How to Size a Position: A Framework for Managing Portfolio Risk
Most investors spend 90% of their research time on stock selection — finding the right idea — and almost none on position sizing — deciding how much capital to commit. This is backwards. The quality of your idea determines your ceiling; position sizing determines your floor. A great idea sized at 1% barely moves your portfolio. A mediocre idea sized at 20% can permanently impair it.
Position sizing is the set of rules that translate conviction + risk into a capital weight. This guide explains the four main frameworks, the two constraints that override any formula, and a worked example you can apply immediately.
TL;DR
- Position sizing — not stock-picking — is the primary lever controlling portfolio risk and drawdown.
- Four frameworks exist: equal-weight, conviction-weighted, volatility-parity, and Kelly criterion. Each makes different tradeoffs.
- Two hard constraints override any framework: concentration limits (single stock, sector, theme) and correlation (how positions move together under stress).
- The "how much to add when wrong?" question has a right answer: pre-define your add-down rules before you enter.
- Volatility-parity is the most mechanical, Kelly is the most theoretically optimal but practically unworkable at full scale; a conviction-weighted approach with hard concentration caps is the practical sweet spot for most portfolios.
1. What position sizing actually does
A portfolio is a collection of bets. Each bet has three parameters:
| Parameter | What it is | Where most investors focus |
|---|---|---|
| Direction | Long or short | 95% of effort |
| Timing | When to enter/exit | 4% of effort |
| Size | How much capital | 1% of effort |
The allocation of effort is almost perfectly inverted relative to impact. Consider: if you're right about a stock doubling but you own 0.5% of your portfolio, you make 0.5%. If you're 60% right about a modest 15% gainer but you own 8%, you make 1.2%. Sizing is a return multiplier.
More critically, sizing is the primary risk control. Every portfolio has a maximum drawdown it can recover from within a reasonable time horizon. The Kelly criterion literature — and empirical studies of fund blow-ups — consistently shows that overleveraging or overconcentrating a single position is a more common path to ruin than picking bad ideas.
The fundamental sizing equation:
Position weight = (Conviction × Liquidity discount) / (Volatility × Correlation penalty)This isn't a precise formula yet — it becomes one in the frameworks below — but it captures the structure: higher conviction and more liquid positions deserve larger weights; higher volatility and higher correlation to the rest of the book deserve smaller weights.
2. The four sizing frameworks
Framework 1: Equal-weight
Every position receives the same capital allocation: 100% / N positions.
| Attribute | Detail |
|---|---|
| Formula | Weight = 1/N |
| Best for | Diversified index-like books, factor portfolios, screener-driven systems |
| Worst for | Conviction-driven concentrated books |
| Key property | Minimizes single-position risk; implicitly says "I have no information edge on sizing" |
Worked example: 20 positions → 5% each. If one goes to zero, you lose 5%.
Equal-weight is the correct default when you have no systematic way to rank conviction. It prevents the most common amateur mistake: outsizing a position because it "feels" like the best idea, when that feeling is often recency bias.
Framework 2: Conviction-weighted
Positions receive weight proportional to your assessed edge or conviction score.
| Conviction tier | Typical range | What qualifies |
|---|---|---|
| High | 8–12% | Deep proprietary work, variant view, long time horizon advantage |
| Medium | 4–7% | Standard research, reasonable variant view |
| Low | 1–3% | Early-stage, limited information, tracking position |
The rule: tiers must sum to 100%, and your highest single position should not exceed your hard concentration cap (see Section 4). Most practitioners cap a single position at 10–15% before the portfolio becomes idiosyncratic.
The risk with conviction-weighting is overconfidence. Research by Kahneman and others shows that subjective confidence scores correlate weakly with actual outcome accuracy. This is why conviction-weighting works best when conviction is grounded in something observable (variant fundamental view, time horizon mismatch, information asymmetry) — not gut feeling.
Framework 3: Volatility-parity
Each position is sized so that its volatility contribution to the portfolio is equal, regardless of capital weight.
| Step | Formula |
|---|---|
| Measure each position's annualized volatility | σᵢ = std(daily returns) × √252 |
| Compute the inverse-volatility weight | wᵢ = (1/σᵢ) / Σ(1/σⱼ) |
| Apply to capital | Capital₍ᵢ₎ = Total portfolio × wᵢ |
Worked example:
| Stock | Annual vol (σ) | Inverse vol | Volatility-parity weight |
|---|---|---|---|
| A | 20% | 5.00 | 37.0% |
| B | 30% | 3.33 | 24.7% |
| C | 40% | 2.50 | 18.5% |
| D | 25% | 4.00 | 29.6% → adjust to 100% |
Volatile positions get smaller weights. Less volatile positions get larger weights. The portfolio's ex-ante volatility contribution is equalized.
Volatility-parity is widely used in systematic multi-asset strategies (it's the core idea behind risk-parity funds). Its weakness: it sizes down your best ideas if they happen to be volatile growth stocks, and sizes up your most stable ideas even if their risk-adjusted expected return is lower.
Framework 4: Kelly criterion
The Kelly criterion is the mathematically optimal fraction of capital to wager on a bet to maximize the long-run geometric growth of the portfolio.
Full Kelly formula:
f* = (p × b - q) / bWhere:
f*= fraction of portfolio to betp= probability of winningq= 1 - p (probability of losing)b= net odds (payout per unit risked)
Simplified stock version:
f* = edge / odds = (Expected return - Risk-free rate) / VarianceThe problem: Full Kelly is extremely aggressive. It concentrates heavily in the highest edge ideas and leads to violent drawdowns during adverse runs. In practice, almost no one uses full Kelly. The standard approach is half-Kelly or quarter-Kelly — which gives up ~20–30% of the theoretical maximum growth rate but dramatically reduces volatility and drawdown.
| Kelly fraction | Max drawdown (historical estimate) | Practical use |
|---|---|---|
| Full Kelly (1.0×) | 50–80%+ | Theory only |
| Half Kelly (0.5×) | 25–40% | Aggressive professional books |
| Quarter Kelly (0.25×) | 15–20% | Conservative professional books |
Kelly is most useful as a sanity check: if your conviction sizing implies a full-Kelly fraction above 15%, you're probably overconfident in your edge estimates.
3. The two constraints that override any formula
No framework should be applied mechanically without two hard constraints.
Constraint 1: Concentration limits
Concentration risk has three layers:
| Layer | Hard cap (illustrative) | Why |
|---|---|---|
| Single name | ≤15% | One bad outcome can't permanently impair the book |
| Sector | ≤30–35% | Sector-specific shocks (regulation, rate cycle, commodity) |
| Theme / factor | ≤40% | Cross-sector correlated drawdown (e.g., "India domestic consumption") |
These caps should be set before you build any position. They are not negotiable because they protect you against the one scenario you didn't model: the fat-tail event that takes out an entire sector or theme simultaneously.
Constraint 2: Correlation under stress
The correlation penalty in the sizing equation is the most underestimated factor. Positions that appear uncorrelated in normal markets often become highly correlated under stress — this is the "correlations go to 1 in a crisis" observation.
A useful mental model: before sizing any new position, ask not "how correlated is this to my existing book?" but "if markets sell off 20%, which of my positions will be sold by the same forced seller?" If the answer includes more than 40% of your book, you are more correlated than you think.
Practical tool — correlation heat map:
Compute a rolling 90-day correlation matrix of your positions. Flag any pair with ρ > 0.6 and treat them as one effective position for concentration purposes.
4. A worked example: sizing a 5-stock portfolio
Suppose you have five ideas with the following characteristics:
| Stock | Conviction | Annual vol | 90-day ρ with rest of book |
|---|---|---|---|
| A | High | 22% | 0.3 |
| B | High | 35% | 0.7 (same sector as A) |
| C | Medium | 28% | 0.2 |
| D | Medium | 18% | 0.4 |
| E | Low | 45% | 0.1 |
Step 1 — Start with conviction weights:
A: 12%, B: 12%, C: 7%, D: 7%, E: 3% = 41% invested, or scaled to 100%: A: 29%, B: 29%, C: 17%, D: 17%, E: 7%
Step 2 — Apply volatility discount:
E's 45% vol is 2× A's — halve its weight relative to conviction. B's 35% vol warrants a modest reduction.
Step 3 — Apply correlation penalty:
A and B have ρ = 0.7. Treat them as one concentrated position. Combined, they might exceed your single-theme cap. Reduce B to medium conviction (8%).
Step 4 — Final allocation:
| Stock | Final weight | Rationale |
|---|---|---|
| A | 25% | High conviction, low vol, low correlation |
| B | 15% | High conviction but correlated with A — capped |
| C | 20% | Medium conviction, moderate vol |
| D | 22% | Medium conviction, lowest vol in book |
| E | 8% | Low conviction + high vol — capped at tracking size |
| Cash | 10% | Residual; reload capacity |
This is not mechanical — it requires judgment. The point is that the adjustments have reasons, not vibes.
5. What to do when a position moves against you
This is where most sizing decisions actually break down. The entry size is set thoughtfully; the decision to add when wrong is made emotionally.
Pre-define three scenarios before you enter any position:
| Scenario | Rule |
|---|---|
| Position down 10–15% | Re-examine the thesis. Add only if nothing has changed and your original edge remains intact. |
| Position down 20–25% | This is a forced decision point. Size is already painful. Add only if you have a specific catalyst or time-bound reason to believe the market is wrong, and you can afford the loss if the position goes to zero. |
| Position down 30%+ | Default rule: cut or hold flat. Adding here requires extraordinary conviction and is usually a mistake. The market has seen something you haven't. |
The standard amateur error is "averaging down" without a systematic rule — turning a 5% position into a 12% position purely because the price is lower. Price being lower is not itself a reason to buy more. Your variant view improving, or the market's thesis being demonstrably wrong, is.
Common mistakes
| Mistake | Why it happens | Fix |
|---|---|---|
| Sizing by conviction without vol adjustment | Conviction and volatility are uncorrelated; a high-conviction idea can still be high-vol | Apply a vol multiplier to all positions |
| Ignoring correlation within the book | Positions are evaluated in isolation | Run the correlation heat map before adding |
| Adding down without a pre-defined rule | Emotional response to being wrong | Write the add-down rule on the day you enter |
| Never rebalancing | Inertia | A position that doubles becomes overweight — trim to target |
| Treating cash as a residual | Cash is a position with its own opportunity cost | Decide upfront: what cash level triggers a full deploy? |
| Copying a famous portfolio | Survivorship bias — you see the winners, not the ones that blew up | Understand the original sizing rules before replicating |
Summary and next step
Position sizing is the translation layer between your research and your results. The four frameworks — equal-weight, conviction-weighted, volatility-parity, Kelly — each have a place, and the right answer is usually a hybrid: conviction drives the ranking, volatility and correlation drive the discount, and hard concentration caps provide the floor.
The most practical starting point: adopt a tiered conviction framework (high / medium / low), define hard caps (15% single name, 35% sector), and run the correlation heat map monthly. That alone will prevent the majority of sizing mistakes.
Related pieces: [What Is Asset Allocation](./what-is-asset-allocation.md) · [What Is Diversification](./what-is-diversification.md) · [What Is a Margin of Safety](./what-is-a-margin-of-safety.md) · [How to Build an Investing Research System](./how-to-build-an-investing-research-system.md)