01 / Data and measurement windows
The downloadable workbook states a 2026 YTD window through September 18, 2026. Its daily calculation ranges use rows 4–182, or 179 observations. The website separately displays an October 6 market snapshot. These are different cutoffs and should not be treated as one synchronized dataset.
| Surface | Current implementation | Interpretation |
|---|---|---|
| Workbook risk statistics | Calculations on stored return observations | Historical sample estimates; underlying price provenance still needs independent validation |
| Web correlation explorer | Fixed, manually entered pair matrix | Illustrative relationships; no live window selection or return-series recalculation |
| Macro and policy sliders | Fixed sensitivity coefficients | Hypothetical shock analysis, not regression estimates |
| Index forecaster | Stored anchor levels, stylized recent paths, assumed event effects | Scenario illustration, not a fitted historical chart or price prediction |
| Quotes and market monitor | Static snapshots | Not live prices; verify independently before use |
A reproducible historical study should use aligned daily total returns over a stated window, document dividends and splits, and match only dates with valid observations. Price-level correlations can be misleading. Inflation is monthly, so it must not be treated as an independent daily observation by repeating each release across days.
02 / Portfolio and risk formulas
Returns and position weights
G = Σ |Vᵢ| · wᵢ = signed Vᵢ / G
rₚ,ₜ = Σ wᵢ × rᵢ,ₜP is an adjusted price or total-return index. G is gross exposure. A short position has a negative signed weight. Holding initial weights constant describes a fixed-weight model, not necessarily an unrebalanced buy-and-hold book.
Correlation versus beta
βᵢ = Cov(rᵢ, rSPY) / Var(rSPY)
βₚ = Σ wᵢ × βᵢCorrelation measures direction and consistency of co-movement, bounded by −1 and +1. Beta measures exposure magnitude: 1.5 implies an estimated 1.5% response to a 1% benchmark move, before asset-specific effects.
Sample covariance and sample variance should use the same denominator. The compatibility-repaired workbook uses COVAR with VAR in some cells, mixing population covariance with sample variance; this slightly understates beta by (n−1)/n. With 179 observations the difference is about 0.56%. This remains a documented limitation, not a claim of fully audited native Excel behavior.
Volatility, drawdown and historical VaR
Wealthₜ = Wealthₜ₋₁ × (1 + rₚ,ₜ)
Drawdownₜ = Wealthₜ / max(Wealth₀ … Wealthₜ) − 1
95% historical daily VaR = −5th percentile of daily returns252 is a trading-day convention. Maximum drawdown is the lowest drawdown. VaR is a historical percentile, not the worst possible loss or a guarantee.
Hedge sizing
Minimum-variance hedge ratio = Cov(rₚ, rh) / Var(rh)For a fraction f of beta removed, βtarget = (1−f) × βₚ. A negative minimum-variance ratio can indicate a long diversifying hedge. Notional, borrow cost, leverage and residual exposure must be considered separately.
03 / Every macro factor, explained
Open a factor for its units and transmission channel. These channels explain intuition; they do not validate the exact numerical coefficients.
Oil / USO · percentage change
A supply-led oil increase can raise inflation and production costs, while benefiting energy producers. A demand-led rally can instead signal stronger growth. USO is futures-based exposure, not spot crude; rolling futures introduces basis and roll effects.
10-year Treasury yield · basis points
Higher yields increase discount rates and can pressure long-duration valuations. Bond prices generally fall when yields rise. A yield increase caused by stronger growth has different equity implications from an inflation shock. 50 bps equals 0.50 percentage points, not 50%.
U.S. dollar · percentage change
Dollar strength affects translated overseas earnings and dollar-denominated financing. The index proxy and a currency pair are not interchangeable. Company hedges and revenue geography alter sensitivity.
Inflation surprise · percentage points
This slider is an assumed inflation shock, not a forecast of a CPI release. Higher inflation can affect margins, real yields and the policy path. A 0.1 pp surprise is distinct from a 0.1% relative price change.
VIX · index points
VIX measures option-implied 30-day S&P 500 volatility. Rising VIX often accompanies equity stress. It is not directly investable; VIX futures and volatility funds can behave differently. See the Cboe reference below.
Gold / GLD · commodity exposure
Gold reacts to real yields, dollar moves and safe-haven demand. It is not a guaranteed inflation hedge over every window and can fall during liquidity-driven selling.
Treasuries / TLT / TIPS · duration and inflation
TLT represents long-duration bonds, not bills or a 10-year yield series. Bills have short maturities, notes intermediate maturities and bonds longer maturities. TIPS link principal to inflation, but prices also respond to real yields; inflation protection does not eliminate duration risk.
Credit / HYG · spreads and growth
High-yield bonds combine Treasury-rate exposure with credit risk. Wider spreads can hurt prices even when Treasury yields fall. Credit spreads and bond price returns should not be confused.
Equities / technology / energy · factor exposures
SPY is broad large-cap equity exposure; QQQ is concentrated in Nasdaq-100 companies. Semiconductors add industry-cycle and concentration risk. XLE and XOM add energy exposure. Banks, healthcare and consumer companies also carry company-specific risks that broad factors cannot fully explain.
How the portfolio sliders combine
+ bUSD × ΔUSD + bInflation × (Δpp × 10) + bVIX × ΔVIX
Contribution (%) = asset shock × weight × long/short sign
Estimated P&L = $50,000 × portfolio shock / 100For NVDA, the assumed coefficients are −0.119, −0.31, −0.24, −0.18 and −0.18 respectively. An oil-only +10% shock therefore produces −1.19%, and a 24% long weight contributes −0.2856%. These are assumed sensitivities, not observed causal effects.
04 / FOMC surprise model
Reaction = −surprise coefficient × surprise units
+ tone coefficient × tone + QT coefficient × QTTone: dovish −1, balanced 0, hawkish +1. QT: slowing −1, steady 0, accelerating +1. An expected decision creates zero rate surprise, but communication can still change the result.
The coefficients are educational assumptions. The current output uses a common percentage-style display even for 2Y and 10Y, so those readings are directional model scores—not Treasury-yield changes in basis points. Do not compare them as equivalent tradable returns.
05 / Index scenario assumptions
Scenario level = current anchor × (1 + S / 100)Horizon factors: 1 month 0.48×; 3 months 1.00×; 6 months 1.32×; 12 months 1.62×. These are manually selected scaling assumptions.
| 3-month event effect | S&P | Nasdaq | Dow | Russell |
|---|---|---|---|---|
| 10Y +50 bps | −3.1% | −5.2% | −2.2% | −4.4% |
| Regional escalation | −2.4% | −2.8% | −2.1% | −3.1% |
| Oil supply shock | −1.8% | −2.1% | −1.2% | −2.6% |
| Dovish Fed | +2.7% | +4.1% | +1.9% | +4.8% |
| Strong earnings | +2.4% | +3.6% | +1.8% | +2.9% |
| Global risk-off | −3.5% | −4.4% | −3.0% | −5.3% |
Growth stocks receive larger assumed rate effects; small caps receive larger easing and risk-off effects. No empirical calibration currently establishes that 2.4% is more accurate than 2% or 5%. The equation explains arithmetic, not predictive validation.
Combining events
Three-month pair adjustments, in percentage points: escalation + oil −0.7; oil + rates −0.5; Fed + rates +0.9; earnings + rates +0.4; earnings + risk-off +0.3; Fed + risk-off +0.4; Fed + oil +0.2; escalation + risk-off −0.8. The same horizon factor scales these assumptions.
The dashed baseline assumes monthly drift of 0.45% for S&P, 0.62% for Nasdaq, 0.32% for Dow and 0.55% for Russell. It is separate from the event scenario. The displayed recent path is stylized, and the forward line includes a sinusoidal visual adjustment. Its final plotted point may differ slightly from the scenario-level card; the tooltip uses a simplified interpolation.
06 / What these models cannot establish
- Correlation is not causation and does not guarantee a hedge will work in stress.
- Static snapshots and manually specified coefficients are not live feeds or fitted forecasts.
- Costs, taxes, spreads, borrow fees, dividends and financing are not fully reflected in the web scenarios.
- Events overlap; linear addition and fixed pair adjustments cannot capture every nonlinear market reaction.
- The same shock can behave differently across growth, inflation and liquidity regimes.
- Source provenance, consistent statistical denominators and out-of-sample validation are required before treating the project as an audited trading model.
07 / Reference sources and provenance
These primary references explain the instruments and economic series. They are not evidence that the website’s coefficients or correlation matrix were estimated from their data.
- FRED: 10-year Treasury constant-maturity yield
- FRED: Consumer Price Index
- FRED: effective federal funds rate
- Cboe: volatility index overview
- State Street: SPY · Invesco: QQQ
- iShares: TLT · iShares: HYG
Implementation reference: the current downloadable workbook and website calculation code. Reproduction requires the exact input series, adjustment policy, dates and model version; the present site does not provide a fully verified external market-data lineage.