By the end of July 2026 our perps program had closed 486 positions on a 104 dollar Hyperliquid account. Before fees it was close to even: -$2.09. Of an 18.19 dollar loss, 16.10 dollars was fees. That is 89%.
| Closed positions | 486 |
| Above zero | 205 (42%) |
| Result before fees | -$2.09 |
| Fees | $16.10 |
| Result after fees | -$18.19 |
| Account size | $104 |
Why the fees grew
Too many orders. Live, the program placed about 30 a day, against 3 to 4 a day in our backtest, and it paid the taker fee on each entry.
Part of the 486 came from a setup mistake: about 5 separate programs were sending orders on the same account at once. We found it on 2026-07-14, stopped them and replaced them with one.
What we changed
- Entries moved from market orders to post only limit orders, cancelled if not filled, so we paid the maker fee instead of the taker fee.
- A 4 hour pause per coin after each position, about 5 times fewer orders.
- A larger size per position, so a one dollar gain was worth the fee.
- On trends, the exit target set at 3 times the stop distance.
How it ended
After the change the account ran in choppy markets at about breakeven, -$0.44 per 24 hours (2026-08-07). We have no record of a period that ended above zero. On 2026-08-09 we parked the account: the edge was too thin at about 96 dollars.
Fees were the leak. We cut them. The account was still too small for the edge.
The rule we keep
Count the fee line first. On Hyperliquid's published schedule (base tier, read 2026-10-24) a taker pays 0.045% per side and a maker 0.015%, so a round trip with market orders costs 0.09% before any price move. A strategy has to beat that on every position before it comes out ahead. Funding is another cost on top.
From our own records, June to August 2026: perps journal entries of 2026-07-14, 2026-07-28, 2026-08-07 and 2026-08-09. Figures are exchange fills, not estimates.
Data from public sources and our own records. Not financial advice.