Sep 22, 2026 · 4 min read
Four rules that keep the AI honest
Max risk per trade, max open positions, a daily loss limit, and a minimum reward-to-risk. Every idea passes all four or it cannot be executed.
An AI that proposes trades needs a leash. Ours is four rules you set once, that every idea is checked against twice: when it is proposed, and again with live figures the moment before an order is sent.
Max risk per trade
The percent of balance you can lose if the stop is hit. It drives position sizing directly. Default 1%.
Max open positions
Once you have this many trades open, new ideas are flagged and cannot be executed. It stops the model from stacking correlated bets when the whole board moves together. Default 3.
Daily loss limit
If today's realised loss reaches this percent of balance, execution stops for the day. This is the rule that saves accounts. Default 3%.
Minimum reward-to-risk
Ideas whose target is too close relative to their stop are shown, so you can see the model's reasoning, but they are blocked from execution. Default 1.5.
What a violation looks like
The card turns amber and lists the rule in plain words: "You already have 3 open positions; your limit is 3." The Open position button disappears, and auto-trade will not touch it. The model is told the same thing, so it explains rather than hides it.
Why we re-check at execution
Between an idea being proposed and you pressing the button, you might open another trade manually, take a loss, or the balance might change. The second check uses what the broker reports right now, so the rules hold even when the picture has moved.