Shadow Valuation Matrix — methodology forward-test framework

The kit chose Greenwald-modified deep value on 2026-05-24 (see methodology-calibration event log). That decision should be tested continuously, not assumed. Every new thesis records what every candidate methodology would have produced — not just the chosen one — so over time the data tells us which framework actually tracks reality on the names AlphaSteve analyzes.

The chosen methodology drives the published central value and trigger. The shadow methodologies are computed in parallel, recorded in the thesis, and aggregated in methodology-calibration. Over 20-30 theses across multiple regimes, the matrix produces a real forward-looking dataset answering "which methodology should we be using."

This is the cheapest good answer to the methodology question — it costs ~20% more time per thesis and produces real signal within 12-24 months rather than requiring 5+ years of single-portfolio attribution. The historical-backtest alternative (methodology-calibration options) has information-leakage problems; the parallel-portfolio alternative costs 4× per thesis. The shadow matrix is the practical middle.

The four candidate methodologies

Each methodology is defined by three things: (1) the valuation engine (how central value is computed), (2) the MoS band required (how much below central value triggers a buy), (3) the default disposition toward growth value and AI-narrative premium.

Notes on each

Pure Klarman/Graham T2. The conservative end. EPV-only; long cash stretches expected. Designed for institutional capital with multi-decade horizons. Will systematically pass on high-multiple growth names.

Greenwald-modified T2. Current chosen calibration. Credits growth value when the three gating tests pass but uses a conservative runway band as the moat-contestation guardrail. Still recognizably deep-value voice.

Buffett-modern T2. Pays up for verified compounders at smaller MoS. Lower turnover. Longer holding periods. The frame Buffett moved to in the early 1970s as Berkshire scaled.

Mauboussin-compounder T2. Focus on durable quality with verified reinvestment economics. Smallest MoS — willing to participate in compounders earlier. Higher tolerance for near-term volatility in exchange for long-term participation.

The matrix format

Every thesis records a shadow valuation matrix as part of the standard valuation section. Format:

### Shadow valuation matrix

| Methodology | Central value | Range | Implied trigger (MoS) | Verdict at current price |
|---|---|---|---|---|
| Pure Klarman / Graham | $A | $A_low – $A_high | $A × (1 − 0.50) | Pass / Pass-with-trigger / Buy |
| **Greenwald-modified** *(chosen)* | **$B** | $B_low – $B_high | **$B × (1 − 0.30)** | **Pass / PWT / Buy** |
| Buffett-modern | $C | $C_low – $C_high | $C × (1 − 0.22) | Pass / PWT / Buy |
| Mauboussin-compounder | $D | $D_low – $D_high | $D × (1 − 0.18) | Pass / PWT / Buy |

**Spread:** $A to $D (ratio: $D / $A)
**At-current-price verdict spread:** how do the four methodologies differ on the action at today's price?
**Where the spread comes from:** brief decomposition of which inputs drive the dispersion (growth-value credit, runway band, MoS requirement)

The chosen methodology's row is bolded. Subsequent rows are shadows — recorded but not acted on.

How each central value is computed (briefly)

Starting from the same underlying inputs (normalized earnings, WACC, ROIIC, reinvestment rate, moat assessment):

Pure Klarman / Graham

• Central = EPV-only = Normalized NOPAT / WACC (then per-share adjustments)
• Growth value: zero
• Range: ± ~15% reflecting normalization uncertainty
• Trigger: Central × 0.50 (50% MoS) on standard quality; lower for fragile names

Greenwald-modified (chosen)

• Central = EPV + Growth Value (when 3 gating tests pass) — narrow-moat runway band (5-7 years)
• Growth value: EPV × (reinvestment × spread / WACC) × runway-adjustment 0.4-0.5
• Range: ± ~20% reflecting growth-value sensitivity to runway assumption
• Trigger: Central × 0.70 (30% MoS) on verified compounders

Buffett-modern

• Central = EPV + Growth Value — wider runway band (8-10 years on verified-moat names)
• Growth value: EPV × (reinvestment × spread / WACC) × runway-adjustment 0.55-0.65
• Range: ± ~25%
• Trigger: Central × 0.78 (22% MoS) on verified-moat compounders

Mauboussin-compounder

• Central = EPV + Growth Value — full runway band (10-15 years on exceptional-moat, documented-reinvestment names)
• Growth value: EPV × (reinvestment × spread / WACC) × runway-adjustment 0.70-0.80
• Range: ± ~30% (widest because longest runway assumption carries most cumulative uncertainty)
• Trigger: Central × 0.82 (18% MoS)

Note: when a methodology's gating tests fail for a particular name, that methodology falls back to its base case (Klarman → EPV-only at $A; Greenwald-modified → EPV-only with 40% MoS; Buffett-modern → EPV-only with 30% MoS; Mauboussin → EPV-only with 25% MoS, but probably should not be paying for a non-compounder under this frame).

Aggregation rules

Aggregated in methodology-calibration under the new "Shadow matrix scorecard" section. At each calibration checkpoint (quarterly, 12-month, 24-month):

For each thesis, record:

• Each methodology's central value at thesis publication
• Subsequent actual price (12-month, 24-month, etc.)
• Absolute error: |methodology central − actual price| / actual price
• Sign of error: did this methodology over- or under-estimate?

Aggregate across theses:

• Hit rate by methodology: % of theses where central value was within ±25% of subsequent actual price
• Median absolute error by methodology: central measure of accuracy
• Bias direction by methodology: are we systematically high or low?
• Regime attribution: do certain methodologies do better in bull / bear / neutral regimes?

The interpretation discipline

Don't react to short-term outcomes. Single-thesis hit rates are pure noise at N<10. Three-thesis runs are noise. Twenty-thesis cross-regime aggregates start to talk.

Don't optimize for the wrong objective. "Best methodology" depends on objective. If the objective is "minimize regret on missed quality compounders," Mauboussin will win. If the objective is "minimize permanent capital loss in any regime," Klarman will win. The chosen methodology should match the kit's stated objective (methodology-calibration explicit objective declaration).

Don't conflate methodology quality with regime fitness. Pure Klarman lost to growth-DCF approaches 2011-2024 not because Klarman was wrong but because growth-favoring regimes rewarded the methodology that participated. The 1992-2010 period rewarded the opposite. A "best methodology" question implicitly assumes regime continuity that is not guaranteed.

Calibration cadence for the matrix

What success looks like

After ~12 months and ~15-20 theses, we should be able to say something like:

• "The chosen methodology (Greenwald-modified) has hit rate X% within ±25% of subsequent 12-month price"
• "Pure Klarman would have hit Y%, Buffett-modern Z%, Mauboypsin W%"
• "The bias direction of the chosen methodology has been [+/−] suggesting [over/under]-conservatism"
• "In the current regime, [methodology] has tracked best"

After ~24-36 months and ~30-50 theses across multiple regime windows, we can credibly say "the kit should be using methodology X" with the evidence to back it. Before that, treat the matrix as descriptive (here is what happened) rather than evaluative (this methodology is best).

What this matrix is not

• Not a backtesting framework. We are tracking forward, not retro-fitting against known outcomes.
• Not a portfolio. We do not deploy capital under the shadow methodologies. Only the chosen one drives Portfolio.
• Not a precise science. Methodology choice has irreducible regime sensitivity. The matrix narrows the uncertainty; it does not eliminate it.
• Not a substitute for the source-discipline and three-gating-test rigor. A shadow value computed loosely is worse than no shadow at all.

Sources

• Greenwald, B. (2001). Value Investing: From Graham to Buffett and Beyond. Wiley. — EPV-plus-growth methodology underlying the Greenwald-modified and Buffett-modern entries.
• Klarman, S. (1991). Margin of Safety. HarperBusiness. — the conservative-end calibration; deep-value MoS bands.
• Buffett, W. Berkshire Hathaway annual shareholder letters, particularly 1989 onward. — Buffett-modern frame.
• Cunningham, L. (2013). The Essays of Warren Buffett (4th ed.). — curated synthesis of Buffett's evolving methodology.
• Mauboussin, M. Expectations Investing (with Rappaport, 2001/2021); "Measuring the Moat" (Credit Suisse / Counterpoint Global). — Mauboussin-compounder frame.
• Mauboussin, M. (2012). The Success Equation. HBR Press. — methodology evaluation discipline (skill vs luck).
• Marks, H. (2011). The Most Important Thing. Columbia. — ch. 17 on honest record-tracking.

Linked

• methodology-calibration — where the matrix data aggregates
• earnings-power-value-greenwald — the EPV-plus-growth math used by three of the four methodologies
• margin-of-safety-pricing — the MoS bands that differ across methodologies
• 02-philosophy-deep-value — the doctrine; this matrix tests whether the doctrine is right
• investment-thesis-template — where the matrix appears in every thesis
• consensus-benchmarking — a related-but-different framework (this is internal methodology comparison; consensus-benchmarking is external)