“I earned 16% last year while the market gained only 10%. I generated 6% alpha.” You can check the difference between those returns. The claim about investing ability needs more evidence.
Perhaps the investor chose well. Perhaps the portfolio was more sensitive to market movements, or held a style that performed particularly well. To assess the record, separate three questions: How much did the account earn? What did it outperform? What remains unexplained by the chosen model?
What you will be able to do
- Calculate an account's return and its difference from a benchmark, explaining what each measures.
- Explain how an assumed beta changes the interpretation of outperformance.
- Save a performance evidence card that separates observed results from unproven skill.
Before you start: Read G06: benchmarks, G08: returns, G09: risk, and G13: uncertainty. This chapter supplies the definitions used in its calculations. Percentage subtraction and multiplication are enough; you do not need to perform a regression, which estimates relationships using observations across periods.
Three records before a ranking of ability
The following records are synthetic. Each account starts with $10,000 and covers the same complete year in U.S. dollars, with no outside contributions or withdrawals. Returns include dividends. Net returns deduct the specified teaching investment costs and exclude personal taxes. The teaching benchmark portfolio has a 10% net total return on the same basis.
A benchmark is a comparison chosen in advance. Does it fit the investment task, and could it be implemented at reasonable cost? Those questions matter before you use the comparison to evaluate someone.
| Account | Gross total return | Net total return | Net return minus benchmark | Additional information |
|---|---|---|---|---|
| A | 13% | 12% | +2 percentage points | Concentrated in a few industries |
| B | 18% | 16% | +6 percentage points | Tilts toward small companies and stocks with recent strong performance |
| C | 9.4% | 9% | −1 percentage point | Its rules require holding more cash |
All three accounts made money. After the teaching costs, A ends with $11,200, B with $11,600, and C with $10,900. These are gains relative to each account's own starting value: their absolute returns. B earned the largest amount in this particular year.
Relative to the benchmark, A led by two percentage points, B by six, and C lagged by one. This difference is commonly called active return. Subtracting 10% from 16% gives six percentage points. It does not mean that the return increased by 6% relative to the benchmark's return.
The table does not establish who had the most skill. You still need returns across periods, portfolio risks, objectives stated in advance, and records of the rules tried. Industry concentration, smaller companies, and cash holdings suggest that these accounts may have taken different risks. A ranking of their returns cannot, by itself, resolve that difference.
A correct subtraction can answer the wrong question
A cash-heavy portfolio and a portfolio invested entirely in stocks may have different jobs. One benchmark can describe their return differences without necessarily evaluating how well each met its own objective.
If C had promised in advance to keep pace with the stock market over time, its lag would need explanation. But the objective, investment universe, and risk constraints should guide the comparison before the outcome is known. Replacing the benchmark afterward with an easier one would change the test.
Dates, currency, dividends, and costs need alignment too. Comparing an account's dividend-inclusive return with an index's price change mixes definitions. Comparing different years or deducting strategy costs without describing benchmark costs creates further ambiguity. State those terms before interpreting the gap. [1]
How much market movement did the portfolio take on?
A portfolio that rises more in a rising market may be more sensitive to that market. Beta describes linear sensitivity to market returns in a simple market model. It is normally estimated from a return series. A higher beta may help explain outperformance in an up market and can also imply greater losses when the market falls.
Within that model, beta of 1 indicates market-level sensitivity; 0.5 indicates less; a negative beta indicates exposure in the opposite direction. Beta of 0 means no exposure to this particular linear market relationship, while the asset could still fluctuate substantially. Beta is neither a fixed daily multiplier nor a measure of every risk. It also places no upper bound on a loss. [2]
Now give B an assumed beta of 1.5. The risk-free return for the same period is assumed to be 2%, and the market's total return is 10%. The market reference here is the earlier teaching benchmark, with the same period, dollar currency, dividend treatment, and cost basis. This setup does not permit inserting any convenient index into the calculation. The risk-free return must also match the currency and period; the label does not guarantee purchasing power.
Start with the market return above the risk-free return: 10% minus 2% is eight percentage points. Multiply by beta of 1.5 to obtain twelve percentage points for the market-exposure term. Add back the 2% risk-free return to get a 14% model reference return.
B's actual net return was 16%. Subtract the 14% reference and two percentage points remain. Some of its six-point lead over the benchmark is accounted for by the assumed market exposure. Beta of 1.5 was an input to this exercise; it was not estimated from this single annual observation.
The calculation illustrates a one-factor market model. It is not an accurate forecast of what a real account must earn. Whether the model describes the portfolio adequately is another question to investigate.
Why the remaining two points do not establish skill
In a common market model, return above the risk-free rate is separated into alpha, a market-exposure term, and a period-specific error. Alpha is the model's intercept: an estimated average return beyond the listed risk explanation. Estimating it requires observations across periods and leaves uncertainty.
Some materials also use “alpha” for a simple benchmark subtraction. In this book, a claim about risk-adjusted alpha needs a stated model and return basis. In B's one-period calculation, the two-point remainder combines alpha and that period's error. One observation cannot separate them or establish stable alpha.
Error is not necessarily pure luck. It can include omitted influences, changing exposures, and measurement problems. An unexplained return does not automatically become a riskless benefit just because the model omitted its source. [3]
For your evidence card, keep the two calculations distinct: six percentage points of active return and two percentage points left after the assumed model reference. Renaming the second number does not produce the additional observations needed to estimate persistence or skill.
The market is not the only shared influence
B's record says it favors small companies and stocks with recent strong performance. Portfolios with similar market sensitivity can still differ because of other shared characteristics. Factor research tries to describe some of these systematic relationships through explicit definitions and data.
Whether the small-company exposure explains B's result requires holdings, returns across periods, and a model check. A style label alone establishes neither the source of the return nor its likely persistence. If adding a defensible style explanation substantially reduces estimated alpha, that reveals the conclusion's dependence on the model.
Adding more variables does not automatically improve the explanation. A model can fit the observed sample too closely, miss important changes, or incorporate information available only later. Choosing the model also needs rules. Continuing to try alternatives until one produces the desired answer does not resolve those concerns.
People may accept certain risks because they expect compensation. The expectation does not ensure that compensation arrives every year. Similarly, market efficiency does not mean that every price is always correct. When many participants study public information and incorporate it into prices, an easily recognized opportunity can face competition. Competition alone does not establish that all active research is useless; finding a plausible explanation does not establish that prices have overlooked it. [4]
Costs and luck can change the ranking
B's gross return was 18% and its net return 16%. The two-point gap is return drag in this synthetic example. It does not say that a broker or manager charged a fee rate of precisely 2%. Real costs can depend on trading frequency, the spread between buying and selling quotes, and execution prices relative to a reference price. Fee bases and payment timing can differ as well.
Gross returns can help examine a strategy's mechanism. Net returns show what remains after the stated investment costs. You can examine both, provided the strategy and benchmark conventions are explicit. Choosing whichever basis makes one record look better would undermine the comparison.
Short-term winners can also arise when many people try different rules, even if none has a persistent advantage. If only successful attempts are shown, the record is incomplete. Skill-versus-luck research requires a defined sample and statistical method; a single winning account cannot substitute for that analysis. [5]
When the evidence is incomplete, keep the skill claim unresolved. The next question is how much support it receives under a stated benchmark, model, cost treatment, and sample. The table gives a starting point for that investigation, rather than permission to treat its ranking as established ability.
Rewrite B's record as an evidence card
A defensible description is: “Under the supplied synthetic annual and cost assumptions, B earned a 16% net total return, six percentage points above the teaching benchmark. Its small-company and recent-winner tilts may involve different risk exposures. The one-period calculation with an assumed beta leaves two percentage points unseparated, which cannot establish stable alpha or skill.”
The next evidence would include returns and holdings across periods, the benchmark and rules chosen in advance, cost calculations, the complete record of attempts, and performance under different market conditions. Until those materials have been checked, this card alone is insufficient grounds for increasing the money committed.
History you have already analyzed cannot be presented again as fresh test material. Keep track of what has been inspected and which choices it influenced. To evaluate the record, check the benchmark, return definitions, shared risk exposures, and evidence of persistence, then request the missing item rather than replace it with a stronger adjective.
Try the calculations in a falling market
Write your answers before reading the answer points.
- Why is B's six-percentage-point lead not, by itself, risk-adjusted alpha?
- Using the supplied 2% risk-free return, 10% market return, and beta of 1.5, calculate the model reference. What does the difference between 16% and that reference represent?
- In a separate synthetic scenario D, the portfolio returns −8% and the benchmark −6%. Both cover the same complete year in dollars, include dividends, and are net of teaching costs, excluding personal taxes. The benchmark is also the model's market reference. Assume beta remains 1.5 and the risk-free return 2%. How much did the portfolio lose, and how far did it lag? Calculate the model reference and remaining difference. Can you infer skill?
Reference answers · reveal
Answer points: First, six percentage points is active return relative to the given benchmark. Risk adjustment requires a model and information about exposures, costs, the sample, and estimation uncertainty.
Second, the reference is 14%: the risk-free term plus the market-exposure term. The remaining two percentage points combine alpha and the period's error. The single observation does not separate them.
Third, the portfolio lost 8% and lagged its benchmark by two percentage points. The reference is 2% plus 1.5 times the difference between −6% and 2%, giving −10%. Actual return of −8% exceeds that reference by two percentage points. Again, that remainder combines alpha and period-specific error; it does not prove skill. A positive remainder, a profitable account, and benchmark outperformance are three separate results.
Save your performance evidence card
Turn your answer for D into four lines: net return; benchmark and active return; model assumptions and remainder; and what remains unproven with the evidence needed next. Use the downward scenario, so the card tests your reasoning beyond the rising-market demonstration.
Check that you separated losing money from lagging a benchmark, labeled beta as assumed, and avoided describing the single-period remainder as estimated alpha. Preserve your first answer and the reasons for any corrections. Copying the correct numbers alone would not show which distinction you needed to repair.
Evaluate the result, the risk, and the evidence
First align periods, currencies, dividends, and costs, then compare with a benchmark chosen in advance. A positive return and an outperforming return answer different questions.
Next examine market sensitivity and other common exposures. Beta describes one model relationship, so low beta need not mean low total risk. Finally, assess the evidence for persistence. Scenario D combines a loss, underperformance, and a positive model remainder. To move from that observation to a claim of ability requires records across periods, complete attempts, and attention to model limits.
References
- CFA Institute: Portfolio Performance Evaluation, 2026 curriculum. The public introduction and summary were used; the paid curriculum was not reviewed.
- Nasdaq: Beta. Supports market sensitivity, estimation limitations, and the distinction from total volatility.
- Laurence B. Siegel: Benchmarks and Investment Management, 2003, printed pages 11–13. Supports the distinction between simple subtraction and regression-based alpha and model interpretation.
- CFA Institute: Market Efficiency, 2026 curriculum. Only the public introduction and summary were reviewed, for information in prices and conditions affecting efficiency.
- Eugene F. Fama and Kenneth R. French: Luck versus Skill in the Cross-Section of Mutual Fund Returns, 2010. Used for the need to distinguish skill from luck with explicit samples and methods, without extending its findings to all investors.
Editorial status
AI-assisted draft, synchronized to the accepted second-round Chinese text. The cards and beta inputs are original teaching assumptions; see the shared synthetic case materials and author review. A separate AI reviewer has completed an independent full-text review. Human financial review and target-reader testing remain pending.