Why We Killed Copy Trading on PerpLog
What we removed
PerpLog shipped a Copy Trading feature earlier this year. Leaders could opt in with a configurable profit-share (0–20%). Copiers could subscribe to a leader, set max risk and allocation, and receive mirrored trades with a 30-second delay (anti-front-running). All copied trades still executed via the user's own wallet through the Position Sizer, so fees flowed through the builder code normally.
It worked. It was well-tested. And it was the wrong feature to maintain.
Why we killed it
1. Hyperliquid Vaults does this natively, and better
Hyperliquid Vaults (app.hyperliquid.xyz/vaults) let a leader deposit capital into a shared vault. Followers join by depositing USDC. The leader trades the vault's capital, and P&L scales proportionally for everyone inside. Everything settles on-chain in one contract — the leader has their own skin in the game because their capital is inside the same vault.
That structure is strictly more trustless than signal relay. A signal-relay leader has no financial downside when a follower takes a losing trade; a vault leader does. That is the kind of alignment that only a protocol-native primitive can provide.
2. We claim to be a HYPE-maxi dapp. We should act like one.
PerpLog's positioning — committed publicly in no-token-philosophy and fee-redistribution— is that we build on Hyperliquid, we do not compete with it. We mirror Hyperliquid's 70/24/6 allocation ratios, route fees through HL's builder code, denominate Season prizes in HYPE, and we have no PerpLog token because HYPE is the alignment token of the ecosystem.
Shipping our own off-chain copy-trading layer while HL ships Vaults on-chain is not consistent with that positioning. It fragments attention. It asks users to choose between a PerpLog product and an HL-native product for the same need. And when we ask that question, HL wins — because Vaults are better.
3. Signal-relay copy trading has structural issues we didn't want to keep solving
A 30-second delay on signals is security theater. If a leader acts maliciously (front-runs their copiers intentionally), the delay does not help — it just costs slippage. If the leader acts in good faith (no front-running), the delay is unnecessary friction. Either way, the delay is a patch, not a solution.
A copy trading leaderboard without on-chain proof of performance is a trust-me system. We could not prove a leader's historical returns without exposing their trade history, which violated our privacy commitment. HL Vaults solve this structurally: vault returns are on-chain, verifiable, and cannot be gamed.
4. It freed ~415 lines of code for features only PerpLog can ship
Removing Copy Trading deleted src/lib/copyTrading.ts (leader/copier registration, signal relay, 30-second delay logic) and src/app/app/copy-trading/page.tsx (the UI). That is infrastructure we no longer have to maintain against HL API changes, test across edge cases, or defend in audits.
Engineering focus now goes to the features where PerpLog has a unique edge:
- Adaptive Sizing engine — quarter-Kelly with Bayesian shrinkage on your own segmented stats
- Discipline Tracker — hot-hand bias and scared-money-after-loss detection on your actual sizing deltas
- Weekly Review with double-blocking — review trades first, then let AI find patterns
- Insights Engine — revenge cycles, green-to-red, winner-cut-short detection
These are the features HL does not ship and will not ship — they sit on top of trade data that PerpLog is uniquely positioned to analyze. That is where our time should go.
What happens next
- The
/app/copy-tradingURL redirects toapp.hyperliquid.xyz/vaults— anyone who had bookmarked it lands where they should have been all along. - Season scoring stops rewarding copy-trading engagement — the
copiersAcquiredandleadersFollowedengagement fields are removed from the scoring engine as of this release. - We will ship a future “Vault Companion” layer— a PerpLog feature that imports a public HL Vault's trades into our journal, insights, and discipline tracker. That is meta-analysis on top of HL primitives, which is positive-sum. No timeline yet; it depends on Season 1 traction.
The principle, stated plainly
PerpLog is a HYPE-maxi dapp built on Hyperliquid. Before adding any feature, we ask: does Hyperliquid already ship this primitive? If yes, we do not duplicate it — we either build a meta-layer on top or decline the feature altogether. Copy Trading was a case where we had built before asking. We corrected the mistake.
We prefer shipping less and staying aligned than shipping more and contradicting our positioning. The Season 1 launch approaches; the surface area we defend matters more than the bullet-point count.
Frequently asked questions
What happens if I bookmarked the PerpLog Copy Trading page?
The /app/copy-trading URL now redirects to app.hyperliquid.xyz/vaults, which is the native Hyperliquid Vaults product. No data was collected from Copy Trading users (feature was not widely used pre-kill), and nothing is lost — HL Vaults is a strictly better product for the same use case.
Are HL Vaults and Copy Trading the same thing?
They solve the same user need (follow a skilled trader) through different mechanisms. HL Vaults let a leader deposit capital that followers can join; profits and losses scale proportionally and everything settles on-chain in the same contract. Traditional copy trading (what PerpLog offered) relays trade signals off-chain with a delay. Vaults are strictly more trustless, more aligned with the leader (they have their own skin in the game in the same vault), and more transparent.
Will PerpLog ever add trading social features?
Yes, but only meta-layers on top of HL Vaults — things HL itself does not ship. For example, importing a vault’s trades into PerpLog’s journal and insights engine to surface discipline metrics, pattern detection, and playbook attribution for vault followers. We build on top of HL primitives, we do not rebuild them.
Is this a purely philosophical decision or does it save money?
Both. Philosophically it removes a feature that duplicated a native HL primitive, which contradicts our positioning as a HYPE-maxi dapp. Operationally it cuts ~415 lines of maintained code, one Firestore collection, and the ongoing effort to keep the feature compliant with HL API changes. Engineering focus now goes to the features only PerpLog can ship: adaptive sizing, discipline tracker, weekly review, insights engine.