Position Sizing: How Much to Bet on Each Trade

Position sizing is where good ideas go to live or die. A trader can be right on direction and still lose money by betting too much, or too little, or on the wrong risk unit. That is why serious investors spend as much time on sizing as they do on signals. If you want a useful companion to this piece, start with risk and return basics, then come back here with a sharper eye for tradeoffs.

The core question is simple: how much capital should you allocate to each trade or asset? The answer is not one-size-fits-all. Fixed-weight sizing, inverse-volatility sizing, Kelly criterion, and risk parity each solve a different problem. Some are designed for simplicity. Some for equalizing risk. Some for maximizing long-run growth. And some, frankly, look better on paper than they behave in real markets.

This article compares the four approaches using real volatility data from major U.S. ETFs and a worked example you can reproduce. It also explains why full Kelly is usually too aggressive in practice, why fractional Kelly is the professional compromise, and how to choose a method based on strategy type rather than ideology. For readers building systematic rules, the discussion pairs naturally with how volatility is measured and rebalancing mechanics.

Why sizing deserves more respect than it gets

Most investors think in terms of selection: which stock, which ETF, which signal. But portfolio outcomes are often dominated by sizing. A concentrated position can turn a modest edge into a large gain — or a small mistake into a permanent loss. That is not a theoretical concern. It is the practical reason many strategies fail after a promising backtest. If you have not already read backtesting pitfalls, do that before trusting any sizing rule that looks too smooth.

There are two separate jobs in sizing. First, you decide how much risk to take overall. Second, you decide how to distribute that risk across positions. Fixed-weight rules answer the second question in the simplest possible way. Volatility-based rules try to equalize risk contribution. Kelly tries to maximize long-run log growth. Risk parity tries to spread risk more evenly across assets or sleeves. The methods are not interchangeable, and the wrong one can quietly distort your strategy.

Four sizing methods, four different philosophies

Here is the cleanest way to think about the four approaches. Fixed-weight sizing is a budgeting rule. Inverse-volatility sizing is a risk-normalization rule. Kelly is a growth-optimization rule. Risk parity is a diversification rule. Each has a place.

Kelly is the most misunderstood of the four. It is not a “maximize returns” formula. It is a growth-optimal rule under specific assumptions: known edge, repeated bets, and a stable distribution. Those assumptions are rarely clean in markets. The original Kelly framework came from information theory, not stock picking, and later finance work showed how sensitive it is to parameter error ^[4][5].

Worked example: the same capital, four different allocations

Let’s use a simple three-asset example with real trailing 252-trading-day annualized volatility estimates from widely traded ETFs as of a recent sample window: SPY for U.S. large caps, IWM for U.S. small caps, and TLT for long-duration Treasuries. These are not perfect proxies for every strategy, but they are familiar, liquid, and easy to verify using public price history from official fund pages or market data libraries ^[6][7].

Footnote: Illustrative calculation based on publicly available daily adjusted close data for SPY, IWM, and TLT over a 252-trading-day lookback window ending on the article’s publication date. Annualized volatility is computed as standard deviation of daily returns multiplied by √252. Universe: three liquid U.S.-listed ETFs. Rebalance frequency: monthly for the sizing illustration. Transaction costs, taxes, and slippage are excluded. Data source: public market price history via official fund pages and/or a public data library such as Stooq or Yahoo Finance-compatible feeds; readers should verify with their own data vendor before trading.

Now suppose you have $100,000 to allocate across these three sleeves.

In this simplified three-asset example, inverse-volatility and a naive risk-parity approximation look similar because we are ignoring correlations. That is the first practical lesson: once correlations matter, risk parity is no longer just inverse volatility. It becomes a portfolio optimization problem. If you want a deeper primer on how correlations change the game, see correlation and diversification.

Worked calculation: inverse-vol weights are proportional to 1/vol. So the raw weights are 1/0.15 = 6.67, 1/0.20 = 5.00, and 1/0.12 = 8.33. The sum is 20.00. Normalized weights are 33.3%, 25.0%, and 41.7%. Rounded, that is close to the table above. The point is not the exact decimal. The point is that the lower-vol sleeve gets more capital because each dollar is expected to contribute less risk.

Kelly criterion: elegant math, brutal reality

Kelly sizing is built around expected edge. In its simplest binary form, if a bet wins with probability p and loses with probability q, and the payoff ratio is b, the Kelly fraction is f* = (bp - q) / b ^[4]. In trading terms, the formula says: bet more when your edge is larger and your variance is smaller. That sounds sensible because it is. The problem is that markets rarely give you p and b with enough precision to trust the output.

Suppose a strategy wins 55% of the time and makes 1 unit when right while losing 1 unit when wrong. The Kelly fraction is 10%. That means the mathematically optimal bet, under those assumptions, is 10% of capital per independent trade. But if your true win rate is 52% instead of 55%, the optimal fraction falls sharply. If your payoff distribution has fat tails, the formula can still overstate the safe bet. This sensitivity is why Kelly is often described as a growth-optimal rule with a very expensive error bar ^[5][8].

Here is the practical tradeoff investors get wrong: Kelly is not dangerous because it is “too aggressive” in the abstract. It is dangerous because the inputs are usually wrong. A small estimation error can turn a theoretically optimal bet into a drawdown machine. That is especially true for short samples, regime shifts, and strategies with unstable edges. If your signal is tied to market regime, read regime detection before you even think about Kelly sizing.

Why do many professionals use fractional Kelly? Because the penalty for overbetting is asymmetric. If you size too small, you may underperform. If you size too large, you can blow up or suffer a drawdown that changes your behavior. Fractional Kelly is a recognition that estimation error, transaction costs, and human tolerance for pain all matter. That is not a philosophical compromise. It is a survival rule ^[5][8].

Risk parity: useful, but not a free lunch

Risk parity is often sold as a smarter way to diversify. The basic idea is to allocate capital so each asset contributes roughly equal risk to the portfolio. In a two-asset world, that can mean giving more capital to the lower-vol asset. In a multi-asset world, it usually means solving for weights that balance marginal risk contributions, often with leverage if the low-vol assets do not add up to 100% of desired risk ^[2][3].

The appeal is obvious. If equities are much more volatile than bonds, equal capital weights can leave the portfolio dominated by equity risk. Risk parity tries to fix that. But there is a catch: risk parity depends on correlations and volatility estimates, both of which move around. When correlations spike in stress periods, the portfolio can become less diversified than it looked in calm markets. That is one reason risk parity is best understood as a framework, not a guarantee ^[2][3].

For investors who want a practical bridge between theory and implementation, the question is not “Is risk parity good?” It is “What risk are you trying to equalize?” If you are building a multi-asset sleeve, risk parity can be sensible. If you are trading a single signal across many names, inverse-volatility may be enough. If you are running a discretionary book with a handful of ideas, fixed-weight sizing may be the right level of complexity. For a broader portfolio context, see asset allocation and why drawdowns matter more than returns.

What investors get wrong about position sizing

The biggest mistake is treating sizing as an afterthought. The second biggest is using the wrong sizing rule for the strategy. A momentum portfolio with many small, independent bets can often tolerate a volatility-aware rule. A concentrated event-driven book may be better served by fixed caps and hard loss limits. A high-turnover strategy with noisy estimates is usually a poor candidate for full Kelly. A low-turnover multi-asset portfolio may be a better fit for risk parity or inverse-volatility.

Another common error is ignoring correlation. Two positions can each look modest on a standalone basis and still create a large combined exposure if they are highly correlated. That is why a sizing rule should be tested at the portfolio level, not just the trade level. If you are building systematic rules, the checklist in backtest checklist is worth using before you trust any allocation logic.

Finally, many investors underestimate the behavioral side. A theoretically optimal sizing rule is useless if it causes you to abandon the strategy after the first ugly drawdown. That is where fractional Kelly, caps, and rebalancing bands become practical tools rather than academic footnotes. The best sizing rule is the one you can execute through a bad month without improvising.

A decision framework by strategy type

Use the following matrix as a starting point, not a law. The right answer depends on edge stability, correlation, turnover, and your tolerance for drawdown. If you are still learning how to think about portfolio construction, the three numbers that matter is a useful companion piece.

Decision tree: If your edge is unstable or hard to estimate, do not use full Kelly. If your positions have very different volatilities, do not use equal dollars without a reason. If your portfolio is multi-asset and you care about balanced risk, consider risk parity. If your process is discretionary and you need clarity, fixed-weight with position caps may be the most honest answer.

How to implement without fooling yourself

Implementation is where elegant sizing rules get messy. Volatility estimates lag. Correlations change. Rebalancing creates turnover. Taxes and spreads eat the edge. That is why a sizing rule should be judged after costs, not before. For a reminder of the hidden drag, see transaction costs and slippage.

Here is a practical checklist you can use before adopting any sizing rule:

If you want to go one level deeper, a simple worksheet helps. Write down the strategy’s expected edge, volatility, correlation with existing positions, maximum acceptable drawdown, and rebalance frequency. Then ask whether the sizing rule still makes sense after costs. That sounds basic. It is also where many backtests fail in real life.

So what should an active investor actually do?

For most active investors, the best answer is not a pure method. It is a hybrid. Use fixed caps to prevent one idea from dominating. Use volatility awareness to avoid accidentally concentrating risk. Use fractional Kelly only when the edge is repeatable and the estimates are defensible. Use risk parity when the portfolio is genuinely multi-asset and you want balanced risk contributions rather than equal dollars.

The real lesson is that position sizing is not about maximizing a spreadsheet. It is about matching the sizing rule to the strategy’s structure and to your own ability to hold the line. A sizing method that looks optimal but causes you to abandon the process is worse than a simpler rule that you can execute consistently.