11 Rule-Based Trading Bots on a Small Account: How We Built and Run Them

Educational only, not financial advice. We run these bots with small amounts we can afford to lose. Trading, and leveraged trading in particular, can lose money fast. Nothing here is a recommendation to buy or sell anything.

What we did

We run 11 rule-based trading bots with real money, about $100 each. AI helped us at every stage: drafting the rules, testing them on past data, writing the bot kit that runs them and checking on them while they run. The decisions themselves are fixed rules, not a model guessing.

The 11 bots are three kinds:

The same strategy runs once per exchange, on five major crypto exchanges, so each exchange gets its own bot with its own $100. Two more bots trade US stocks and funds on a practice (paper) account only; they are not part of the 11.

The first bots started on 23 September 2026. The rotation bot was the last to go live, on 2 October 2026. Every bot looks at daily price bars and checks in every 15 minutes.

The steps

1. Write the idea as a rule. "Buy when X, sell when Y, never hold more than Z." If you cannot write it down, a bot cannot follow it. AI is good at turning a loose idea into precise conditions and pointing out the cases you forgot.
2. Test the rule on past data. Run it over years of history before any money is involved, and look hardest at the worst stretch, not the average.
3. Pick the settings once. Our futures grid bots, for example, use 24 price levels across a range 20% wide, set from the last 20 daily bars, with a 50-day trend filter. The test decides the settings; the live bot then just follows them.
4. Run it on a practice account. Every bot in our kit starts on paper by default. Live trading has to be switched on for each bot on each exchange separately.
5. Go live with a small, fixed amount. We give each bot about $100. It sizes orders from that amount only, so one bot cannot spend the rest of the account.
6. Watch the plumbing first. Our first week was mostly about connections, not trading: keys with the right permissions, the right futures mode on each exchange, the right margin settings. Since 1 October 2026 every one of the 1,801 automatic checks across the 11 bots has completed, and 10 orders have been placed and filled with no one clicking anything.
7. Review and adjust slowly. Change one thing at a time, test it on past data again, and only then change the live bot.

The safety rules we follow

• A fixed amount per bot. Each bot has its own allocation of about $100 and never sizes from the whole account.
• A kill switch. One switch stops every bot from placing new orders. It is off only while everything is healthy.
• A daily loss limit. If a bot's money falls 5% in a single day, the kill switch engages automatically and trading stops until a person looks at it.
• Isolated margin on every futures bot. A losing position can only use the money given to that position, not the rest of the account.
• Leverage caps. The futures grid bots run at 1x. The rotation bot is capped at 2x per contract. There is no "unlimited" setting: a futures bot will not run until an amount, a leverage and a margin mode have been set.
• Exit orders can only close. Orders that close or shrink a position are sent as close-only, so a mistake cannot accidentally open a new, bigger position.
• Paper first, live per bot. Live trading is opted into per bot and per exchange, never for everything at once.
• Alerts on every failure. If a check fails, for example because an exchange rejects a request, we get a message straight away instead of finding out days later.

What we would do next

1. Keep the bots small and boring until they have run cleanly for a long stretch.
2. Review each bot's rules against fresh past data every few weeks, one change at a time.
3. Move a practice bot to real money only after it has run cleanly on paper for weeks.
4. Write up each rule type in plain language so you can test the same ideas on a practice account yourself.