I Do Not Predict the Market
Because I am a developer, acquaintances often ask me questions like which crypto to buy or if it is too late to enter Nvidia. My answer is always the same.
I do not know, and I do not try to know. I do not watch charts to guess tomorrow’s fluctuation or watch macroeconomic news to predict interest rates. I am an engineer and not a prophet.
To a developer, the market is a non-stationary system where the statistical properties of its time-series data change constantly. Because yesterday’s rules break today in this chaotic system, past data is just a map and never an immutable law. Attempting to predict this with human intuition is pure arrogance. I did not build a prediction model. Instead, I designed a controller. A system must not capsize on unpredictable waves but must restore its own balance. This is the core of my investment philosophy.
Before designing the system, I had to define the biggest risk factor. The most fatal bug is actually the investor’s own brain rather than a market crash or inflation. The human brain overweights recent and intense data. When stocks rise, we believe they will rise forever and click the buy button at the peak. When they crash, we believe the world is ending and click the sell button at the bottom. Data science calls this phenomenon overfitting. A model fitted too tightly to local patterns shatters when it faces new conditions. Intuitive investing driven by the amygdala is an algorithm destined to fail.
I removed my own judgment from the system’s variables by relying only on two mathematically grounded principles called covariance and negative feedback. My portfolio is like a department store holding over 20 asset classes including Korean stocks, U.S. stocks, bonds, commodities, gold, the dollar, and cash. This is not simply about following the classic proverb of not putting all eggs in one basket. It is about covariance, representing the correlation structure between assets. I do not look at returns first. I check if they crash together.
When stocks crash, safe assets like the dollar and government bonds often rise. When currency values drop, real assets like gold and commodities defend value. Combining assets with different movement vectors offsets volatility across the entire portfolio. Of course, this correlation structure itself is not a constant. In extreme phases like a financial crisis, correlations spike and everything collapses together. Thus, I view diversification not as an omnipotent shield but as a safety mechanism. It converts catastrophic failure into manageable damage by eliminating the Single Point of Failure.
Once the structure is built, an engine must run the system. While many people view rebalancing as a simple portfolio check-up, to a developer, it is a negative feedback loop that maintains a set point. Let us assume a 50 to 50 split between stocks and bonds. If the market surges where stocks hit 60 and bonds drop to 40, an error occurs. While the crowd yells to buy more, my system coldly triggers a routine. It sells stocks and buys bonds to restore the set point.
Conversely, if stocks fall to 40, the system does not panic or sell at a loss. It sells safe assets that have reached 60 to buy cheap stocks. This mechanical act forces me to sell high and buy low without emotion through what financial engineering calls volatility harvesting. Just like Shannon’s Demon where prices fluctuate with mean reversion without a clear direction, the system turns that turbulence into energy.
This effect is not magic. It works best under specific conditions where asset correlations are low and transaction costs are small. It requires management rather than blind faith. Theory also differs from reality since a model that was perfect in back-testing often creaks in production. Friction coefficients like costs and taxes drag it down. Because frequent rebalancing erodes excess returns through transaction fees and taxes, I apply band-based rules. I intervene only when the allocation deviates by more than 5 percent from the target weight. I also minimize tax friction by using cash flow from dividends and earned income to fill the underweight assets. I avoid selling assets to balance the ratio whenever possible.
Some ask if this complex diversification makes the returns too boring. The answer is yes. This system is indeed boring. I do not know the thrill of doubling my assets overnight. Conversely, I also do not know the terror of losing half my assets overnight and suffering from insomnia. My objective function in investing is not maximum return. It is optimizing drawdown control and sleep quality under the constraints of taxes and costs.
To a developer, a good system is not an unstable racing car boasting top speed. It is a highly available system operating like a server running 24 hours a day and 365 days a year through rain or snow. I do not try to beat the market. I simply control the system so it does not collapse regardless of where the market bounces. If the market goes up, it is good because my asset values rise. If it goes down, it is good because I can buy more at a lower price.
I will sleep soundly again tonight. Tomorrow the Nasdaq may soar or crash. My system will balance itself upon those waves.