The Trading Bot That Runs 24/7: How We Built BTC + QQQ Risk-Managed Automation

We run two trading bots in production, every market hour, every day. One handles BTC perpetual futures with 3x leverage. One handles QQQ-related equities with multiple intraday strategies. Both have been running for months without intervention, with risk management baked deep into the architecture. Here's how they're built.

The four guardrails every trading bot needs

Per-trade risk limit. No single trade risks more than 2% of total equity. This caps the worst-case outcome of any individual decision. Bad trades are inevitable; ruin shouldn't be.

Daily loss limit. If cumulative losses for the day exceed 3% of equity, the bot stops trading until the next session. Prevents bad days from becoming catastrophic days. Most blow-ups happen when a losing streak compounds across a session.

Drawdown kill switch. If total equity drops more than 15% from peak, the bot stops trading entirely until manual review. Forces a circuit breaker into the system that prevents total ruin.

Consecutive loss pause. Three consecutive losing trades trigger a temporary pause. Often the market regime has shifted and the bot's strategies have stopped working — better to pause and reassess than keep firing in unfavorable conditions.

Each guardrail by itself is intuitive. Stack all four and the bot becomes resilient to almost every common failure mode.

The strategy stack

The QQQ bot doesn't run one strategy — it runs three:

Opening Range Breakout (ORB). The first 30 minutes of the trading day establish a range. Breakouts above or below that range trigger position entry, with stops on the opposite side.

Gap Fade. Large overnight gaps tend to fill. Position the opposite direction of the gap with a tight stop.

ATR Range Trade. Markets oscillate within volatility-adjusted ranges. Buy at the lower band, sell at the upper band, exit on breakout.

Each strategy fires independently, with its own risk parameters. Total daily risk across all three strategies is capped at the 3% daily loss limit. The bot allocates 50% of capital to TQQQ/SQQQ and 50% to TSLA/TSLS pairs, so wins and losses are diversified across instruments.

The BTC bot is different

The BTC bot uses a Candle Range Theory approach combined with bear-flip regime detection. Different market, different mechanics. Same risk-management overlay.

It runs on BloFin's signal-bot infrastructure with 3x leverage on perpetual futures. Daily margin rebalance. Entries based on session sweeps. Stops at 1.5x ATR. Targets at 2.5x risk. Backtest performance over six months: +126% with 22.8% max drawdown — and crucially, six of seven months profitable, meaning the upside isn't dependent on a single lucky month.

The notification layer

Both bots send push notifications via ntfy.sh after every trade and a daily summary at market close. The summary includes: trades taken, P&L, total equity, drawdown from peak, and any risk-management triggers fired.

This isn't optional. A trading bot you can't monitor is a trading bot you don't trust. Real-time visibility into what the bot is doing, in plain language, is what makes it possible to actually run these things 24/7 without stress.

The journal

Every trade is logged to a JSONL file with full context: timestamp, instrument, strategy, entry, exit, stop, target, P&L, balance after. This becomes the primary debugging artifact when something goes wrong, and the primary input to weekly performance review.

If you can't reconstruct exactly what your bot did and why, you can't trust it. The journal is the difference between operating a black box and operating a system you understand.

What we don't do

We don't claim the bots are guaranteed to make money. They're not. Markets shift, regimes change, strategies that worked in 2024 stop working in 2026. The risk management is what allows them to fail gracefully when their strategies fail — not a license to ignore performance.

We don't manage other people's money with these bots. They're internal tools. We build them for clients who want their own bots, but the running and monitoring is the client's responsibility.

We don't promise specific returns. Backtests are approximations. Live performance varies. The architecture protects against ruin; the strategies provide the upside, and that upside fluctuates.

The build pattern

If you want a similar bot built for your own trading, expect to spend $6,000-15,000 depending on complexity. Backtesting infrastructure, risk management overlay, notification stack, and a paper-trading phase before going live — all of that is part of the work. Skipping any of it is how trading bots blow up. The architecture isn't optional. The strategy is.