I have been studying investors lately. Reading how they think, how they bet, how they hold. One pattern keeps surfacing.
The best ones think evolutionarily, whether they say so or not.
Dalio is the most explicit about it. The frame I keep coming back to is r vs. K selection.
Y Combinator funds 200+ companies each batch. Most won’t reach profitability. A handful hit escape velocity. The expectation is a power law: a few investments return everything and more, so the rest can go to zero.
That’s r-selection. In biology, r-strategists maximize offspring count at low investment per individual. Rabbits. Bacteria. Most won’t survive, but enough will to carry the line forward.
Warren Buffett runs the opposite system. Berkshire Hathaway holds roughly 50 positions, but five carry most of the value. He buys companies he plans to hold forever. The filter is strict. The conviction is deep. Each position earns its slot through a rigorous process before capital moves.
That’s K-selection. K-strategists produce few offspring and invest heavily in each one. Elephants. Great apes. High investment, high survival rate, high ceiling per individual.
The instinct is to pick a side. Both are adapted to different environments.
YC works because the market is the unknown. Even great teams lose when the market is wrong. The only way to find which markets win is to fund broadly and let results decide. The batch is the filter.
Buffett works because the signal is strong. He operates on decades of pattern recognition. The companies he evaluates have:
- track records
- moats
- management teams
Narrow selection is rational when input quality is high.
The strategy follows the environment.
r-selection when signal is weak and options are cheap. K-selection when signal is strong and depth is what creates the edge.
I run my own work this way. I start many small experiments at once. The ones that show results get my full attention.
Each makes the other possible.
The failure mode is applying r-selection to a situation that needs K, or K to a situation that needs r. Running hundreds of experiments when you already know what works. Or picking one bet before the data tells you which one.
The question is which environment you’re in.
🧬
Based on your current environment, which strategy should you run?
