Account Risk Management for Traders

Most trading discussions focus on entries. Most account blowups come from sizing. This post lays out how account-level risk is handled in practice: the vocabulary, the math, the constraints each market imposes, and a set of parameters worth defining before any system trades real money.

The trap: risk-per-trade alone is unbounded

The standard advice is to “risk 1% per trade.” Position size then comes from the stop distance:

units = (equity × riskPct) / |entry − stop|

That formula has no ceiling. As the stop tightens, size grows without limit. On a $10,000 cash account buying a $100 stock:

Every row risks exactly 1% to the stop, and the last row is impossible in a cash account. A backtester that sizes this way without further checks will happily report results for positions the account could never hold.

The lesson: risk-per-trade tells you how much you want. Separate constraints decide how much you’re allowed. Final size is the minimum of every constraint, never the output of one formula.

Vocabulary

The pair most often confused is heat and exposure. You need limits on both. A heat limit alone allows the 4.6× case above. An exposure limit alone lets wide-stop trades quietly risk too much.

The math behind small risk numbers

Losing streaks happen

For a system with loss probability p, the expected longest losing streak over N trades is roughly:

longestStreak ≈ ln(N) / ln(1/p)

Plan for a streak somewhat longer than the expected one, because it’s an average, not a maximum.

What a streak costs at different risk levels

Equity after k consecutive losses risking r each is (1 − r)^k.

Recovery is asymmetric

A drawdown of d requires a gain of d / (1 − d) to get back to even.

Together, these tables explain why swing systems usually keep risk per trade in the 0.5–1% range. A 10-trade losing streak at 1% is an unpleasant month. At 5% it’s a 40% hole that needs a 67% gain to fill.

The layered model

Each layer can only reduce size or reject the trade. No layer can increase what an earlier layer set.

signal
  │
  ▼
[1] Sizing intent        risk-per-trade → desired units
  │
  ▼
[2] Account constraints  buying power / margin for this market type
  │
  ▼
[3] Position limits      max notional per position, lot rounding
  │
  ▼
[4] Portfolio limits     heat, gross exposure, correlation, max positions
  │
  ▼
[5] Account state gates  drawdown ladder, daily/weekly loss limits
  │
  ▼
order
  │
  ▼
[6] Post-trade monitor   margin utilization, closeout distance, HWM

1. Sizing intent

riskAmount   = equity × riskPerTrade × sizeMultiplier   // multiplier from layer 5
perUnitRisk  = |entry − stop| + expectedSlippage + perUnitCost
desiredUnits = riskAmount / perUnitRisk

• Convert to account currency. In forex, stop distance is in the quote currency. Skipping the pip-value conversion silently mis-sizes JPY pairs and crosses.
• Enforce a minimum stop distance, for example half an ATR. Stops inside normal noise both oversize the position and get hit anyway.
• Use gap-adjusted risk for instruments that gap: max(stopDistance, gapFactor × ATR). A stop is an order, not a guarantee.

2. Account constraints

maxUnitsByMargin = availableBuyingPower / (price × marginRate)

marginRate is 1.0 for a cash account. Available buying power has to account for every open position and every pending order. When multiple strategies or instruments share one account, buying power must be reserved at the moment an order is approved, or several simultaneous signals will each size against the whole account.

3. Position limits

• Cap any single position at a percentage of equity.
• Set an absolute maximum order size as a fat-finger guard.
• Round down to the broker’s lot size. If that produces zero, reject rather than rounding up to the minimum.
• For thinly traded instruments, cap size at a fraction of average daily volume.

4. Portfolio limits

• Maximum portfolio heat: total R across open trades.
• Maximum gross leverage: set well below what the broker allows.
• Maximum open positions.
• Correlation buckets: cap heat per bucket. In forex, break each pair into per-currency exposure (long EUR/USD is +EUR, −USD) and cap net exposure per currency. Trading EUR/USD and GBP/USD long at 0.5% each is closer to one 1% USD-short trade than to two independent ones.

When a new trade would breach a portfolio limit, either reject it or shrink it to fit. A reasonable policy is shrink-to-fit with a floor: if the trade would be less than about a quarter of its intended size, skip it. A sliver of a trade is not what the strategy signaled.

5. Account state gates

A drawdown ladder measured from the high-water mark, plus shorter-horizon loss limits:

Three design rules matter more than the exact numbers:

• Halts don’t resume on their own. Recovering above the threshold should not silently restart trading. Require a deliberate decision to rearm.
• Measure with equity, not balance. Otherwise a large losing open position hides the drawdown until it closes.
• Define the session boundary once, such as 5 p.m. New York for forex, so “daily” means the same thing everywhere.

6. Post-trade monitoring

• Recompute margin utilization and distance to broker closeout on every price update.
• Above a utilization threshold, block new entries.
• As closeout approaches, reduce positions yourself, largest risk first, rather than letting the broker pick what to liquidate.
• Update the high-water mark and drawdown state.

How margin works by market

Assume one market type per account, since the rules differ enough that mixing them in one sizing model causes errors.

Cash equities

• Buying power is settled cash minus cash committed to pending orders.
• Leverage is at most 1.0, and there’s no shorting.
• US equities settle T+1. Reusing unsettled sale proceeds too quickly can cause good-faith violations in a real cash account.

Margin equities (Reg T)

• Initial margin is 50%, giving 2× overnight buying power. Maintenance is at least 25% under FINRA rules and often 30% or more at the broker, higher for volatile names.
• The pattern-day-trader rule has historically required $25,000 for frequent intraday round trips. FINRA has been revising it, so check the current rule.

Forex

• Margin is set per instrument. US retail limits are 50:1 on major pairs (2% margin) and 20:1 on others (5%). Brokers publish the exact rate per instrument.
• Brokers close out positions when equity falls to a set fraction of margin used; at OANDA, for example, that’s half.
• Trading is continuous during the week, but the weekend is a real gap.
• Overnight financing (swap) accrues daily and belongs in any backtest.

Futures

• Margin is a fixed dollar amount per contract, set by the exchange with possible broker add-ons.
• Contract notional is large, so whole-contract rounding dominates at small account sizes. Sometimes the correct answer is that the account is too small for the contract.
• Positions are marked to market daily, and contracts must be rolled before expiry.

Crypto

• Spot behaves like cash equities, but trades around the clock with no settlement delay.
• Perpetuals and margin products have per-position maintenance margin, a liquidation price, and periodic funding payments. Liquidation is typically faster and less forgiving than a forex closeout.

Parameters worth defining

These defaults are conservative starting points for swing trading, not recommendations. Percentages are of equity unless noted.

Per trade

Portfolio

Margin

Account state

Operational

In practice only a few of these get tuned regularly: risk per trade, max heat, and the drawdown thresholds. The rest should have sensible defaults that rarely change.

Easy-to-miss risks

• A stop isn’t a maximum loss. Gaps make portfolio heat an underestimate. A useful stress figure is the loss if every position gaps two ATRs against you.
• Correlations rise in a crisis. Buckets built from calm-market correlations understate risk exactly when it matters. Simple, conservative fixed groupings often beat estimated correlation matrices.
• Pending orders use buying power. Reserve it when the order is placed and release it on cancel.
• Stale data is a fat-finger in disguise.
• Everything must be in account currency: P&L, risk, and margin.
• Costs matter more than they look, especially for strategies with small, bounded profit targets like mean reversion.
• Reconcile with the broker. Periodically compare your recorded positions and cash with the broker’s and stop trading on a mismatch.
• Log every rejection and resize along with the constraint that caused it. Otherwise “why didn’t it take that trade?” has no answer.

Sanity checks for any backtester

Run these as invariants across every bar of every backtest:

1. On a cash account, gross exposure never exceeds equity.
2. Total risk to stops never exceeds the heat limit.
3. Several simultaneous signals never produce combined notional above buying power.
4. A very tight stop produces a position capped by buying power or the position limit, not an enormous one.
5. Positions in cross and JPY pairs risk the intended amount in account currency.
6. The drawdown ladder reduces size, then halts, and doesn’t resume on its own.
7. Simulated margin closeouts follow the same rule the broker uses.

If a backtest can’t pass these, its results describe a strategy no real account could have run.