Trend Following Screens: Test the Empty-Portfolio Policy

Trend following screens: Declare what happens to cash when no instrument passes a mechanical screen.

A trend screen needs an explicit empty-set cash policy. When no security qualifies, the account might remain in cash, hold prior positions, or use a declared fallback. Each choice changes the portfolio path. A comparison that changes both the screen and the fallback cannot cleanly isolate what the screen contributed.

The Reddit discussion compared a mechanical trend screen with simpler controls. Reported outcomes remain unverified; some replies were inaccessible. This article proposes a separate original experiment, not a replication or a verdict on discretionary trading.

Trend following screens: separate the components

Write four specifications before running anything: historical universe membership, qualifying rules, portfolio construction, and exits. Add the empty-set policy to construction rather than leaving it hidden inside a loop. State whether existing holdings must still qualify at every rebalance. A screen that selects candidates does not by itself define entries, stops, targets, or how much capital each candidate receives.

Keep warm-up eligibility identical across variants. If a long moving average needs more observations, begin comparisons only when their declared inputs are available. A security joining the universe late must not receive a fictional earlier membership history. Today's survivors are not a substitute for historical membership.

A synthetic two-by-two study

Suppose an invented screen selects three securities in month one and none in month two. Define two screens, full and simplified, and two policies, cash on empty and retain prior holdings. Register the four combinations before examining their returns. Each uses the same dates, costs, initial capital, rebalance timing, and executable price assumptions.

For illustration, an empty month with $10,000 equity and a declared zero cash return ends at $10,000 before fixed costs. A retain-prior-holdings variant with an invented 4% loss ends at $9,600. The $400 difference belongs to that policy comparison in this synthetic path. It does not establish that either policy will work better elsewhere.

Add exits as a later research dimension

A stop experiment introduces additional choices: trigger data, timing, order type, gap treatment, and remaining-position accounting. SEC order guidance distinguishes order behavior. In your fixture, make the market jump across the intended stop and require the simulator to use its declared executable event. Do not assume the intended threshold is also the fill price.

If you later add two exit policies to the four original combinations, document all eight variants. Changing the test after seeing which month hurt most consumes development information. Selection research motivates keeping search history. No fixed number of variants makes an experiment automatically valid or invalid.

Inspect the matched differences

Compare each policy against the same screen first, then compare screens under the same policy. Retain turnover, time in cash, exposure, transaction counts, and terminal open positions beside return and drawdown. If one variant lacks input history, report missing coverage instead of shortening only its evaluation window. Keep failed and neutral outcomes in the ledger.

Open Tradewink Advisory Research for the offline guide. Your first useful action is to write the four-row experiment table and a tiny empty-screen fixture. Review that table before supplying historical data. These diagnostics do not establish a complete discretionary strategy or authorize a live screen replacement.

Continue the research

Copy Trading Wallet Rankings: Freeze Eligibility and Trading Strategy Replication: Resolve Conflicting Exits develop adjacent checks. Use a small synthetic fixture first; these educational diagnostics do not validate a trading edge or change account settings.