Bitget published a report in 2025 claiming 93% of their copy traders were profitable.

An independent study analyzing over 100,000 copy trade outcomes found 48.48%.

One of these numbers is a marketing statistic. The other is what actually happens to real followers. They're not measuring the same thing, and knowing the difference will save you from a very predictable mistake.

How Copy Trading Works — The Honest Version

Copy trading is simple in principle: you connect your account to a "leader" or "signal provider," and their trades automatically replicate in your account at a proportional size.

A leader opens a 5% ETH-PERP long. Your account automatically opens a proportional ETH-PERP long. When they close, you close.

The pitch is compelling. You access the skills of profitable traders without developing those skills yourself. You sleep while the system trades. You participate in upside you couldn't have identified alone.

The problem is structural, not conceptual. The mechanics of replication introduce costs and distortions that don't appear in the leader's own results. This is where the gap between platform marketing and independent research opens up.

The Actual Numbers

The most cited figure in copy trading marketing comes from platform self-reporting. These numbers range from Bitget's 93% down to claims in the 70-80% range from other platforms.

What independent research actually finds:

An analysis of over 100,000 copy trade outcomes — tracking actual follower P&L, not leader P&L — put the profitable outcome rate at 48.48%. That's essentially a coin flip.

A separate 90-day study of featured leaders on major platforms found that only 43.61% produced positive net returns for their followers, even when the leaders themselves showed positive returns. The divergence between leader performance and follower performance is the whole story.

Bybit's own follower data — disclosed under less favorable conditions than their marketing materials — showed a follower win rate of 43.65% across tracked periods.

Academic research confirmed the structural finding. Apesteguia, Oechssler, and Weidenholzer published a study in Management Science in 2020 examining copy trading behavior, concluding it "reduces ex-ante welfare" — meaning followers would have done better making their own decisions, even imperfect ones, than delegating to observed traders.

Why the Numbers Diverge: Five Structural Problems

1. Execution Latency

Every copy trade arrives after the leader's trade. The leader gets filled first. By the time replication triggers, the order reaches you, and you get filled, the market has moved. On high-volatility assets or during rapid price action, the follower's average entry price is consistently worse.

This isn't a fixable implementation problem. It's physics. You cannot simultaneously be the first fill and a copy of the first fill.

On liquid, slow-moving markets the gap is small. On the altcoins that tend to produce the highest leader returns — the ones that show up at the top of leaderboards — the gap is largest.

2. Position Sizing Mismatch

Proportional sizing sounds fair. If the leader puts 5% into a trade, your account puts 5% in. But the leader's position sizing is calibrated to their account size, their risk tolerance, their overall portfolio, and potentially their view of how liquid a market is at that scale.

A leader with a $500,000 account sizing 2% into a position is entering differently than a follower with a $5,000 account copying that same 2%. Transaction costs, slippage, and minimum order sizes hit differently at different account scales.

More importantly: when leaders size large because they have high conviction and can absorb volatility, follower accounts may be forced to partially fill or size down, missing the best part of the move.

3. The Observer Effect

Leaderboards change behavior. When traders know their positions are being copied by thousands of followers, they make different decisions than when they're trading for themselves.

Some leaders start trading for the leaderboard — taking positions that look impressive on a tracker, manage drawdowns more conservatively to preserve public status, or trade with less aggression than they would privately. The version of the trader you're copying is the observed version, not the authentic version.

This effect is documented across contexts wherever performance is publicly ranked. Crypto copy trading isn't immune.

4. Survivorship Bias in Leader Selection

Platform leaderboards show the traders who are currently winning. You have no visibility into the distribution of past leaders who were once on those leaderboards and blew up.

The trader ranked #1 this month has one good month visible. The traders who held that position last quarter, performed badly in drawdown, and got dropped from the promoted list are invisible.

This creates a systematic bias: you're always copying from a pool selected for recent good performance, which has negative predictive power. Reversion to the mean after outlier performance is one of the most replicated findings in performance research. Copy trading leaderboards select for the traders most likely to revert.

5. Performance Decay After Public Exposure

When a trader's strategy becomes widely copied, the edge often deteriorates.

If 10,000 followers are replicating the same entry on a thin altcoin, the combined order flow of the copies affects the fill prices — and in some cases affects the price enough to partially front-run the original leader, reducing the move available to everyone including the leader.

Strategies that work when they're private work differently when they're public. Platform-featured leaders are, by definition, public.

What Platforms Are Actually Measuring

When Bitget publishes 93% profitability, they're not lying. They're measuring something real — just not what followers experience.

Platform profitability figures typically measure win rate at the leader level (what percentage of leader trades had positive P&L), over select time periods, among promoted leaders (already filtered for recent good performance). They may also measure count of profitable positions rather than portfolio P&L, which allows many small wins alongside catastrophic losses to produce a misleadingly high "win rate."

Follower P&L is the relevant number. It almost never appears in platform marketing, which is itself useful information.

The Alternative Model: Consensus Detection vs. Copying

There's a conceptual difference worth drawing clearly.

Copy trading inherits everything from one trader — their bad habits, their crowded entries, their observer-effect behavior, their survivorship-bias selection. You get the result, not the reasoning.

What makes smart money data useful isn't replicating one wallet's positions. It's detecting when multiple independently-acting wallets with demonstrated track records converge on the same direction. That convergence is the signal — independent confirmation from sources that didn't coordinate.

This is part of why I built the HyprSwarm dashboard. Not to surface a single trader to copy, but to identify when the wallets with the strongest performance histories are quietly building positions in the same direction. If three wallets in the elite tier are all adding ETH short exposure while price is stable, that's a different quality of information than one prominently ranked leader making a visible trade.

The smart money tracking guide covers how this works on Hyperliquid specifically — where every position is on-chain and verifiable, not self-reported.

What Good Traders Use Copy Trading For

Copy trading platforms aren't worthless. They serve specific use cases well.

For complete beginners who understand they're paying an education premium, copy trading offers market exposure with limited decisions required. If the expectation is "learn how professional traders size and manage positions" rather than "generate reliable returns," the cost of underperformance is partly offset by observation.

For managing a small portion of capital you're genuinely indifferent to, the set-and-forget simplicity has value.

What it doesn't do: generate the returns platforms imply at the leader level, transfer edge from one account context to another, or protect you from the structural degradation over time.

The Question You Should Actually Ask

Before copying any trader, the question isn't "what's their win rate?" It's: "what's the follower P&L for people who copied this trader over a full market cycle, not just the recent bull run portion?"

That number is almost never published. When it is, it's rarely the same as the advertised headline figure.

The honest version of copy trading marketing would say: on average, followers earn somewhat less than leaders, most leaders who appear on leaderboards revert to average or worse performance within a few months, and the structural costs of replication apply regardless of leader quality.

That's not a marketing pitch. It's also the truth.

Nothing in this article is financial advice. Past performance of copy trading leaders does not predict follower returns. All trading involves risk of loss.