One trader, or a basket?

Drawdown pass-through

With one leader, your drawdown is their drawdown, undiluted. A 40% peak-to-trough decline in their account is a 40% decline in the mirrored portion of yours, arriving at exactly the moment you are most likely to abandon the strategy.

A basket does not eliminate drawdown — correlated leaders can decline together — but it removes the single point of failure. One leader's bad month is a fraction of the book rather than the whole of it.

One leader is one regime

Traders are rarely regime-agnostic. A breakout trader thrives in trending, high-volatility conditions and bleeds in chop. A mean-reversion trader does the opposite. Funding-harvesting strategies depend on persistent funding skew that vanishes when positioning flattens.

Copying one of them is a bet on their regime persisting for as long as you hold the strategy. Holding several with different archetypes reduces how much of your outcome rides on that single condition.

• Trend-following: strong in sustained directional moves, weak in range-bound markets.
• Mean reversion: strong in ranges, exposed during sustained trends.
• Funding-driven: dependent on a skew that can normalise without notice.

Why leaderboards push you toward concentration

Public leaderboards rank on raw PnL over short windows. That metric selects for leverage and luck: the fastest way to the top of a 7-day board is a large leveraged position that happened to work. Copying the top row is closer to momentum-chasing a lottery winner than to allocating to skill.

This is why selection here starts from consistency, profit factor and survivability rather than headline PnL, and why ROI is reported as unavailable when public history is not sufficient to compute it honestly.

A good record is not evidence of robustness

The central error in single-trader copying is treating a strong history as proof that the process is durable. A record is a sample, and the sample was drawn from one market environment by one decision process. Both of those are conditions, not properties. When either changes, the record stops describing the future without any visible warning that it has stopped.

Two failure modes compound this, and neither is observable in advance. The first is behavioural: sizing up after losses, holding a losing position past the plan, revenge trading after a bad week. None of it appears in a PnL curve until it has already cost money. The second is survivorship — you are looking at an account that is still visible, which is a filtered sample by construction, and the accounts that ran the same strategy into liquidation are not on the leaderboard to be compared against.

The concentration problem is therefore about tail dependence rather than about average returns. Your entire outcome depends on one account surviving. If it does not, your exposure does not underperform; it ends.

• One process means perfectly correlated errors, however many trades it takes.
• Behavioural failure is invisible until it is realised in the curve.
• Visible accounts are a survivorship-filtered sample of the strategies that were tried.
• If the leader stops trading, your exposure goes to zero with nobody making that decision.

What diversification cannot do

Diversification reduces idiosyncratic risk, not market risk. If the whole basket is long into a liquidation cascade, isolation and weighting will not save the outcome. Nor does a basket protect against a systemic exchange, oracle or bridge failure.

Past performance is not indicative of future results. Perpetual futures are leveraged instruments and carry a substantial risk of loss, including the loss of your entire position.