Position Sizing and the Death of Conviction
Most investors think the hardest part of their job is picking the right companies. It isn't. The hardest part is knowing how much to own when you finally find one, and then staying still long enough for the thesis to play out. I have watched people blow up accounts on stocks that were fundamentally correct the entire time. They just held too big a position and panicked at the first real downturn. The single most important uncommon sense principle for anyone who actually wants to build wealth through investing is this: your position size should be inversely proportional to the degree of your uncertainty, not proportional to your excitement. This sounds simple until you try it in practice, because human beings are not built to reduce position sizes when they feel most confident. That is exactly when conviction peaks and it is exactly when you want to go all in. The mechanism that prevents you from doing this is what I will walk through below.
The Most Important Thing Uncommon Sense For The Thoughtful Investor
The framework is called Kelly Criterion sizing, adapted for public equities and ETFs. It originated in gambling theory in the 1950s but it was brought into investment circles by people like Ed Thorp and later Paul Samuelson. The original formula is: f* = (bp - q) / b Where f* is the fraction of your capital to bet, b is the net odds received, p is the probability of winning, and q is the probability of losing. In investable terms this translates roughly to comparing your estimated edge against your estimated variance. The full math gets complicated fast when you are dealing with correlated stocks or multi-factor models.
Here is what actually works in practice. I stopped using the raw Kelly formula around 2012 because it produces numbers that are too aggressive for any real portfolio. A half-Kelly or quarter-Kelly approach is far more realistic. You estimate your expected return on a position, estimate the standard deviation of returns, and divide expected return by the square of standard deviation. That gives you a fractional Kelly number. Then you reduce it again by half as a safety margin. The actual workflow looks like this. I write down three numbers for every position before I enter: the bull case return estimate, the base case return estimate, and the bear case return estimate. I assign rough probabilities to each scenario. From those three numbers I calculate an expected value and an approximate standard deviation. Then I apply the quarter-Kelly adjustment. The resulting percentage is my starting position size. This process usually takes about twenty minutes per position if you are doing it properly. Most retail investors spend two hours researching a stock and zero minutes thinking about how large their position should be. That is backwards. The research tells you whether to buy. The sizing tells you whether you will survive long enough to profit.
Get the Full Details
I ran into a specific problem a few years ago that I had not anticipated. I was running a concentrated portfolio of about fifteen positions, and the quarter-Kelly model was telling me to hold a 7.3% stake in a biotech stock. The stock dropped 18% in a single day on a failed clinical trial readout. My model had assumed a standard deviation based on five years of historical volatility, which completely missed the binary event risk inherent in that sector. I lost about 1.3% of my total portfolio on that one trade, which would have been manageable. But the real damage came from the psychological response. I had to sell the position the next day because the thesis was broken, and I locked in a loss at a terrible moment. The model was not wrong about sizing for normal market conditions. It was wrong about the probability space I was operating in. The workaround was straightforward but annoying. I added a maximum single-position cap of 10% regardless of what the model said. For binary-event-heavy sectors like biotech and certain commodity names, I reduced the effective position size by an additional 40% on top of the quarter-Kelly number. That cut my biotech allocation from 7.3% down to about 4.4%, which meant the same 18% drop only cost me 0.2% of total portfolio value instead of 1.3%. It felt like leaving money on the table at the time. It saved me from making an emotional decision under stress. There are legitimate downsides to this approach that nobody talks about much. The first is that it requires you to honestly quantify your uncertainty, which most people cannot do. You will look at a stock and think you have a 70% probability of a bull outcome when you probably have more like 40%. The model will give you garbage in and garbage out unless you calibrate your probability estimates against actual outcomes over time. I keep a spreadsheet where I track my predicted scenario probabilities against what actually happened. It took me about eighteen months of consistent use before my calibration improved to the point where the model was useful. Before that it was worse than guessing.
The second downside is opportunity cost. A quarter-Kelly framework will systematically underweight high-conviction opportunities compared to a concentrated approach. If you are right about a massive compounder and the model tells you to own 5% when you could have owned 15%, you will significantly underperform during bull markets. This is by design. You are trading upside for survival. The math works in your favor over time because the people who take the bigger positions are the ones who eventually go bust from sequence of returns risk or emotional breakage. The third downside is that this approach assumes you can estimate expected returns and standard deviations. For broad index funds this is trivial. For individual stocks in efficient markets it is extremely difficult. The better your research process is, the more meaningful the output of the model becomes. If your stock picking edge is essentially noise, you should just buy a low-cost index fund and forget about this entirely. Position sizing models add complexity without adding value when you do not have a real informational or analytical edge over the market. For most thoughtful investors the practical takeaway is simpler than the full framework. Here is the distilled version: pick a position size that would let you hold the stock through a 40% decline without selling, without stress, and without changing your thesis. If you cannot sleep through that kind of move, you own too much. This rule of thumb captures about eighty percent of the benefit of full Kelly sizing with maybe five percent of the effort.
The broader point is that investing is mostly a risk management problem dressed up as a stock picking problem. You can be right about the stock and still lose money if your position size breaks your ability to stay rational. The uncomfortable truth is that the market pays you for bearing uncertainty, not for expressing confidence. Position sizing is how you translate that understanding into actual portfolio behavior. Everything else is secondary.
