A complete trend rule
Build and test a simple trend system end to end · 10 min
- First four years
- Last four years
Same rule, same market. The first and the second four years look like different systems.
Your backtest of a trend system shows a Sharpe ratio of 1.2 over eight years. You chose its settings by looking at those eight years. On the next eight years, its Sharpe ratio is most likely…
A complete trend system answers every question before the first trade: which markets, when to enter, when to exit, how big, and what it costs. Anything left open gets decided by mood later.
- Markets: a fixed list chosen before testing, as uncorrelated as possible.
- Entry and exit: for example a 50-day channel to enter and a 20-day channel to exit.
- Size: each market scaled so the portfolio targets a volatility, say 10% a year, using only past returns.
- Costs: commissions and slippage on every trade, charged in the test.
Then read the backtest the way a skeptic would. Split it in halves; compare before and after costs; ask how big the standard error of the Sharpe ratio is (about √(1 ÷ years) for low ratios). A result that only exists in one half, or only before costs, is not a result.
- After costs
- Before costs
| Period | Sharpe | Return a year | Volatility | Max drawdown |
|---|---|---|---|---|
| First half | 1.51 | 16.1% | 10.3% | -7.9% |
| Second half | 0.68 | 5.7% | 8.8% | -9.7% |
| All, after costs | 1.10 | 10.6% | 9.5% | -9.7% |
| All, before costs | 1.28 | 12.4% | 9.5% | -9.2% |
Six generated markets with modest trends, eight years, the first 130 days only warm up the rules. Costs paid: 12.4% of starting capital; turnover 16.6× capital a year.
Generated markets, not real data. They show how a rule depends on the market it meets; they do not predict how real markets behave.
Sizing one market inside a four-market system
- Capital, portfolio target
- $100,000, 10% volatility a year
- Four roughly uncorrelated markets
- each gets 10% ÷ √4 = 5% volatility
- This market’s volatility
- 25% a year
- Position value
- $100,000 × 5% ÷ 25% = $20,000
- Standard error of an 8-year Sharpe
- √((1 + 1.2² ÷ 2) ÷ 8) ≈ 0.46
Capital $200,000. A market should carry 5% annual volatility of the account, and its own volatility is 20%. What position value does the rule hold?
A system says: "Buy 55-day breakouts on these 12 markets, risk 10% volatility, charge 0.1% per trade." What is missing?
A system has a Sharpe ratio of 1.1 in the first half of its backtest and 0.1 in the second. What is the most honest reading?
Missed questions come back for review after one day, then three, then a week.
Capital $100,000. You target 15% annual volatility on a market whose volatility is 20% a year. How much should the position be worth (in $)?
Hint: Position value = capital × target volatility ÷ market volatility.
Three years of a fresh generated market are on screen; the next two years do not exist yet. Tune the rule on what you can see, commit to two predictions, then generate the future.
- Your rule so far
Over the next two years, after costs, this rule will…
What kind of market is this?
Generated markets, not real data. They show how a rule depends on the market it meets; they do not predict how real markets behave.