Quick Answer: Hyperliquid trader ROI verification means recomputing a trader's return, win rate, PnL, and drawdown from their public fill history instead of trusting the leaderboard headline. The open-source Hyperliquid Trade Bot does this in a read-only terminal app, needs no API keys, and flags martingale-style sizing. Treat the output as due diligence, not financial advice.

A leaderboard showing +412% ROI looks like a hiring decision. It's really a headline number, and it can hide fees, funding costs, one lucky trade, or a position size that doubles after every loss. Hyperliquid trader ROI verification is the step most copy-traders skip. The Hyperliquid Trade Bot is an unofficial community project that tries to audit those numbers before you mirror anyone. Here's how it works, where it can mislead you, and what to check yourself.

Why Leaderboard ROI Overstates What You'd Actually Earn

Leaderboard ROI is a percentage change over a window. It says nothing about the path, the costs, or how concentrated the gains were. Two traders can both show +200% and be very different to copy: one peaked and lost 15% along the way, the other fell 70% before recovering.

The project's README claims that a 30–40% gap between headline and verified ROI is common. That's the author's claim, not a figure I can independently confirm, so treat it as a reason to check, not a benchmark. The gap usually comes from three places: fees that compound over thousands of fills, funding payments on open perpetual positions, and a handful of trades that carry most of the profit.

The most useful way to read ROI is alongside drawdown. Drawdown is the peak-to-trough decline in equity. A 40% drawdown needs a 66.7% gain just to get back to even, because 60 has to grow to 100. Most people underestimate that asymmetry, and it's why a high ROI with deep drawdowns often isn't worth copying.

The most common advice is "pick the highest ROI." That's exactly backwards for copy-trading. The high number tells you the outcome. The drawdown tells you whether you'd have survived to see it.

How the Verifier Rebuilds the Equity Curve From Fills

The tool doesn't trust the leaderboard figure. Given a trader's address, it pulls the public fill history and reconstructs equity from first principles. According to the README, the verification pipeline covers:

  1. Fills-based PnL, marked to market on every fill and validated against L1 state
  2. Fee- and funding-adjusted returns, which are the tradeable number
  3. A drawdown envelope with peak-to-trough equity and time underwater
  4. A sizing regime detector for martingale and revenge-sizing patterns
  5. Outlier dependence, meaning how much total PnL comes from the top-N trades
  6. A significance test on Sharpe and expectancy
  7. Counterparty heuristics that score wash-trade and self-trade suspicion

Here's the practical point. Each of these checks answers a question the leaderboard doesn't. Fee adjustment answers "what would I have kept?" Outlier dependence answers "would this trader's edge survive without one lucky week?" The significance test answers "could this be luck over the sample size?"

The verdict is stamped as 🟒 TRUSTED, 🟠 REVIEW, or πŸ”΄ REJECT, based on a rubric in ~/.hl-verify/rules.toml. Because the thresholds are editable, the rubric is a set of opinions you can inspect. That's a feature, but it also means the stamp is only as good as the thresholds behind it.

That said, the verifier is only as good as its inputs. If a trader's activity spans spot and perp accounts, you need to understand what's included before trusting the totals.

Martingale and Grid Recovery: The Pattern That Looks Like Skill

Martingale sizing means doubling your stake after each loss, so one win recovers all prior losses. Grid recovery is a cousin: the trader adds to a losing position at fixed price intervals, lowering the average entry. Both produce a smooth equity curve with very high win rates, which is exactly why they look like skill.

Here's the arithmetic. Start with a base stake of 1 unit and double after each loss. After six consecutive losses you've staked 1 + 2 + 4 + 8 + 16 + 32 = 63 units in total, and the seventh stake is 64 units. A single losing streak that's longer than your capital can carry ends the account, not the strategy. You won't see that in a win-rate number.

Counter-intuitively, the trader with a 90% win rate can be the riskier copy. A high win rate in a martingale book means many small wins and a rare, very large loss. The win rate counts the wins and never sizes the loss. This is why the README pairs win rate with drawdown and a martingale score.

The detector looks for patterns such as:

  • Position size increasing after losing trades, especially by a consistent multiplier
  • Repeated adds to losing positions at regular price intervals
  • Size spikes that cluster after drawdown periods
  • A high win rate combined with a fat left tail in trade PnL

The default rubric sets max_martingale = 0.4. That number is a configurable threshold, not a universal standard. Set it too loose and you'll miss recovery behavior. Set it too tight and you'll reject traders who simply scale into conviction.

Drawdown, Sharpe, and Sortino: What Risk-Adjusted Returns Actually Tell You

The Sharpe ratio divides excess return by the standard deviation of returns. It penalizes all volatility, including upside swings you'd happily keep. The Sortino ratio replaces standard deviation with downside deviation, so it only penalizes returns below a target. For copy decisions, Sortino is often the more honest measure, because you don't mind upside volatility.

Both ratios need enough data to mean anything. The config defaults in the README use a 90-day window and a min_trades = 100 floor. Ninety days of a strategy that trades once a week gives you about 13 observations, which is far too few for a stable ratio. Check the trade count before you check the ratio.

Here's the comparison most people miss. A trader can have a higher Sharpe than another and still carry a worse max drawdown. Sharpe averages risk across the whole window. Max drawdown captures the single worst stretch, which is the one that decides whether you stay in the trade.

Most people stop at the ratio. Don't. Pull the drawdown and the trade count next to it, then ask whether you could have sat through the worst period without closing your position.

Headline ROI vs. Verified Metrics: A Comparison

The table below maps each signal to what it reveals, its blind spot, and the config key that controls its threshold. Thresholds are the README's defaults.

MetricWhat it revealsBlind spotDefault threshold
Headline ROIRaw percentage gain on the leaderboardIgnores fees, funding, and the path takenNone (display only)
Max drawdownWorst peak-to-trough equity lossDepends on the window you chooseReject worse than βˆ’35%
Profit factorGross profit divided by gross lossA few large wins can dominate itMinimum 1.6
Outlier dependenceShare of PnL from the top-N tradesThe cutoff N is a judgment callMaximum 0.30
Significance p-valueWhether the edge is distinguishable from luckAssumes trades are roughly independentp ≀ 0.05

The table isn't a ranking. It's a checklist. A trader who passes four rows and fails the drawdown row is still a risky copy.

Where the Tool Can Mislead You

This project is an unofficial community build, not affiliated with Hyperliquid, and the README says so. The disclaimer matters because you can't assume the numbers have been audited by the protocol or any third party.

A few specific limitations to keep in mind:

  • Thresholds are opinions. The [trust_rubric] section is a starting point. A stamp of 🟒 TRUSTED reflects the rubric you're running, not a universal truth.
  • Survivorship bias is still there. A verified history tells you how a trader performed, not whether their next 90 days will look the same. Traders who blew up and deleted their addresses don't appear on the leaderboard at all.
  • The screenshot is a preview. The dashboard example in the README is a static ASCII render, and its numbers aren't real results. Don't quote them.
  • Release binaries are still planned. The README lists release/ as planned, so the one-click launchers are the current path.

The project is read-only. It contains no order-placement code path and doesn't request private keys. The live monitor and trade log use an optional read-only wallet, and the README states that keys are never stored.

Run Your Own Check in Under an Hour

You don't need to trust my reading or the project's stamp. You can run the verifier yourself with a few commands.

  1. Clone the repository and install dependencies:
    git clone https://github.com/tigerogijzooav/Hyperliquid-Trade-Bot.git
    cd Hyperliquid-Trade-Bot
    pip install -r requirements.txt
    
  2. Start with the bundled offline dataset to learn the interface: python main.py --demo
  3. Open ~/.hl-verify/rules.toml and read every threshold before you trust a verdict.
  4. Pick one leaderboard address and run the verification with a 90-day window.
  5. Compare the verified ROI, max drawdown, and trade count against the headline figure, and write down the gap.

Doing this once teaches you more than reading any ranking. You'll see how fees and funding change the number, and whether a trader's record depends on a few trades.

Frequently Asked Questions

What is Hyperliquid trader ROI verification?

Hyperliquid trader ROI verification is recalculating a trader's return, PnL, win rate, and drawdown from their public on-chain fill history. It replaces the leaderboard headline with a figure that includes fees and funding, and it checks whether the result depends on a few outlier trades.

How do I verify a Hyperliquid trader's ROI without an API key?

Use a tool that reads public Hyperliquid data by trader address. The Hyperliquid Trade Bot needs only that address for verification, so you don't need a key. Set a window of at least 90 days and check the trade count, because a small sample makes every ratio unreliable.

Why is verified ROI lower than leaderboard ROI?

Leaderboard ROI usually ignores trading fees, funding payments on open positions, and how much profit comes from a few trades. Verified ROI includes those costs. The README describes a 30–40% gap as common, but that figure is the project's claim, so check it on the traders you care about.

Is martingale trading always a red flag?

Not always, but it's a pattern that needs a closer look. Martingale and grid recovery produce high win rates and smooth curves while hiding a large left-tail risk. If a trader's size rises after losses, check their max drawdown and largest single loss before deciding.

Audit Before You Copy

Hyperliquid trader ROI verification takes minutes once the tool is running, and it's the cheapest protection you have against copying a number that looks better than it is. This week, run python main.py --demo, pick one leaderboard address, and compare its headline ROI with the verified figure. Note the gap and the max drawdown, then pass this to someone about to copy a leaderboard wallet.