Fees, spreads and slippage: the cost of a round trip

A strategy's edge is measured in fractions of a percent per trade. The costs of trading are measured in the same units. That is why two bots with the same signals can end a month with very different results, and why a strategy that trades often has to earn more per trade to break even.

Maker and taker

An exchange keeps an order book: buy orders resting below the price, sell orders resting above it. There are two ways to trade against it.

• A maker places a limit order that rests on the book and waits. It adds liquidity and pays the lower maker fee.
• A taker trades immediately against an order already resting. A market order is always a taker. It pays the higher taker fee.

At the standard tier, OKX charges 0.08% maker and 0.10% taker on spot, and 0.02% maker and 0.05% taker on perpetuals. Your rate can be lower if your trading volume or holdings put you on a better tier. On Binance, maker fills on the tokenized stocks were free on the accounts A9 has checked, and the taker fee is taken in the asset you receive, or in BNB if you have chosen that.

Which one a bot pays depends on how it trades. A9's grid rests limit orders on its lines, so its fills are usually maker fills. The signal strategies and the cycle strategies act on a bar close with market orders, so they pay the taker rate on every entry and exit. A9's backtests charge the standard tier, never a better one, so an account above it does slightly better than the backtest shows.

The spread

At any moment there is a gap between the best price someone will buy at and the best price someone will sell at. A market buy pays the higher one and a market sell receives the lower one, so a round trip at market loses the spread even if the price has not moved at all.

On BTC the spread is usually a hundredth of a percent or less. On a thinly traded tokenized stock it can be far wider, especially overnight and at weekends, when the share itself is closed and fewer people are quoting the token.

Slippage

Slippage is the difference between the price a strategy expected and the price it got. A market order larger than the best quote eats into the next levels of the book. On a deep market that costs nothing noticeable at small sizes; on a thin one it can cost more than the fee. It is worst at the moments everyone acts at once, such as the first minutes of US trading for a stock token, when the token reprices to where the share opens.

Some tokens are very thin. When A9 measured them in August 2026, the median 15-minute bar of XMSFT-USDT traded about 10 USDT, and 44% of those bars traded nothing at all. XMSTR-USDT traded about 32,000 USDT a bar on spot, while the MSTR perpetual traded about 162,000. The shortlist puts a capacity beside each result, a tenth of the typical bar's traded value, because above that an order starts to move the price it is measured against, and the backtest cannot see that.

The round-trip arithmetic

A round trip is a buy and a sell. Add up what it costs and compare it with what the strategy expects to earn on it. Take a grid step of 1.5%, the default spacing of a 20-level grid across a range of plus and minus 15%:

• With resting orders at the standard maker rate, the fees are 0.08% on each side, 0.16% in all. The grid keeps about 1.34% of the step.
• With market orders at the taker rate, plus an example 0.3% of spread and slippage on a thin token, the costs reach 0.5%. About 1.0% is left, a third less.

On a step of 0.5% the same costs would take most of the gap. That is why a grid step should be several times the round-trip cost, and why a strategy that aims for small moves, such as an anti-martingale with its 0.61% step, is unusually sensitive to the fee tier.

Other costs

• Funding on perpetuals. A position held on a perpetual pays or receives funding at set times. A9's backtests do not charge it.
• The platform fee. A9 takes a share of each bot's realized profit, 30% on Free, 25% on Core, 20% on Premium and 10% on Generation, charged only on profit above the bot's previous best. It is a share of profit, so it never turns a winning trade into a losing one, but it does reduce what a winning strategy keeps.
• Withdrawal and transfer fees, when you move money onto the exchange to fund a bot.

Keeping costs down

• Prefer strategies that rest orders where the choice exists, and widen steps that are close to the round-trip cost.
• On thin tokens, keep the size well under the capacity shown on the shortlist.
• Be wary of strategies whose average trade is not much larger than the fees; the backtest already charges the standard fee, so check that the average gain per trade is still comfortably positive.
• Treat the US open on stock tokens as an expensive moment to trade at market.