Hyperliquid

How to review your Hyperliquid trades: fills, funding and the three numbers that matter

Updated November 12, 2026

A Hyperliquid wallet address exposes everything a trade review needs — every fill, every open position, every funding payment — through a public endpoint that takes no key of any kind. The catch is in the shape of the data: a fill is not a trade, the closedPnl Hyperliquid stamps on a closing fill is before fees, and funding never appears on a fill at all. Reviewing well means rebuilding trades from fills, netting fees and funding, and then computing three numbers before anything else: your net expectancy per trade, your fee drag, and your win rate on the trade after a loss. This guide walks all of it, then shows what Trade Copilot does with the same address.

What a Hyperliquid wallet actually exposes

Hyperliquid's /info endpoint answers questions about any wallet, keyed on the address alone. There is no API key to create, nothing to sign and no permission to get wrong — it is the same information the explorer shows for any account. The calls a review needs:

clearinghouseState
Your open perp positions — size, entry, leverage, unrealized P&L, liquidation price — plus margin used and account value. Each position also carries cumFunding.sinceOpen: the funding paid or received since it opened.
userFills and userFillsByTime
Your 2,000 most recent fills, or the fills inside a time window at 2,000 per request over the 10,000 most recent. This is Trade History, and it is what a review is rebuilt from.
userFunding
Every funding payment in a time window: the coin, the USDC amount, your position size at the time and the rate applied. This is Funding History.
frontendOpenOrders and historicalOrders
What is resting now, and what you placed, filled and cancelled. Cancelled orders are evidence too — they are where hesitation and moved stops show up.

The app shows the same rows under different names: Trade History is userFills, Funding History is userFunding, and the Portfolio page has its own portfolio request for account-value and P&L history. Any tool that asks for a private key to show you your own trades does not need it — every one of these is a public read.

How fills become trades

A fill is one execution against the book. A trade — the unit worth reviewing — is a position from open to flat, which may take one fill or twenty. Hyperliquid gives you what you need to stitch them together: dir says what the fill did (“Open Long”, “Close Short”), startPosition is your size before it, and sz and px are its size and price. Three details trip most reviews:

  • closedPnl is stamped only on closing fills, and it is before fees. Opening fills read zero. A trade's gross result is the sum of closedPnl across its closing fills; its net result is that sum minus every fee on every fill of the trade, opens included.
  • Fees carry their sign. fee is positive when you paid and negative when you received a maker rebate, a volume-share tier most traders never reach; feeToken is what it was charged in, and builderFee appears separately when you traded through a front end that charges one.
  • crossed tells you which side of the spread you were on. true means your order took liquidity and paid the taker fee; false means it rested and paid the maker fee. Trade Copilot does not record it, so it is the field to keep for yourself.

The reviewing mistake is grading fills. A partial close at a profit followed by the rest closing at a loss is one trade with one result, one decision and one lesson — reviewing the two fills separately produces a win and a loss where there was neither.

Where funding went

Funding on Hyperliquid is paid every hour — the conventional eight-hour funding formula divided by eight — directly between longs and shorts. It is added to or subtracted from your account balance and it never shows up as a fill — so a review built from Trade History alone silently omits it. On a scalp it rounds to nothing. On a position held for days on the crowded side it can be the difference between a winning trade and a losing one, and it is the most common reason a trader's own tally disagrees with their account value.

While the position is open, cumFunding.sinceOpen on the position is what it has cost or earned so far. After it closes, userFunding over the trade's window is the record. A trade's real result is closedPnl minus fees, plus or minus funding — and for anything held overnight you want all three columns.

The three numbers that matter

Win rate is the first number every dashboard shows and the least useful one: a 70% win rate loses money if the losses are big enough, and a 35% one makes money if they are not. Compute these three first. Everything else is detail.

1. Net expectancy per trade

(Σ closedPnl − Σ fees ± funding) ÷ number of trades

What one more trade of yours is worth on average, after everything. It needs a sample: below about ten trades it swings with every result and tells you nothing, which is the gate Trade Copilot's own performance metrics use. Positive means you have an edge and the job is to protect it. Negative means no amount of trading more will help.

2. Fee drag

Σ fees ÷ gross profit on winning trades, and the share of fills with crossed: true

At the base tier a taker pays 0.045% per side against a maker's 0.015% (at the time of writing) — three times as much, on a round trip that already has to beat the spread. A method that is profitable on resting orders and not on market orders does not have a strategy problem, it has an execution problem, and this is the number that tells you which one you have.

3. Your win rate after a loss

win rate and average P&L of every trade that immediately follows a losing one, against your overall

Sort trades by close time, take the trade after each loser, and compare that subset with everything. A gap has one meaning: the setups are not the problem, the state you enter them in is. This is the revenge-trading number, and it is the one traders most reliably never compute. Five trades in the subset is enough to start looking.

If the third number is the one that is off, the fix is not a better setup. Read why you keep revenge trading and how to actually stop — the interventions are all process, none of them willpower.

A review checklist for each closed trade

The numbers tell you whether something is wrong. The per-trade review tells you what. These six questions are the ones Trade Copilot's written review answers for every closed trade; they work just as well as six spreadsheet columns, as long as you answer them for every trade and not only the ones you feel like revisiting.

  1. Setup and plan quality. Was there a real reason — a catalyst, a structure, a pattern from your own playbook — or did the entry exist because price moved? Did the higher-timeframe context agree? What was the planned reward-to-risk, and did you have a stop and a target at entry?
  2. Rule fidelity. For each rule you have written down: followed, violated or not applicable, with one line of evidence from the trade. If you have no written rules, that is the finding.
  3. Execution quality. Four dimensions, one word each. Entry timing: on plan, early, late or chased. Stop discipline: held, moved tighter, moved wider, removed or never set. Target management: at plan, cut early, held past target or none. Sizing: within your parameters, oversized or undersized.
  4. Decision quality, with the outcome held aside. Would you take this exact trade a hundred times with only what you knew at entry? Four quadrants: good decision and good outcome (repeat it); good decision and bad outcome (do not change a rule off one loss); bad decision and good outcome (the dangerous one — a lucky win teaches a bad habit); bad decision and bad outcome (find the root cause).
  5. Mistakes, from a fixed list. Tag from a vocabulary you do not expand mid-review: chasing, late entry, moved the stop to avoid a loss, cut a winner short, revenge trade, trading to make back the day. A fixed list is what makes the patterns countable across months.
  6. One process goal. A single observable thing you will do differently next session. Not "be more disciplined" — "after any loss, the next position is half size."

How Trade Copilot does this from your address

Trade Copilot connects to Hyperliquid with the wallet address and nothing else — no API key, no private key, no signature. The integration only ever calls /info; there is no code path to the order endpoint, so it cannot place, cancel or withdraw even in principle. Connecting runs one check — clearinghouseState for the address, to catch a typo or a wallet Hyperliquid has no account for — then walks your last 90 days of fills, rebuilds your closed trades and emails you a coaching-style report that leads with your recurring mistake patterns. From then on every closed trade is journaled automatically when it closes, or the next time you open the app if it closed while the app was shut. Of those, up to 5 closes a day that happen while the app is open get an automatic written review against the six questions above and against the rules and plan you gave it.

It reads your open positions, your fills with their fee and closedPnl, your open and historical orders, and your spot balances. On the journal it computes your expectancy and profit factor (shown from ten trades), your drawdown, and your win rate after a loss (from five). The Mistake Radar ranks the mistakes your trade reviews keep finding, by how often they come back, and once you have ten closed trades in 90 days its Discipline view adds your habits, with overtrading and fee drag in dollars. It does not place orders, and it does not need to.

Two honest limits. The journal row keeps Hyperliquid's own closedPnl, which is before fees, though every review and statistic nets the fee on each fill out of it. And, specific to this venue, it does not read your funding payments: the funding endpoint it calls returns the market's rate by coin, not what your account paid or received. Spot fills are not journaled either; the journal, the reviews and the 90-day report are perps. So the copilot gets you the trades, the written review and the after-loss number net of fees, and you bring the funding column.

To see the after-loss numbers before connecting anything, paste any address into the free discipline report for any Hyperliquid address. It rebuilds the wallet's closed perp trades from up to 90 days of fills and measures revenge trading, sizing up after losses, letting losers run, overtrading and fee drag, every number net of trading fees. Funding payments are not included. It needs nothing else: no account, no key, no card.

See exactly how the Hyperliquid connection works, or how it works across all five exchanges.

Frequently asked questions

Do I need an API key or a private key to read my Hyperliquid trades?
No. Every read a review needs — fills, positions, funding, orders — is a public /info request keyed on your wallet address, the same one any explorer shows. Hyperliquid has no API key and secret pairs — its API wallets are signing keys — and a signature is only ever needed to act: place or cancel an order, transfer, withdraw. A tool that asks for a private key in order to show you your own history is asking for something it does not need.
Is closedPnl on Hyperliquid net of fees?
No. closedPnl is stamped on closing fills before fees; the fee is a separate field on every fill, opening fills included. A trade's net result is the sum of closedPnl over its closing fills minus the sum of fee over all of its fills — and for a multi-day hold, plus or minus funding, which appears on neither.
Does Hyperliquid's trade history include funding payments?
No. Funding is settled every hour straight into your account balance and never appears as a fill, so Trade History (userFills) does not contain it. Use Funding History (userFunding) for the record, or cumFunding.sinceOpen on an open position for the running total.
How many fills can I pull from Hyperliquid?
userFills returns your 2,000 most recent fills. userFillsByTime returns up to 2,000 per request within a time window, and only the 10,000 most recent fills are available that way. That bound applies to every tool reading the public API, including Trade Copilot's 90-day walk — a very active wallet may find its oldest fills out of reach and needs the CSV export or an indexer for the full history.
How many trades do I need before the numbers mean anything?
About ten before expectancy or profit factor is worth reading, and at least five in the after-a-loss subset before comparing it with your overall — the gates Trade Copilot's own metrics use before they show a number. Below that, one outlier moves everything. Win rate is the exception: it is stable early and not very useful at any sample size.
Can I run this checklist on a wallet that isn't mine?
Yes — every read here is public, so the same steps work on any address, which is exactly how public wallet audits are done. What changes is the value of the last three questions: decision quality, mistakes and a process goal only mean something to the person who made the decisions. For a wallet you are studying, stop at the three numbers and the execution-quality read.

Related

Trade Copilot for Hyperliquid — the connection, what is read, and a support table against the journals Hyperliquid traders use. Revenge trading — what to do when the third number is the problem. All guides.