Buffett once said investing has two iron rules: first, never lose money; second, never forget the first rule.
Previously I treated this saying as mere “mindset building”, thinking it only emphasized risk control. But after deeply studying the Log-normal Distribution model, I discovered this saying actually has an extremely cold and rigorous mathematical foundation.
For intuitive understanding I wrote an interactive visualization tool; I suggest reading it alongside this article: 👉 The Truth of Compounding: Log-normal Distribution Lab
1. Starting from the Normal Distribution
Before entering complex financial models, we must first understand nature’s cornerstone — the normal distribution.
The Galton Board: Addition Visualized
Imagine a board covered with staggered pegs; balls drop from the top, and each time a ball hits a peg it randomly bounces left or right. This physical process is essentially the accumulation of random steps. With enough balls, the pile shape at the bottom inevitably presents the perfect “bell curve”.
The Central Limit Theorem (CLT)
This is mathematics’ Central Limit Theorem: large numbers of mutually independent, randomly distributed variables — as long as they are related by summation — will inevitably tend toward a normal distribution.
In the world of addition, gains and losses are symmetric. If you have 100 yuan, add 10 today and subtract 10 tomorrow, you still have 100 yuan.
2. The Model of Investment Returns: The Multiplication Trap
However, the investment world follows not addition but multiplication.
Summation vs Product
Your wealth growth is not $P_0 + r_1 + r_2$, but: $$P_t = P_0 \times (1 + r_1) \times (1 + r_2) \times \dots \times (1 + r_t)$$
This multiplicative random process leads to a fatal asymmetry: if you lose 50%, you need to gain 100% to get back to even.
The Log Transform: Bridging Two Worlds
To use the powerful Central Limit Theorem, mathematicians took the logarithm ($\ln$) of prices: $$\ln(P_t) = \ln(P_0) + \ln(1+r_1) + \ln(1+r_2) + \dots$$
Now the right side of the equation becomes the summation of log returns. By the Central Limit Theorem, $\ln(P_t)$ follows a normal distribution. Derived conclusion: if the logarithm of asset prices is normally distributed, then the asset prices themselves follow a “log-normal distribution”.
3. Graphical Traits of the Log-normal Distribution
Observing the log-normal distribution’s probability density curve, we find it completely different from the normal distribution:
- Asymmetry (right skew): it has a long tail extending right (theoretically infinite upside), but the left side is firmly stuck at 0 (you can lose at most everything).
- Center of gravity shifted left: this is the article’s most core observation.
In the log-normal distribution there are three “centers”:
- Mean: the mathematical average return.
- Median: the outcome 50% of people can reach.
- Mode: the most probable, most likely outcome.
4. Deep Properties: Why “Don’t Lose Money” Is the First Law
By observing the interaction of model parameters $\mu$ (expected return) and $\sigma$ (volatility), we can draw the following practical insights:
4.1 The Volatility Trap: Losses Are Compounding’s Killer
Under the log-normal distribution, the compound annual growth rate $g$ is approximately: $$g \approx \mu - \frac{\sigma^2}{2}$$
See that $\frac{\sigma^2}{2}$? Volatility directly “deducts money” from your returns. Even if the average return $\mu$ is high, as long as volatility $\sigma$ is large enough, your real growth $g$ may be negative. This is the mathematical reason why “steadiness” matters more than “explosiveness”.
4.2 Don’t Use Leverage: The Collapse of the Median
Leverage increases $\mu$ linearly but increases $\sigma^2$ at square speed. Reflected in the graph: although the Mean is pulled rightward by the tiny minority of get-rich survivors, the curve’s peak (Mode) and midpoint (Median) rapidly collapse leftward (into the loss region). Conclusion: the higher the leverage, the greater your probability of becoming that “averaged-away” cannon fodder.
4.3 The Illusion of Chasing Highs: The Price of High Volatility
Hot stocks (high $\mu$, extremely high $\sigma$) often present extreme left skew. Although they “look” like they rise fast, their Mode (the most likely outcome) is often far below 1.0. What you chase is that sparse, extremely hard-to-catch “long right tail”, while what you face is the high-probability loss peak.
4.4 Principal Must Be Sufficient
In the multiplicative model, all returns are based on $P_0$. Because of volatility drag, the probability of small principal doubling through high-risk volatility is far below the certainty of large principal growing through low volatility. Principal is not only capital but also risk-resisting “redundancy”.
4.5 There’s Always a Chance to Win: Time as the Antidote
As long as your compound growth rate $g > 0$, as time $t$ increases:
- The growth term increases linearly with $t$.
- The random fluctuation term increases slowly with $\sqrt{t}$. As long as you’re not cleared out by big losses or leverage, time will eventually dilute short-term randomness, letting the wealth distribution’s peak slowly climb past the “break-even point”.
5. Summary
I may have made some imprecise assumptions in building this model, but it reveals a profound truth: Investing is not a game of who runs fastest, but a game of who survives longest.
“Don’t lose money” is not about pursuing conservatism, but about protecting that fragile compounding engine from being forcibly stalled by the demon $\sigma^2$. In this asymmetric probability world, mediocre wins + very few losses = great compounding.
If you’re interested in how these parameters affect the distribution, welcome to my visualization tool to adjust the parameters yourself — you’ll develop a whole new reverence for “risk”.