Scoring methodology: how leaders are selected and weighted

Why headline PnL is not one of the factors

Total profit is the most available number about a trader and the least informative. It conflates skill with size, with leverage, and with how favourable the period happened to be. Two accounts with identical process and different position sizing produce wildly different PnL; two accounts with identical PnL can have entirely different probabilities of surviving the next month.

Every factor below is chosen because it is harder to produce by accident than a large PnL figure is. None of them are impossible to game, and the page says so where that is true.

Factor 1 — realized PnL consistency

Consistency measures how evenly realized profit is distributed across a record rather than how large it is in total. A record where most of the profit arrived across many periods is materially different evidence from one where a single window produced everything and the rest is flat or negative.

It is a proxy for repeatability. A repeatable process leaves a repeated signature; a lucky window leaves a spike. This factor is the primary defence against the single-lucky-month problem, and it is why a trader with lower total profit can score above a trader with a larger, lumpier number.

• Catches: one-window records dressed up as track records.
• Does not catch: a consistent process that is consistently exposed to one regime.

Factor 2 — win rate

Win rate is the proportion of closed trades that were profitable. On its own it is close to meaningless — a trader can win 90% of the time and lose money by letting the losers run — which is exactly why it is never used alone here.

Its role is as a cross-check against profit factor. High win rate with weak profit factor describes a trader whose losses are larger than their wins and who is one bad trade from giving back the record. Low win rate with strong profit factor describes a trend-follower, which is a legitimate shape. The pair separates the two; either number alone cannot.

• Catches: martingale-shaped records and unmanaged losers, when read against profit factor.
• Does not catch: anything by itself — treat any platform that ranks on win rate alone with suspicion.

Factor 3 — profit factor

Profit factor is gross profit divided by gross loss. It answers a question raw PnL cannot: how much was risked to earn what was earned. A profit factor barely above 1 describes an account grinding out a result with almost no margin for a change in conditions.

It also suppresses one-trade records directly. If a single position produced the entire gross profit, the ratio is fragile in a way that a total-profit ranking would reward rather than penalise.

• Catches: thin-margin grinding and single-trade records.
• Does not catch: a strong ratio built over a period too short to be meaningful — which is what record-length floors are for.

Factor 4 — position discipline

Discipline looks at behaviour rather than outcome: whether position sizing is stable or erratic, whether leverage escalates after losses, whether positions are held into liquidation proximity, whether the account behaves the same way after a bad stretch as before it.

This is the factor that addresses the failure mode selection alone cannot see. Tilt does not appear in a historical PnL curve until it has already cost money. Sizing behaviour is visible earlier, and it is the closest available proxy for how a trader will act in the drawdown you have not yet observed.

• Catches: leverage escalation after losses, erratic sizing, brinkmanship near liquidation.
• Does not catch: a first-time behavioural failure with no precedent in the record.

Factor 5 — account survivability

Survivability asks whether the account has actually survived adverse conditions, rather than whether it has avoided encountering them. An account whose entire history sits inside one favourable trend has not demonstrated survivability; it has demonstrated a trend.

This is the factor most directly aimed at copy trading's specific asymmetry. A leader who is liquidated does not merely underperform — the strategy ends, and every account mirroring it ends with it. Survivability is therefore treated as a precondition rather than as a contribution to a score.

• Catches: accounts that have never been tested, and accounts that have already failed a test.
• Does not catch: a regime that has no precedent in any available history.

Why the floors are pass/fail

Weighted scoring systems have a specific pathology: they let strength on one dimension pay for failure on another. A trader with an extraordinary profit factor and no survivability evidence can still clear a weighted average, and the resulting allocation is precisely the allocation you did not want to make.

So the floors run first, as filters, not as inputs. Fail any floor and the trader is out of the candidate set. The composite score is computed only across survivors, which means the score answers 'how strong is this evidence?' and never 'is this trader eligible at all?' Those are different questions and combining them is how bad allocations get justified by arithmetic.

How the design resists the single-lucky-month problem

One favourable month can dominate almost any performance statistic. Four properties of this design are aimed at it specifically, and they work together rather than individually.

None of this removes luck. It raises the amount of luck required to pass, which is the most any selection process can honestly claim.

• Minimum record length: short histories are rejected outright rather than scored optimistically.
• Consistency over totals: an evenly distributed record outranks a lumpier one with a larger total.
• Profit factor: a record produced by one outsized position is penalised, not rewarded.
• Continuous rescoring: a score is a current reading, so a fading edge loses weight without waiting for a review cycle.

From score to allocation

Capital is distributed across qualified leaders in proportion to composite score, subject to per-leader notional ceilings. The ceilings exist because proportionality alone is unbounded: a leader whose score pulls away from the field would otherwise accumulate a concentration that defeats the point of holding a basket.

Weights are applied at rebalance rather than continuously, because every reweighting closes and opens positions and both sides pay fees and slippage. Turnover is a real cost paid out of your capital, so the sticky basket is deliberately patient on ordinary variance — a higher score alone never forces a replacement — and decisive only on emergencies.

Starter mode mirrors one leader; Full mode mirrors up to 10 and unlocks at $100,000 of mirrored volume. The gate is mechanical rather than commercial: a small balance split across ten margined sub-accounts produces positions too small to track their leaders faithfully after minimum order sizes and rounding.

What the score is not

The score is a statement about a record, not a forecast. It is computed from public, backward-looking Hyperliquid trading history, and every factor in it is conditional on the regime in which that history was produced. A high score means the evidence is strong and consistent under the measured dimensions; it does not mean the next period will resemble the last one.

Two further limits are worth stating explicitly. Scores are derived from public history, so anything a trader does off-venue or across other accounts is invisible to them. And the model portfolio and leaderboard figures are estimates derived from that public history — they are not a record of what any individual account earned.

Nothing here is financial advice. Perpetual futures are leveraged instruments: a position can be liquidated in full, and past performance of any trader is not indicative of future results. Copy trading does not remove that risk — it changes who makes the decision, not what the market can do to it.

• Not a prediction, and not a recommendation to copy any specific trader.
• Backward-looking and regime-conditional by construction.
• Blind to activity outside the public on-chain record.
• Model figures are estimates, not your account's realized results.