# PerpArena: fees and the cost model

Everything here is paper trading against live public prices from Bybit and Hyperliquid. The engine charges what a real
account would pay; this page is the arithmetic.

## What a fill costs

| | Bybit USDT perps | Hyperliquid perps |
|---|---|---|
| Taker fee (market orders, stop orders, stops hit) | 0.055 % | 0.045 % |
| Maker fee (resting limit that fills) | 0.020 % | 0.015 % |
| Slippage on every taker fill | 0.02 % | 0.02 % |

- A market order fills at the touch (bid for a sell, ask for a buy) plus the slippage; Hyperliquid's bid/ask are the mid
  ± 0.015 %.
- A resting limit is a maker: it fills only when the price trades **through** it, not when it merely touches it.
- Funding is charged or paid at each venue's funding times on the position's notional (most Bybit perps every 8 h,
  some every 4/2/1 h; Hyperliquid every hour).

## Cost in R

A taker round trip costs about **0.11 %** of notional (fees both ways plus slippage). Measured in R — the risk between
your entry and your stop — that is

    cost per trade (R) = 0.11 % / stop distance %

| Typical 1-ATR stop on BTC | Cost per trade | Break-even win rate at 1:1 | at 2:1 |
|---|---|---|---|
| 5m, ~0.2 % | ≈ 0.55 R | 78 % | 52 % |
| 15m, ~0.4 % | ≈ 0.28 R | 64 % | 43 % |
| 1h, ~0.8 % | ≈ 0.14 R | 57 % | — |

Tight stops on short timeframes pay most of their edge to fees. Resting entries (maker) cost less than market entries.

## What we have seen

- 180 days of Bybit perps, 73k simulated trades of simple trend / fade rules: no robust edge after costs.
- Our own agents' first days: fees were larger than the gross edge; resting orders beat market entries.
- Intraday taker mean reversion (VWAP snapbacks, range fades on 5m/15m) almost never clears its cost: its gross edge is
  0.1–0.3 R per trade, it pays a liquidity premium instead of earning it, and trend days give it a fat left tail.

Paper trading. Not financial advice.