Position sizing: survival is a math problem.

Size kills accounts before strategy does

A mediocre strategy sized conservatively loses money slowly, leaving time to notice and adapt. A genuinely good strategy sized recklessly can still destroy an account — because every strategy, including profitable ones, produces losing streaks, and oversized positions convert an ordinary streak into an unrecoverable drawdown. Ruin is a sizing phenomenon, not a strategy phenomenon.

The base rates make the point concrete. A trader with a 70% win rate — a strong premium-selling profile — will experience a streak of five consecutive losses with near certainty over a few hundred trades. At 2% risk per trade, that streak costs about 10% of the account: uncomfortable, survivable, statistically ordinary. At 15% risk per trade, the same ordinary streak removes more than half the account and the strategy never gets to demonstrate its edge.

Everything that follows on this page is one idea in different forms: decide, before every trade, the most it is allowed to cost — and make that number small enough that no plausible sequence of losses is fatal.

Fixed-fractional risk: 1% / 2% / 3%

Fixed-fractional sizing caps the maximum loss of each trade at a fixed percentage of current account value. The conventional bands: 1% conservative (new strategies, undefined edge, low drawdown tolerance), 2% balanced (the default for most defined-risk options traders), 3% aggressive (proven edge, high conviction, and a demonstrated ability to sit through the resulting drawdowns).

Because the cap is a percentage of current equity, the system self-adjusts: positions shrink in dollar terms during drawdowns — slowing the bleed exactly when it matters — and grow as the account grows. For defined-risk options structures the arithmetic is exact, since max loss per contract is known at entry: contracts = (account × risk%) ÷ max loss per contract, rounded down. If the answer is zero, the trade is skipped.

Kelly: the theoretical ceiling, not the target

The Kelly criterion computes the risk fraction that maximizes long-run compounded growth given a win rate and payoff ratio. For a trade that wins probability p with payoff b times the amount risked, the Kelly fraction is f* = (p × (b + 1) − 1) / b. It is a genuine mathematical result — and full Kelly is far too aggressive to trade.

Three reasons. Kelly assumes you know p and b exactly; in markets both are estimates, and overestimating edge with Kelly sizing is catastrophic, because betting beyond the true Kelly fraction reduces growth and can produce ruin. Kelly also tolerates violent drawdowns — full-Kelly paths routinely lose half their value on the way to their theoretical growth rate. And it assumes independent sequential bets, which a book of simultaneous correlated positions is not.

The practical convention is fractional Kelly — quarter-Kelly as the default — which sacrifices a modest amount of theoretical growth for a dramatic reduction in drawdown depth and in sensitivity to estimation error. A trade whose Kelly fraction works out to 10% of the account gets 2.5% at quarter-Kelly: right back in the fixed-fractional band, which is not a coincidence.

Worked example: sizing on a $25,000 account

Account: $25,000, risk policy 2% ($500 per trade). Candidate: a 5-wide put credit spread collecting $1.40, so max loss is (5.00 − 1.40) × 100 = $360 per contract:

Two contracts of the first spread would risk $720 — 2.9%, a silent upgrade from “balanced” to “aggressive” that nobody decided on purpose. And the second spread illustrates the rule traders resist most: when one contract exceeds the budget, the correct size is zero. Either choose a narrower structure or wait; the risk cap is a constraint, not an opening bid.

The correlation trap: five trades that are one trade

Per-trade caps assume trades are independent. Sell put spreads on five large-cap tech names in the same week and you do not have five 2% positions — you have one 10% bet on the same underlying factor, because in a sharp selloff their correlations converge and all five short puts get tested together. Diversification measured by ticker count is an illusion; risk lives at the factor level.

The defense is a second cap above the per-trade cap: limit total risk across correlated positions — same sector, same factor, same direction of volatility exposure — to roughly 2–3 single-trade units. A book of short premium across different tickers is still one short-volatility position when the VIX spikes, and should be counted as such. Before adding a trade, the question is not only “is this within my per-trade cap?” but “what does my book lose if its common factor has a bad week?”

Drawdown math: losses are asymmetric

The recovery required after a loss grows faster than the loss itself, because the gain must be earned on a smaller base. Lose 10% and you need 11.1% to get back to even — annoying. Lose 50% and you need 100% — a different order of problem entirely:

This asymmetry is the ultimate argument for small position sizes: the deep right side of that table is effectively a one-way door. A trader risking 1–2% per trade can absorb a long losing streak and still be operating in the region where recovery is a matter of routine; a trader who reaches −50% now needs to double the account merely to be back where they started, usually while sized down and demoralized. The goal of sizing is never to maximize the best sequence — it is to make the worst plausible sequence boring.

• What is the max loss?
• What is the expected shortfall?
• What is the worst realistic scenario?
• What happens if the stock gaps?
• What happens if the event goes against me?
• How should I size this trade?

• Am I too concentrated in tech?
• Are all my trades correlated?
• What is my total portfolio delta?
• What is my total vega?
• What happens if SPY drops 2%?
• What hedge should I use?