 / 55 / 25 earn the weights

One threshold no longer does two jobs

The honest backtest boundary

Probabilistic underwriting sandbox

Fast takeoff

Industrial scaling

Plateau / regulation

Forecast outcome ledger

A finite six-step track

Turn curiosity into a risk decision.