ON-CHAIN

Open any perpetuals venue and there is a leaderboard of the most profitable accounts over the last 30 days. It is right there, sorted, and it is almost entirely a trap.

The trap is not that the numbers are wrong. The numbers are accurate. The trap is that sorting every account by recent profit selects for exactly two things: skill, and huge amounts of variance. And the second one is far more common at the top of the list.

What you are actually looking at

A leaderboard is a survivor list. You are seeing the accounts that made it. You are not seeing the thousands that took the same kind of risk and lost, because they are not on the list and their addresses are not memorable. If a strategy has a 5% chance of a spectacular month, then out of ten thousand accounts, five hundred will show a spectacular month. And every one of them will look like a genius from the outside.

Six checks, in order

1. Length of history

Filter out anything younger than three months. A new address that made money in a pump is unreadable. There has not been enough time for a bad environment to test it.

2. Size of the sample

Ten trades cannot distinguish skill from luck. Neither, honestly, can fifty. What you want is enough trades that consistency becomes visible, and enough time that more than one market regime is represented.

3. Drawdown, not just profit

An account that made 300% by being down 80% at some point is a different proposition from one that made 120% while never down more than 15%. Most venues will show you profit. Fewer will show you the path. If you cannot see the drawdown, you do not know the risk that produced the return.

4. Position concentration

Look at how much of the profit came from one position. A single oversized winner that accounts for most of the return is a bet, not a method. And it tells you the next outcome will depend on whether a similar bet works again.

5. Leverage and liquidation proximity

A profitable account running near-liquidation distance on every position is a strategy that has not been paid for yet. The profit is real; the survival is not demonstrated.

6. Whether the behaviour is repeatable by a human

Some wallets are profitable because they are first in a block, or because they are part of an operation you cannot replicate. Copying the visible actions of an automated strategy, without its speed or its capital, reproduces the losses and not the wins.

The bias that catches almost everyone

Even a wallet that passes all six checks gives you a subtle statistical problem: you found it after you saw its results. You selected on the outcome. That means the observed performance is inflated relative to the true skill, by an amount that depends on how many wallets you looked at before choosing this one.

The practical correction is to stop treating a leaderboard position as evidence and start treating it as a hypothesis. Watch the wallet forward, for real, without acting, and record what happens. Twenty forward-looking observations tell you more than a year of history you selected for being good.

What a good wallet actually looks like

Less exciting than the top of the leaderboard, and consistent:

• A long history across more than one market regime.
• Modest but persistent returns, with drawdowns that are shallow and recovered.
• No single position that dominates the profit.
• Recognisable, repeated structure in what it does. The same kind of trade, sized the same way, in similar conditions.
• Behaviour you could describe in two sentences without hand-waving.

If you cannot describe the edge in two sentences, you have found a number, not a strategy.

Why this matters more than picking well

The cost of copying a lucky wallet is not just the loss. It is that you learn the wrong lesson from it. You conclude a risk posture was safe because it happened to work, and you carry that forward into the next decision with more confidence than you had before.

Screening is not about finding the best performer. It is about refusing to draw conclusions from evidence that cannot support them.