Position Sizing for Stock Pickers: A Conviction Framework

Most position sizing advice for individual investors stops at "don't put more than 5% in any one stock." That's a guardrail, not a framework. It tells you what to avoid but nothing about what to actually do when you've done the work and believe one idea is twice as good as another.

If you're picking individual stocks, you've already decided diversification alone isn't enough — you want to express views. Position sizing is how those views show up in your returns. Get it right and your best ideas drive the portfolio. Get it wrong and your worst idea gets the same weight as your best, which means you're closet-indexing with extra steps.

Here's a framework built around three inputs: conviction, asymmetry, and correlation.

Input 1: Conviction (and how to actually measure it)

Conviction is not enthusiasm. It's the inverse of how much new information would change your mind. A high-conviction position is one where you've stress-tested the bear case, mapped what would have to be true for you to be wrong, and concluded the probability of being wrong is low.

A simple test: write down three things that, if they happened over the next twelve months, would make you sell. If you can't write them, your conviction is actually a vibe. If the three things are vague ("if the thesis breaks"), same problem.

A rough conviction tier system:

Tier 1 (high): You can articulate the bear case better than most bears. You've owned it through at least one drawdown without flinching. The thesis is structural, not a catalyst trade.
Tier 2 (medium): You like the setup, the numbers work, but the thesis depends on something you can't fully verify — management execution, a product cycle, a macro tailwind.
Tier 3 (low): You're interested. Maybe a starter position to force yourself to follow it.

Most retail portfolios have positions across all tiers. That's the problem — too many Tier-2 positions without enough Tier-1 conviction to drive outsized returns.

Input 2: Asymmetry — the Kelly-lite version

Full Kelly criterion math is overkill for most stock pickers because we can't estimate edge to three decimal places. But the intuition matters: position size should scale with the ratio of upside to downside, weighted by the probability of each.

A simplified version: for each position, estimate three scenarios over your holding period (call it 2-3 years):

Base case return — what you actually expect
Bear case return — what happens if you're wrong, not a rounding error wrong but thesis-broken wrong
Bull case return — what happens if you're right and the market notices

If base is +30%, bear is -25%, and bull is +80%, that's a respectable setup. If base is +15%, bear is -40%, and bull is +25%, that's a bad position regardless of how exciting the story is. You want positions where the bear case is bounded and the bull case is uncapped.

Two Tier-1 convictions can have very different sizes if their asymmetry differs. A high-conviction defensive compounder with limited upside gets sized smaller than a high-conviction compounder with optionality.

Input 3: Correlation — what you actually own

Owning seven semiconductor names is owning one position. Owning Visa and Mastercard is one position. Owning four "AI infrastructure" plays moves together on the same days as a single cluster.

Before sizing a new name, ask: what's already in the book that this resembles? Same end market, same customer base, same macro sensitivity (rates, oil, China), same factor exposure (high-growth unprofitable tech all behaves like one stock during regime changes).

A practical rule: treat a cluster of correlated names as a single position for sizing purposes. If your "max position" is 8%, and you own three names that all benefit from the same data center capex cycle, that cluster shouldn't exceed 8% combined — not 8% each.

Putting it together: a sizing grid

Here's a starting grid for a concentrated individual-stock portfolio (15-25 names):

Tier 1, strong asymmetry, low correlation to rest of book: 6-10%
Tier 1, moderate asymmetry: 4-6%
Tier 2, strong asymmetry: 3-5%
Tier 2, moderate asymmetry: 2-3%
Tier 3 (starters): 1-2%

The top 5 positions should be roughly half the book. If your top idea and your tenth idea are sized the same, you don't really have a top idea — you have a watchlist with money in it.

One more rule: let winners grow, but cap them. A position that runs to 15% on appreciation has earned its size, but it's also now driving your portfolio more than your analytical edge probably justifies. Trim back to your sizing-grid level, not to some arbitrary "take profits" number.

What to watch next

Audit your current book. List every position with a conviction tier and a base/bear/bull return estimate. Positions you can't tier are positions you don't understand. See how to approach this methodically by separating process error from outcome error.
Find your correlation clusters. Group holdings by what actually drives them. Reduce overlaps where the combined cluster exceeds your single-position limit.
Re-rank quarterly, not daily. Conviction tiers should update with new information (earnings, management changes, thesis violations), not with price action.
Track the gap between conviction and sizing. If your highest-conviction names aren't your largest positions, that's the most fixable mistake in the portfolio.