Module 13: Trading Systems and Automation
Putting It Together: Building Your First System
Chapter 12 of 12 · about 5 minutes
12.1 From idea to running system
Over this module, Mint had built a system step by step. For the final chapter, Hash asked Mint to put the whole journey into one document. "If you can describe your system completely on a few pages, you understand it. If you can't, it isn't ready."
12.2 Mint's system document
1. Hypothesis (Chapter 2). Bitcoin trends persist on daily timeframes because of momentum and crowd behaviour. A trend-following system should capture large moves while limiting losses in downtrends.
2. Rules.
| Component | Rule |
|---|---|
| Market | BTC spot, on one reputable exchange |
| Timeframe | Daily, UTC close |
| Entry | Close above the 50-day SMA, with the 50-day SMA rising |
| Exit | Close below the 50-day SMA |
| Initial stop | Entry minus 2 × 14-day ATR |
| Sizing | Risk 1% per trade at full size; no leverage |
| Costs assumed | 0.15% per trade, updated from forward testing |
3. Backtest summary (Chapters 3 and 4), illustrative.
| Check | Result |
|---|---|
| Annual return versus buy and hold | Lower return, much lower maximum drawdown |
| Robustness to moving average length (40 to 60) | Results broadly similar |
| Out-of-sample period | Positive, somewhat weaker than in-sample |
| Tested on ETH | Similar behaviour |
| Costs doubled | Still positive |
| Number of trades | 48; modest sample, noted as a limitation |
4. Forward testing (Chapter 5). Three months of paper trading, then live at 0.25% risk, then 0.5%. Costs found to be higher than assumed; updated. All signals taken.
5. Automation (Chapters 6 to 10).
| Element | Setup |
|---|---|
| Execution | Simple script checks the daily close and places orders through an API |
| API key | Trade-only, no withdrawals, IP whitelisted, dedicated sub-account |
| Safety limits | Maximum order size, price sanity check, maximum one entry per day, daily loss cap |
| Alerts | Every order, every error, bot offline |
| Kill switch | Documented and practised on testnet |
| AI use | Research and code drafting only; no AI-driven execution |
6. Expected ranges and stop rules (Chapter 11).
| Metric | Expected range | Pause if |
|---|---|---|
| Win rate | 30% to 40% | Below 20% after 50 trades |
| Longest losing streak | Up to 8 | More than 12 |
| Maximum drawdown | Up to 10% | More than 13% |
| Cost per trade | 0.15% | Consistently above 0.3% |
7. Review schedule (Module 12, Chapter 13). Weekly: check alerts and rule adherence. Monthly: metrics versus expectations and versus holding BTC. Quarterly: full review; changes only if tested out-of-sample.
8. Place in the overall plan (Module 12, Chapter 14). The system trades within the satellite portion of Mint's crypto allocation. Core BTC and ETH holdings stay in self-custody and are not touched by the system.
12.3 Hash's review
Hash read through Mint's document.
"It's simple, tested honestly, with a known weakness you've accepted. Automation is limited and protected. You know what normal looks like and when to stop. And it fits inside your broader plan rather than replacing it."
"It's honestly a bit boring," said Mint.
"Most good systems are," said Hash. "Exciting systems tend to be overfitted, overleveraged, or both."
12.4 Common first-system mistakes
| Mistake | Better approach |
|---|---|
| Starting with complexity | Start with one simple, testable idea |
| Skipping forward testing | Paper and small live trading before scaling |
| Automating before understanding | Follow the system manually first |
| Giving bots broad permissions | Trade-only keys, sub-accounts, limits |
| Abandoning the system in the first drawdown | Compare with expected ranges first |
| Tinkering constantly | Scheduled reviews and out-of-sample testing |
| Letting the system replace the overall plan | Keep it within a defined allocation |
12.5 Module 13 wrap-up
Mint, Node, and Hash met at the end of the module.
"I've learned what a trading system is, how to design one from a hypothesis, how to backtest it, how to avoid fooling myself, how to forward test and go live gradually, how APIs, bots, on-chain automation, and AI agents work, how to manage operational and security risks, and how to know when a system has stopped working," said Mint.
"And the lesson?" asked Hash.
"A system is just my rules, written clearly enough to test and follow. Automation can execute them, but it can't make bad rules good, and it adds its own risks. The edge comes from honest testing and the discipline to keep following it."
Node added, "In my old job, I'd have called that process control. Test it, limit it, monitor it, and have a way to shut it off."
"That's exactly what it is," said Hash.
12.6 What's next
With the trading and investing modules complete, the remaining modules of Proof of Learning turn to the practical foundations that support everything else: taxation and regulation in different regions, crypto in your personal finances, and postmortems of crypto's biggest failures, with the warning signs that preceded them.
12.7 Try it yourself
Write your own system document using the eight sections in section 12.2. Fill in every section, even if some parts are still "not yet done." Then identify the next step you need to complete, whether that's backtesting, forward testing, or setting up safety limits, and schedule it.
Key takeaways from this chapter
- A complete system document covers hypothesis, rules, backtest results, forward testing, automation setup, expected ranges, stop rules, review schedule, and its place in your overall plan.
- Good systems are usually simple, honestly tested, and protected, with known weaknesses accepted in advance.
- Automate only after understanding and manually following the system, with narrow permissions and safety limits.
- Keep systems within a defined allocation of your broader plan.
- Testing, limiting, monitoring, and being able to shut down are the foundations of any automated system.
Coming up in Module 14: Taxation and Regulation. How crypto is taxed and regulated in different regions, record-keeping, common taxable events, and staying compliant as rules evolve.