Automated Copy Trading Strategy: The Layer You Actually Control

Three accounts sit on the same source. Account one runs a flat 0.50 lots per copy. Account two runs a 0.5x multiplier off the source's size. Account three sizes every copy to 0.75 percent of its own equity against the actual stop distance. The source places exactly the same trade on all three, and by the end of the month the three accounts have nothing in common — different exposure per trade, different worst days, different survival odds through the same losing streak.

The source did not change. The replication policy did. That policy is what "strategy" means in copy trading, and it is the part you own outright.

Two layers, and only one is yours

Copy trading has an alpha layer and a replication layer.

The alpha layer is whoever generates the trades — your own discretionary trading on a source account, a colleague's system, a channel you follow. You choose it, you can stop following it, but while you are following it you do not control it.

The replication layer is the transform between that source and your accounts. Every trade passes through it, and every parameter in it is a decision you made, explicitly or by accepting a default. People spend months evaluating the first layer and thirty seconds on the second, which is backwards, because the second is deterministic, testable and entirely yours.

There are five knobs.

Knob 1: the sizing rule

This is the one that changes outcomes most and gets the least thought.

Fixed lot. Every copy is the same size regardless of instrument, stop distance or account balance. It is the default in cheap tools and it is almost always wrong: 0.50 lots on EURUSD with a 20-pip stop and 0.50 lots on gold with a six-dollar stop are not comparable risks, and neither adjusts as the balance moves.

Multiplier. The destination copies the source's volume times a factor. Coherent when the accounts are similar in size and the source sizes sensibly; meaningless when they are not. A 0.5x multiplier from a 50,000 account to a 5,000 account still leaves the small account at roughly five times the source's percentage risk.

Balance or equity ratio. The destination scales by the ratio of its balance to the source's. This keeps percentage exposure aligned automatically as accounts grow apart, and it is the sane default for copying between your own accounts of different sizes.

Risk percentage. The destination ignores the source's volume entirely and computes its own from its equity and the trade's actual stop distance. Here is the arithmetic, using a 40-pip EURUSD stop on a standard contract where a pip is worth about 10 dollars per lot:

Account equity   Risk %   Risk in cash   Stop cost per lot   Lots
   10,000 USD     1.0%        100            400 USD         0.25
   10,000 USD     0.5%         50            400 USD         0.12
   25,000 USD     1.0%        250            400 USD         0.62
  100,000 USD     0.5%        500            400 USD         1.25

Those figures are illustrative and depend on your broker's contract specification, but the property they demonstrate is exact: risk-percentage sizing is the only rule that stays correct when the balance moves, when the instrument changes, and when the stop distance changes. It is also the only rule that requires the source to supply a stop — which is a feature, because a trade with no stop has no defined risk to size against.

Knob 2: filters

A filter is a rule for not copying. Every one of them trades coverage for control, and the defaults are rarely right for your situation.

• Symbol allowlist or blocklist. The single highest-value filter. If your source trades exotics you do not want, or an index your broker prices badly, exclude it rather than accepting it for completeness.
• Session windows. Copy only during the sessions you can supervise, or only during the sessions where your broker's spread on that instrument is tolerable.
• News windows. Skip copies inside a window around high-impact releases. On funded capital this is frequently not optional but a rule you are contractually held to.
• Maximum spread. Reject a copy when the destination's spread at that moment exceeds a threshold. This is the filter that protects you from copying into a broker's 03:00 spread widening.
• Stop-distance bounds. Reject copies whose stop is implausibly tight or absurdly wide. A parse error or a fat-fingered source trade usually presents as one of those two.
• Maximum concurrent positions and maximum per symbol. Caps total exposure regardless of what the source does.
• Direction filter. Occasionally useful — for example, running long-only on an account whose mandate says so.

Every filter needs to produce a visible, reasoned skip. A filter that silently drops trades is indistinguishable from a bug, and you will eventually spend a weekend proving which one it was.

Knob 3: aggregation and netting

The moment more than one source feeds one destination, you have a portfolio problem rather than a copying problem.

Opposing trades. Source A buys EURUSD, source B sells it. On a hedging account you hold both and pay both spreads for a net position of zero. On a netting account the second order reduces or reverses the first, and now one source's exit will close a position the other source thinks it still owns. Neither behaviour is wrong; running without deciding which you want is.

Correlation stacking. Five sources each risking 1 percent sounds like diversification. If four of them are long the dollar on the same morning, it is a 4 percent directional bet with extra steps. Copy trading spreads execution across accounts; it does not spread risk across uncorrelated ideas unless the sources genuinely are uncorrelated, and most retail sources are long the same three instruments.

Attribution. With multiple sources into one account, every position must carry a tag saying which source created it. Without that, an exit instruction from source B can close source A's position, and the account's history becomes unreadable after the fact.

Knob 4: update policy

What travels after the open is a policy choice, and most tools bury it in a checkbox.

• Stop and target modifications. Copy them, or manage your own levels after entry.
• Partial closes. Copy proportionally — the usual correct answer — or ignore, which lets the destination drift away from the source with every scale-out.
• Pending order cancellations. Copy them, or accept stale pendings that trigger next week.
• Positions already open when you connect. Copy them at current price, or copy only what opens from now on. Copying an existing position means entering at a price the source never paid, with a stop that may already be closer than the entry. "From now on" is the safer default.
• Manual trades on the destination. Managed, ignored, or closed as orphans. "Ignored" is the only answer that does not surprise people, and it requires the attribution tags from the previous section to be reliable.

Knob 5: circuit breakers

The rules that stop everything without asking you.

• Daily loss cap, enforced by the system, requiring a manual re-enable. A plan that depends on you noticing a bad day is a plan that depends on you being awake.
• Maximum drawdown from peak equity, which is the one that matters on funded capital.
• Maximum copies per day, which catches a source that has started malfunctioning.
• Consecutive-failure halt. If five copies in a row fail to place, something is broken. Stop and raise it rather than retrying into the dark.
• A kill switch you have tested mid-cycle, not just while idle.

And the breaker everybody forgets: what happens to open copies when copying is disabled. Close them all, leave them running, or stop new copies only. All three are defensible. Not knowing which one your tool does is not.

Testing a replication policy

The useful thing about the replication layer is that it can be tested without predicting anything.

Take the source's actual trade history — entries, exits, stops, partials, timestamps — and replay it through different policies. You are not testing whether the entries were good; you already know what they did. You are testing what the transform does to them: what exposure each policy would have carried, what the worst day would have looked like under each, how many trades each filter would have removed and which ones.

That is a deterministic comparison of your own rules against a fixed input, which is a far more honest exercise than a backtest of somebody else's entries. It answers questions like "would the maximum-spread filter have removed my three worst fills" and "does risk-parity actually reduce my worst day here, or just move it". A journal that records the reason for every skip makes this replay possible after the fact rather than only in theory.

The short version

Write your replication policy down as five lines — sizing rule, filters, aggregation, update policy, breakers — and put the document somewhere you will see it. Most people cannot state theirs from memory, which means they are running whatever their tool shipped with. The defaults were chosen by a developer who has never seen your balance, your broker or your drawdown rule.

Where this fits

If the destinations are funded accounts with their own rules, /blog/prop-firm-copier-sizing-across-accounts goes into how the sizing knob interacts with challenge and drawdown limits. And for what is actually being transmitted between the two ends — and what deliberately is not — see /blog/signal-providers-and-followers.