What is a copyability score?

A 1-5 rating of whether an account's entries can be mirrored before the edge disappears, given the real delay between the leader's fill and your order. It is not a quality score: a brilliant scalper can rate 1, and a mediocre but slow trader can rate 5.

Why not just rank by PnL or ROI?

Because raw PnL mixes luck, leverage, account size and reward-farming, and none of that tells you whether you could reproduce it. A leaderboard sorted by profit will happily put a high-frequency account on top — the one account a copier can never follow.

What makes an account uncopyable?

Three things, in order: hold times shorter than the signal-to-execution delay; many concurrent entries laddered into one move; and position sizes so small that your proportional share falls under the venue's minimum order.

Does a high copyability score mean the trader is good?

No, and treating it that way is the mistake it exists to prevent. Copyability answers 'could I mirror this?'. Profitability and risk are separate questions, which is why every account is also shown with its drawdown, its Sharpe at its real cadence and a check for whether its profit came from actual edge.

Copyability vs copy score: what a 1-5 rating should measure

Short answer: a copy score usually ranks how much someone made. Copyability ranks whether you could have made it too. They disagree far more often than people expect, and when they disagree the profitable one is usually the one you cannot follow.

The question each one answers

Public leaderboards sort by PnL or ROI. That single number blends together at least four things: the trader's skill, their leverage, the size of their account, and how much of the profit came from incentives rather than trading. Two accounts with identical PnL can be opposite propositions for a copier.

Copyability asks one narrow question instead: given the real delay between their fill and your order, can your entry still get a similar price? It is a timing verdict, not a compliment.

Why a great trader can be a bad leader

Take an account that closes positions about a minute after opening them. Its record can be superb. But a copy has to travel: the leader fills, we detect it, we size it against your capital, we submit. By the time your order rests, the move that made them money is over — and you are entering a position they are already leaving. Copying it does not produce a smaller version of their result. It produces a different, worse one.

Now take an account that holds for days and enters in discrete, visible steps. Less spectacular, entirely reproducible. For someone copying, the second is the better leader — and any ranking sorted by profit puts the first one on top.

What we actually measure

Hold time — how long a position stays open, versus how long a signal takes to become your order.
Entry cadence — whether entries are discrete events you can react to, or a continuous stream.
Concurrency — an account laddering dozens of simultaneous entries into one move is not one copyable decision, it is dozens of partial ones.
Size against the venue floor — if your proportional share of their trade lands under the venue's minimum order, the copy cannot be proportional at all.

The band definitions are published as machine-readable JSON at /api/ai/copyability, so an agent can read the rubric instead of guessing what a 4 means.

How to use it without fooling yourself

Copyability is a filter, not a recommendation. It tells you which accounts are even eligible to be followed. What decides whether you should follow one is separate and has to be checked separately: drawdown, consistency across months, whether the profit came from real trading or from farming incentives, and whether a single lucky position carries the whole record.

The honest order is: filter by copyability first, judge quality second. Doing it the other way round is how people end up mirroring an account they could never have executed.

Frequently asked

What is a copyability score?: A 1-5 rating of whether an account's entries can be mirrored before the edge disappears, given the real delay between the leader's fill and your order. It is not a quality score: a brilliant scalper can rate 1, and a mediocre but slow trader can rate 5.
Why not just rank by PnL or ROI?: Because raw PnL mixes luck, leverage, account size and reward-farming, and none of that tells you whether you could reproduce it. A leaderboard sorted by profit will happily put a high-frequency account on top — the one account a copier can never follow.
What makes an account uncopyable?: Three things, in order: hold times shorter than the signal-to-execution delay; many concurrent entries laddered into one move; and position sizes so small that your proportional share falls under the venue's minimum order.
Does a high copyability score mean the trader is good?: No, and treating it that way is the mistake it exists to prevent. Copyability answers 'could I mirror this?'. Profitability and risk are separate questions, which is why every account is also shown with its drawdown, its Sharpe at its real cadence and a check for whether its profit came from actual edge.

See how each metric is computed in the methodology, or read why the venue itself limits what you can copy in Hyperliquid vs Polymarket.