The Copy-Bot Tax Hiding In Your Hyperliquid Trades
New ShieldTX analysis finds that top perp traders being copied lose a median of 63bps per trade. And the bots copying them aren't faring much better.
Copy trading has become one of the most popular ways to participate in on-chain perps. But how much does copying actually cost, both for the copiers and for the traders being copied?
We dug into 30 days of Hyperliquid perp data to find out.
What we did
We started with every wallet that traded perps on Hyperliquid over a 30-day window: 332,317 wallets in total.
From that universe, we identified potential copiers: wallets that opened the same coin and side within 60 seconds of a leader, on at least 10 matching fills over the window.
We then measured the economic cost on both sides of the trade:
Cost to copiers. For each matched fill, we calculated the entry edge given up by arriving late. We measured how much worse the copier's entry was than the leader's, multiplied by the copier's size:
`max(copier_price − leader_price, 0) × copier_size` for buys, mirrored for sells.
We summed these values across copiers, this captures the slippage copy-bots pay by chasing the leader's price. We counted only the fills where the copier got a worse price, so this is an upper bound on what copiers pay for arriving late.
Cost to leaders. We joined copier fills to each leader burst within the 60-second window, then split bursts into those that had copier fills during execution and those that didn't. Controlling for coin, size and volatility, we compared price drift between the two groups. If copied bursts drift meaningfully more, that difference is the cost of being copied.
The headline number. For a random sample of 7 wallets on the Hyperliquid leaderboard (30-day PnL), we compared the price drift on bursts that were copied against bursts that weren't. The gap between the two is the cost of being copied. We started with 7 wallets and are expanding coverage in future reports.
What we found
Across our sample, copied bursts drifted a median of 63bps more against the leader than comparable uncopied bursts.
On a $1M position, that is roughly $6,300 lost before the trade has a chance to work. The effect was more prevalent on Altcoins.
That is the cost of being copied: the adverse drift leaders experience when copy-bots pile into their positions mid-execution. It is a hidden tax on the best traders on the platform, and it adds up on every burst that gets copied.
Check your own wallet
We studied top-100 wallets, but the same mechanics apply to any public Hyperliquid wallet that bots can watch.
We built the Hyperliquid Copy-bot Scanner so you can see the estimated impact on your own trading. It uses a rolling 30-day window of all trading activity, perps only, to show which bots are mirroring your trades and what it is costing you. Make sure you have enough HL perp trading activity over the last 30 days to see the results.
Shield your strategy
Copy bots work by watching a known wallet. ShieldTX opens each position from a fresh Hyperliquid account, so there is no wallet for them to follow.
We built ShieldTX to provide enterprise-grade transaction shielding for high-volume perp traders on Hyperliquid with an API that integrates natively into existing trading setups. ShieldTX privately routes your trades to Hyperliquid, protecting your signal, reputation and strategies from copy bots, competitors and media. Your funding wallet stays separate from your trading wallet, and nothing on-chain links the two.
Methodology note:
Universe: 332,317 wallets.
Copiers: at least 10 fills on the same coin and side within 60 seconds after a leader.
Leader sample: 7 wallets drawn at random from the top 100 on the [30-day PnL] leaderboard.
Leader cost: median drift gap, in bps, between copied bursts and matched uncopied bursts.
Copier cost: one-sided entry slippage (upper bound).
Limitations: possible coincidental matches on high-volume coins, and drift may partly reflect momentum.
Scanner uses a rolling 30-day window of perps activity.