Risk Management in Automated Trading: Limits, Sizing and Kill Switches

Automated trading isn't passive income. A bot executes the rules you gave it, so weak risk rules get executed with machine precision, and a bug can repeat a mistake faster than you can react. Here's the risk setup I'd want before turning anything on, with hypothetical numbers.

Capital preservation comes first

A strategy's edge, if it has one, only shows up over a large sample. If drawdowns end the account before the sample is large enough, you never find out. Every risk decision should pass one test: does this keep me trading long enough to learn whether the strategy works?

The daily loss limit

Every bot needs a hard daily loss limit that stops new orders for the session. One conservative convention is 1.5% to 2% of equity. On a hypothetical $10,000 account that's $150 to $200. Ten straight maximum-loss days would cost roughly 18% to 20% of the account, depending on whether you compound. The limit is a trigger and not a guarantee, because slippage, gaps or a disconnect can push the realized loss past it.

Per-trade risk and tick math

Set a maximum dollar loss per trade, for example 0.5% to 1% of equity, then work backward to the stop. Contracts = risk dollars divided by (stop points times dollar value of a point). ES is $50 per point (tick $12.50), MES $5 ($1.25), NQ $20 ($5.00) and MNQ $2 ($0.50).

Add commissions and an allowance for slippage. Margin is collateral and not a cap on loss, so size from risk and not from how many contracts margin allows.

Correlated bots

ES and NQ tend to move together, so two directional bots on both are closer to one bet at double size. Measure the correlation on your own data and don't assume it. Mixing strategies built for different conditions, like mean reversion and trend or level-based logic, can reduce overlap, but strategies can still lose together in a sharp move.

News windows

Scheduled releases like FOMC, CPI and jobs reports can cause fast moves and slippage on stops. One common choice is to pause automation for a window around high-impact releases, such as 30 minutes either side. That's a policy decision and not a finding. Check that the filter actually fires by running it in SIM against the calendar, and consider turning the bot off entirely on days you don't trust.

Expectancy, not win rate

Expectancy = (win rate times average win) minus (loss rate times average loss). A hypothetical 40% win rate with a $150 average win and a $50 average loss gives 0.40 x 150 = $60 minus 0.60 x 50 = $30, so +$30 per trade. A hypothetical 80% win rate with a $20 average win and a $200 average loss gives $16 minus $40, so -$24 per trade. These only mean something if the averages are measured after costs. See the backtesting guide for how to measure them.

Drawdown recovery arithmetic

Ten straight losses at 1% risk leave about 90.4% of the account, a 9.6% drop that needs about 10.6% to recover. Ten straight losses at 10% risk leave about 34.9%, a 65% drop that needs roughly 187%. That arithmetic assumes fixed percentage losses in sequence. Real outcomes are variable, losses cluster, and variable dollar risk needs simulation instead of a simple formula.

Test the kill switch

1. Configure the limit in the strategy, the platform or the broker, and write down which one enforces it.
2. In SIM, deliberately breach it and watch what happens: are new orders refused, and what happens to an open position?
3. Confirm it fails closed. If the data feed drops or an error occurs, trading should stop instead of continuing blind.
4. Check order results explicitly. Orders can be rejected or partially filled without an obvious error, so don't infer a fill.
5. Restart the bot mid-session in SIM and confirm it can't double-fire the same signal.

Check the account's rules before automating at all. Apex prohibits automated or algorithmic preset bots that place, modify or cancel trades without your personal real-time review, per its prohibited activities page. Other firms differ by account product.

Reduce size in drawdowns

Raising size to win it back fast is the usual mistake. One protocol, with thresholds you'd set yourself: after a daily limit is hit, stop for the day. After a 5% weekly drop, trade half size the next week. After a 10% drop from the account high, pause automation and review whether conditions or the strategy changed. Scale back up with a defined scaling plan. For a specific bot's built-in controls, check its specs page.