Ten accounts feels like diversification. If they all run identical copies of one strategy, it’s the opposite: it’s one bet at ten times the size, dressed up to look like ten bets. On the day that bet loses, they all lose together — and a portfolio of clones has no survivor.
Copies are perfectly correlated by design
Diversification works because uncorrelated positions don’t fail at the same time — when one is down, another is often up, so the portfolio is steadier than any single holding. That’s the entire mathematical benefit, and it depends on the positions being different.
A copier makes them identical on purpose. Same instrument, same direction, same entry, same exit, scaled to each account. That’s the feature — you want your decision expressed faithfully everywhere. But it also means the correlation between your accounts is effectively 1.0. There is no offsetting holding, no position that’s up while another is down. When the master wins, every account wins; when the master loses, every account loses, in lockstep.
So a copied fleet gives you two real things — spread across firms (counterparty risk) and spread across timing if you stagger — but it gives you zero strategy diversification. Pretending otherwise is the core error.
One drawdown day, total portfolio hit
The consequence is blunt: your fleet’s worst day is your single strategy’s worst day, multiplied by the account count. There is no averaging-out, because there’s nothing different to average against.
If your strategy has a bad session that risks tripping a daily loss limit, it doesn’t trip one account’s limit — it trips every account’s, at once, because they all took the same losing sequence. A losing streak that a single account could absorb becomes a portfolio-wide breach when it lands on ten identical copies simultaneously.
This is why the failure of a copied fleet is rarely gradual. It’s a single day where the one strategy underneath had its expected-but-painful drawdown, and every clone crossed its floor together. The equity curves don’t diverge and rescue each other; they’re the same curve drawn ten times.
Why more accounts isn’t more diversification
The intuition “more accounts = safer” is exactly backwards for identical copies. Adding a clone doesn’t reduce your risk of ruin — it raises the stakes on a single outcome. You’re not spreading the bet; you’re increasing the size of the one bet you already have.
Contrast the two shapes:
| Portfolio shape | Correlation | Effect of a bad strategy day |
|---|---|---|
| Ten different strategies | Low | Some accounts down, some flat or up — portfolio absorbs it |
| Ten identical copies | ≈ 1.0 | Every account takes the same loss — breach together |
The table is the whole argument. Account count only buys safety when the accounts hold different risk. Ten copies of a coin-flip is still one coin-flip — you’ve just agreed to lose ten times when it comes up tails. The only thing that genuinely lowers your correlated risk is either fewer, larger accounts, or genuinely decorrelated positions.
Decorrelating: staggered entries, varied strategies
If you want real diversification inside a multi-account setup, you have to reintroduce difference on purpose. There are two honest levers, and both cost you the clean simplicity of pure copying:
- Staggered entries and exits. Instead of firing every account at the identical tick, introduce modest variation in timing and level. This won’t decorrelate a strategy that’s simply wrong about direction, but it softens the execution-driven simultaneity — the “every account breached in the same second” problem during spikes.
- Varied strategies per account or group. Run two or three genuinely different edges across sub-groups of accounts, so a bad day for one isn’t a bad day for all. This is the only thing that actually lowers correlation, and it costs you real work: each strategy needs its own proven expectancy, or you’ve just added an unvalidated bet to dilute a good one.
Be honest about the trade-off. Perfect copying maximizes operational simplicity and minimizes diversification. Decorrelating buys resilience at the price of complexity — more strategies to validate, more to monitor. Before you split into multiple strategies, confirm each one clears the bar: a positive expectancy with a real sample behind it. Run the numbers with the expectancy calculator and think in R-multiples so you’re comparing edges on the same scale, not adding noise.
The math of simultaneous-breach probability
You don’t need heavy probability to feel this, just one illustrative example. Suppose — purely for illustration — a single account has some chance of a breach-level day; call it p.
- If accounts were independent, the chance of all of them breaching on the same day would be p multiplied by itself once per account — a tiny number that shrinks fast as you add accounts. That’s the diversified world you think you’re in.
- With identical copies, correlation is ≈ 1.0, so they breach together. The probability that all of them breach on a given day isn’t p × p × …; it’s just p. Ten accounts don’t make the breach ten times rarer — they make a single-account breach a ten-account breach.
That’s the entire correlation trap in one line: independence would make simultaneous ruin astronomically unlikely, and copying throws that protection away. Your fleet’s probability of a total wipeout is the probability of one account’s bad day — you’ve just attached ten accounts’ worth of consequence to it.
Practically, that means the thing protecting a copied fleet isn’t the account count; it’s the per-account floor. Size each copy as a fraction of that account’s remaining drawdown, model the floors with the prop-firm drawdown calculator, and enforce a hard limit at the broker per account so no single strategy day can walk the whole fleet across the line at once. Shibiki keeps those limits separate and enforced per account, and tracks live edge health with a Wilson confidence interval on the one strategy underneath — because when every account is the same bet, the only question that matters is whether that bet is genuinely positive. Confirm each firm’s drawdown mechanics directly, since they differ and change.
Related: understanding R-multiples · expectancy calculator · prop-firm drawdown calculator