Contrary to popular belief, earnings growth has a very low correlation with decadal stock returns. The primary driver is the change in the valuation multiple (e.g., P/E ratio expansion or contraction). The correlation between 10-year real returns and 10-year valuation changes is a staggering 0.9, while it is tiny for earnings growth.

Mathematical models like the Kelly Criterion are only as good as their inputs. Historical data, such as a stock market's return, isn't a fixed 'true' value but rather one random outcome from a distribution of possibilities. Using this single data point as a precise input leads to overconfidence and overallocation of capital.

Despite the confidence with which they are presented, annual stock market predictions from major investment banks are notoriously unreliable. Data from 2003-2023 shows the median forecast was off by 14 percentage points, highlighting the futility of trying to precisely time the market based on expert commentary.

Investors often judge investments over three to five years, a statistically meaningless timeframe. Academic research suggests it requires approximately 64 years of performance data to know with confidence whether an active manager's outperformance is due to genuine skill (alpha) or simply luck, highlighting the folly of short-term evaluation.

Investors often expect an average 8-10% annual return from stocks. However, historical data shows the most common yearly outcomes are monster returns of +15-20%, with +20-35% returns also being frequent. This demonstrates that market performance is characterized by periods of extreme gains, not steady, average growth, a concept investor Ken Fisher termed "normal market returns are extreme."

Historical analysis of investors like Ben Graham and Charlie Munger reveals a consistent pattern: significant, multi-year periods of lagging the market are not an anomaly but a necessary part of a successful long-term strategy. This reality demands structuring your firm and mindset for inevitable pain.

Contrary to a common myth, high equity valuations do not reliably revert to a historical mean. An analysis of 32 different valuation scenarios found only one case of statistically significant mean reversion. Structural economic shifts, like reduced GDP volatility since the 1990s, justify higher sustained valuation levels.

Across 200 years and 56 countries, the single most important factor for long-term investing success is the starting valuation. Buying portfolios with low P/E ratios or high dividend yields consistently outperforms buying expensive assets by 3-4% annually over the long run.

Academic studies show that company growth rates do not persist over time. A company's past high growth is not a reliable indicator of future high growth. The best statistical prediction for any company's long-term growth is simply the average (i.e., GDP growth), undermining most growth-based stock picking.