Risk warning. Algorithmic trading can lose money quickly when risk controls are absent or misconfigured. The rules in this article reduce the probability of catastrophic loss but cannot eliminate it. Past performance is not indicative of future results, and no risk-management framework guarantees profits or even prevents all losses.
Why AI traders fail more often than they should
Most AI trading accounts that are wiped out are wiped out by risk-management failures, not strategy failures. The strategy might have been mediocre. The risk-management was actively bad. Survival in algorithmic trading is overwhelmingly a function of the risk framework wrapped around the strategy, not the strategy itself. The seven rules below are the core of any survivable AI trading framework.
Rule 1: Risk no more than 1–2% of capital per trade
This is the foundational rule of all professional risk management and the rule beginners most often violate. Per-trade risk should never exceed 1–2% of total tradeable capital. On a $10,000 account, that is $100–$200 of risk per trade — not position size, but the dollar amount you would lose if the trade hits its stop loss. With a 5% stop, that supports position sizes of $2,000–$4,000. Anything larger is sizing past the survival threshold, and the mathematics of drawdown will catch up with you sooner than you expect.
Rule 2: Always use a hard stop loss
Every position must have a stop loss configured before the trade is opened, and that stop must be hard — placed as a working order at the exchange, not held mentally. Mental stops fail at the moments they matter most: during fast moves, during account drawdowns when discipline weakens, during news events when liquidity evaporates. Hard stops execute regardless of whether you are watching, awake, or emotionally able to act.
Stop losses should be placed at levels where the trade thesis is invalidated, not at arbitrary percentages. A trend-following entry’s stop should sit below the structural low that defined the trend; a mean-reversion entry’s stop should sit beyond the range boundary that defined the setup. Stops placed by feel, or at round percentages that bear no relationship to market structure, get hit by normal market noise rather than genuine thesis invalidation.
Rule 3: Cap daily losses
A daily loss cap stops the account from being damaged catastrophically by a single bad day. The cap should be set at 3–5% of total capital — meaning that once losses for the day reach that threshold, all positions are closed and no new trades are taken until the next day. The discipline this enforces is the most important behavioural discipline in algorithmic trading: it stops the account from compounding bad decisions during a bad session.
Most platforms with serious risk controls support automatic daily loss caps. If your platform does not, the cap should be enforced manually with a written rule and an alarm at the threshold. Either way, the cap is non-negotiable. Days that get away from traders are days when the cap was either absent or ignored.
Rule 4: Set a maximum drawdown circuit breaker
The drawdown circuit breaker is the daily loss cap’s larger sibling: a hard rule that pauses the strategy entirely if cumulative drawdown reaches a defined threshold. A 15–20% drawdown is the typical threshold for retail strategies. When the threshold is hit, the strategy stops, the trader reviews what happened, and trading resumes only after a deliberate decision rather than reflex.
The purpose of the circuit breaker is not to prevent drawdowns — every strategy has them — but to prevent the trader from compounding a strategy failure by continuing to deploy a system that has degraded. Genuine strategy degradation is hard to distinguish from variance in real time. The circuit breaker forces the conversation. See how Duneriat exposes risk-control parameters. Open a Duneriat account in minutes →
Rule 5: Diversify across uncorrelated strategies
Running a single strategy on a single asset is concentration risk regardless of how good the strategy looks in backtest. Running three uncorrelated strategies — a trend follower on Bitcoin, a mean-reversion strategy on EUR/USD, a grid bot on a sideways altcoin — distributes risk such that a regime change adverse to one strategy does not destroy the entire portfolio. Correlation matters. Two crypto trend strategies on different assets are not uncorrelated; they will both lose money in a crypto bear market.
For most retail traders, two or three strategies is the practical sweet spot. More than that and the operational overhead of monitoring each one exceeds the diversification benefit. Fewer than two and you are exposed to single-strategy risk regardless of how careful the position sizing is.
Rule 6: Limit total exposure to 30% of capital
Total open exposure across all positions and strategies should not exceed 30% of total tradeable capital at any time. The remaining 70% is dry powder — capital available to absorb drawdowns, capital available to deploy when conditions improve, and capital that prevents a correlated drawdown across positions from cascading into account-ending losses.
This rule is the one that sounds most conservative and that most reliably keeps accounts alive across multi-year horizons. Traders who run with 80–90% of capital deployed are running close to the edge of survivability. Traders who run with 30% deployed have multiple recovery paths from any single bad period. The mathematics of compounding favour the patient capital structure heavily over the long run.
Rule 7: Review weekly, adjust quarterly
Risk parameters should be reviewed weekly — not changed weekly, but examined to confirm they remain appropriate given recent strategy behaviour. Strategy parameters themselves should be adjusted on a quarterly cadence at most. The temptation to tweak strategies after every losing week is one of the most reliable ways to over-fit a system to recent noise. Strategies need time to express their statistical edge; weekly parameter changes destroy the time required.
The quarterly review should include drawdown analysis, win-rate stability check, regime classification of the prior period, and explicit decision to continue, pause, or retire each strategy. Decisions made on this cadence preserve the strategic intent of the system; decisions made daily destroy it.
How these rules compound
The seven rules look simple individually. Their power is in compound application. A trader who applies all seven cannot blow up an account in any single trade, any single day, or any single bad week, and is structurally protected against the strategy degradation that destroys traders who apply none of them. The downside is that the seven rules together cap maximum performance during favourable regimes — the framework is built to survive bad regimes rather than maximise good ones. That trade-off is the right trade-off for almost every retail trader.
Frequently asked questions
Are these rules too conservative for aggressive traders?
The rules are calibrated for survival across multi-year horizons. Traders who want to run more aggressively can scale up the parameters, but the failure rate increases non-linearly as the parameters loosen. Aggressive sizing is the reason most retail accounts blow up.
Do these rules apply to fully automated strategies and discretionary trading equally?
Yes. The rules are about capital preservation, which applies regardless of whether the trades are placed by a human or an algorithm. The implementation differs — automated systems enforce the rules through configuration; discretionary traders enforce them through written rules and discipline — but the principles are identical.
What if my platform does not support drawdown circuit breakers?
This is a meaningful platform shortcoming. Either enforce the circuit breaker manually with strict written rules and alarms at the threshold, or migrate to a platform that supports automated risk controls. A platform without configurable drawdown protection is operating below the floor of acceptable risk infrastructure for serious algorithmic trading.
How often should I rebalance position sizes as the account grows or shrinks?
Position sizing should be calculated on the current account balance, not the starting balance. Many serious traders recompute position sizing weekly. As the account grows, position sizes can grow proportionally; as the account shrinks, they should shrink proportionally to preserve the percentage-based risk discipline.