Before the 1929-1943 bear market came the 1920s 'Roaring Twenties' where stocks were described as at an 'all-time permanent Plateau' by economist Irving Fisher, showing how investors justify expensive valuations before major crashes.

Value is currently trading near the 90th percentile of cheapness historically, which is not a guarantee of outperformance but provides better odds than alternatives based on the principle that valuations predict future returns with some statistical reliability.

If an investor listened to Jeremy Grantham's 2013 warning about market overvaluation and became anxious due to confirmation bias about threats like high P/E ratios, they would have missed one of the greatest decades of stock returns in history.

The strategy you must get right is diversification across many unique, persistent, pervasive, and implementable risk premiums that survive transaction costs, with logical reasons for persistence.

An investor who signs an Investment Policy Statement promising to rebalance according to a plan should be reminded by holding a 'revolver' and saying 'Go ahead and make my day' if they consider abandoning the plan, using Dirty Harry's famous line to emphasize the commitment required.

Vanguard's market return projections have ranges that are too narrow—they might say U.S. stocks should return 5-7% when the true range should be 0-10%, understating the actual uncertainty investors face.

Over any 20-year period, the outperformance of value stocks over growth stocks is probably only 5-6% maximum, but in any single year the difference can be 20-40%, so longer time horizons make diversification less important from a performance perspective.

Buffett's advice to stay invested and ignore all information—just follow the course—is the correct strategy if you had perfect foresight about the future, because specific outcomes are unpredictable even when conditions are known.

Confirmation bias causes investors to embrace forecasts that confirm their preconceived notions as 'brilliant' while dismissing contrary forecasts as worthless, making forecast-driven investing particularly dangerous.

At the end of 2023, 20 Wall Street analysts predicted S&P 500 returns ranging from -12% to +13% with an average of +2%, but the S&P 500 actually returned +23-25%, making the average forecast off by 21 percentage points—more than double historical average returns.

In the last seven years, the consensus forecast for S&P 500 returns has been off by at least 14% every year, representing the minimum error across the entire spectrum.

Jeremy Grantham, despite being celebrated for forecasting the 2000 and 2008 crashes, warned in 2013 that the market was 75% overvalued and predicted the S&P would lose almost 1% annually for the next 7 years, but it actually gained almost 11% annually—an error of 12 percentage points—yet he remained influential.

Research after Schiller's 10-year CAPE showed that 5-year, 6-year, and 8-year cyclically-adjusted PE ratios have roughly the same ~40% correlation to future returns as the 10-year version, suggesting the exact averaging period matters less than previously thought.

Experts making medical diagnoses based on MRI scans who report 90% confidence are correct only about 65% of the time on average, while AI makes far fewer mistakes in the same task, demonstrating that overconfidence is systematic.

Valuations measured by current P/E or Shiller CAPE 10 ratios show only ~0.4 correlation with future 10-year returns, meaning valuations have nearly zero predictive power for one-year returns and cannot be used to time markets even over decades.

The proper approach to forecasting is to treat any forecast as the median of a wide dispersion of potential outcomes and run Monte Carlo simulations to test whether a portfolio can withstand all possible outcomes, including tail risks, rather than relying on a single median forecast.

In 2008, a client's portfolio that included worst-case scenarios in its planning (such as a 2008-style financial crisis) allowed them to be well-prepared psychologically and to have discussed Plan B options (cutting expenses, working longer, selling the second home) before crisis hit, which is why Monte Carlo simulation planning is essential.

Investing is fundamentally different from rolling dice because we do not know the actual odds of future returns; we can only estimate future returns and consider a wide dispersion of outcomes given major uncertainty about risk premiums and geopolitical/economic regime changes that nobody can forecast.

The Schiller CAPE ratio was developed based on Benjamin Graham and David Dodd's principle that you shouldn't look at only one year of earnings because the economy is cyclical—you might be at the end of a boom (high earnings) or end of a recession (low earnings), so multi-year averaging is necessary.

Your investment time horizon must be at least 10 years if you want to invest in risk assets, because only at that horizon can you be confident that stocks will outperform risk-free treasury bills; otherwise, you don't have certainty of positive returns and should not take equity risk.