Reading a Book: Factor, Concentration, and Tail-Risk
Decomposition of a Hypothetical Long/Short Equity Portfolio

In one line: A fund that looked diversified was really a market bet, with one stock driving 41% of its risk.

Abstract. I decompose the risk of a single hypothetical long/short equity book (ten US large-cap names, six long and four short, gross 100%, net +30%) to separate a portfolio’s dollar description from its risk description. Regressed on the Fama–French five factors plus momentum with Newey–West errors over 5,384 trading days (2005–2026), the book carries a realized market beta of 0.56 on +30% net dollars, roughly 1.9× its net-implied direction, and an Investment (CMA) beta of −0.50. The six factors explain 61.3% of daily variance, leaving 38.7% idiosyncratic, and residual alpha is +8.1% annualized (t = 3.75) in sample. By risk rather than dollars the book is concentrated: NVDA is 16.0% of gross but 41.4% of variance. One-day 99% historical value-at-risk is 2.67% (expected shortfall 3.29%). A portfolio that reads as diversified stock selection is in fact directional, concentrated, and factor-driven; on a survivorship-biased universe every in-sample figure is an upper bound.

1. Introduction

A portfolio is a set of claims. The manager asserts that returns originate in his ideas; the risk function establishes what the book actually owns. This note runs that decomposition end to end on a single portfolio, following the sequence a platform risk desk applies when a new manager’s positions arrive: take the book, decompose its risk, return the findings. The portfolio is hypothetical and illustrative, constructed to be taken apart rather than traded, and is not a real position or a recommendation. It is a constant-weight, daily-rebalanced construction held over the full price history on a universe of current large-cap survivors, so it carries hindsight of which names survived. Every in-sample figure, the residual alpha above all, is an upper bound rather than a forecast.

2. The portfolio

The book holds ten US large-cap names, six long and four short. Weights are fractions of NAV set by hand for the example; long dollars concentrate in mega-cap growth, short dollars in defensives.

3. Factor decomposition

The constant-weight book’s daily return is regressed on the Fama–French five factors plus Carhart momentum, with Newey–West standard errors, over 5,384 days from 2005-01-04 to 2026-05-29. The intercept is the residual alpha; the slopes report the exposures the book carries whether or not they were intended.

3.1 Fit and residual alpha

The six factors explain 61.3% of the book’s daily variance, leaving 38.7% idiosyncratic. Annualized residual alpha is +8.1% with a t-statistic of 3.75, positive and significant in sample. On a survivorship-biased universe it is a ceiling, not a track record.

3.2 Unintended exposures

• Market (Mkt-RF) beta +0.56. Net dollar exposure is only +30%, yet the realized market beta is 0.56; the shorts are low-beta defensives, so netting dollars did not net out direction. In beta terms the book is about 1.9× more market-directional than its net book implies. Unintended.
• Investment (CMA) beta −0.50. A tilt toward aggressive, high-reinvestment growth (long mega-cap tech) funded by short conservative payers. A style bet, not stock selection. Unintended.
• Value (HML) beta −0.17. Net short value, long growth. A style bet, not stock selection. Unintended.
• Profitability (RMW) beta −0.13. Mildly short quality; the shorts are the profitable defensives. A style bet, not stock selection. Unintended.
• Size (SMB) beta −0.03. Negligible; both sleeves are large-cap, so the book is correctly size-neutral. Intended (clean).

4. Concentration and sizing

By dollars the book appears spread out; by risk it does not.

Dollar weights imply 9.3 effective bets, but one name, NVDA, supplies 41.4% of the variance. Marginal risk contribution by name, from the annualized covariance of the positions:

5. Value-at-risk and expected shortfall

One-day historical simulation on the book’s 5,422 daily returns, expressed as a percent of capital (gross 100%). Value-at-risk is the loss exceeded on all but 1% (or 5%) of days; expected shortfall is the mean loss on the days that breach it.

6. Historical stress

These are realized returns, not model losses: the book’s actual return over historical windows, computed by compounding its constant-weight daily returns across each range.

The COVID window is the instructive one: the book gave back only −4.8% peak-to-trough because its low-beta shorts fell alongside its longs, while the −19.0% Global Financial Crisis draw shows the same hidden market beta biting once the shorts stopped hedging.

7. Findings and recommendations

Three issues dominate the teardown, each with a concrete remedy.

1. Hidden market beta. The book runs a 0.56 market beta on +30% net dollars, roughly 1.9× more directional than the net book suggests, because every short is a low-beta defensive. Fix: size to a beta target rather than a dollar target, substituting higher-beta shorts or overlaying an index short until realized beta matches the intended net.
2. Risk concentrated in one name. NVDA is 16.0% of gross dollars but 41.4% of portfolio variance; the book is diversified by dollars and concentrated by risk (9.3 effective bets by dollars, far fewer by risk). Fix: cap single-name risk contribution, not only single-name weight, trimming NVDA or adding an offsetting position.
3. A style bet in a stock-picker’s coat. Only 38.7% of daily variance is idiosyncratic; 61.3% is the six factors, led by an Investment (CMA) beta of −0.50, a large long-growth / short-conservative tilt. Fix: neutralize the CMA/growth tilt if the thesis is single-name selection, or own it deliberately as the strategy and budget risk to it.