After discussing transaction-price regularities and weekday effects, the author explicitly says that reported conclusions may be too small relative to costs. He also notes that a significant result in one period can disappear in another.
Keep the historical tense
“The data showed an effect” is different from “the effect works now.” Market participants adapt, and the mix of conditions can change. Any extension beyond the book’s period needs new evidence; this companion has not supplied that evidence.
Worked example
A fictional weekday effect averages 0.10% before costs. A 0.15% cost leaves a negative net average, even if the historical difference is real.
Case connection
Results depend on the period. A decade-long comparison is not evidence for an identical result in every shorter window.
Buffett’s ten-year fund wager
Source-grounded facts
Buffett’s 2017 letter reports that the S&P 500 index fund beat each of five funds-of-funds over the wager.
Context
Buffett and Protégé Partners compared an S&P 500 index fund with five funds-of-funds over ten years, from 2008 through 2017. The comparison included the returns investors retained after fees.
Outcome
One comparison fund was liquidated in 2017, a fact noted in the table. The wager illustrates costs and evaluation periods, not a universal result for every manager or decade.
- All five funds-of-funds outperformed the index fund in the difficult first year, 2008.
- The comparison continued through the agreed decade rather than stopping after that initial result.
- Buffett’s final table reported a 125.8% gain for the index fund; none of the five funds-of-funds matched it.
Case analysis
Changing the evaluation window can change the apparent winner. Compare the first-year result with the agreed full period to see why the horizon belongs in the definition of a claim. For a calendar effect, ask separately whether it exists in one sample, persists elsewhere and survives costs. A favourable period does not answer all three questions.
Try it
Write three separate questions: did the effect exist in the sample, did it survive another period, and was it large enough after costs?