When evaluating others' success, ask if their strategy would work for most people who adopt it, or if it relied heavily on luck. If a strategy isn't reproducible and leaves many casualties behind, it's not a model to be learned from, regardless of the impressive outlier outcome.
People justify high-risk strategies by retroactively fitting themselves into a successful subgroup (e.g., 'Yes, most investors fail, but *smart* ones succeed, and I am smart'). This is 'hindsight gerrymandering'—using a trait like 'smartness,' which can only be proven after the fact, to create a biased sample and rationalize the risk.
In domains with extreme outcomes (music, startups), success is heavily influenced by luck, making it difficult to replicate. A more effective strategy is to study the common failure modes of the vast majority of talented people who tried. This provides a clearer roadmap of what to avoid than trying to copy a lucky winner.
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.
The fastest skier isn't who wins, but the fastest among those who finish. Irreversible losses, like a career-ending injury, eliminate all future opportunities for gains. Therefore, over the long term, ensuring survival is mathematically more important than maximizing short-term performance, a concept known as ergodicity.
Maximizing daily output does not maximize yearly output. Long-term success requires investing in activities like building trust, relationships, or skills, which often yield no immediate returns and may seem inefficient day-to-day. Consistently choosing short-term tactics over long-term strategies ultimately limits growth.
For an event with a 1-in-N chance of happening, if you try N times, the probability of it occurring at least once is roughly 63%. While this highlights the danger of repeated low-probability risks, it also applies positively. Consistently performing small, beneficial actions can compound to make eventual success almost a mathematical certainty.
Instead of relying on population averages for risk (e.g., car accidents), monitor your own close calls and mistakes. These 'near misses' are latent data points that provide a much better personal estimate of your true risk profile and how long you can last before a critical failure occurs if habits don't change.