Execution costs for trading agents: spread, fees and slippage

In nearly every automated trading setup, the first cost estimate is spread plus commission. That estimate misses what happens between the decision and the fill: the quote can move, the order can consume several price levels, and a limit order can sit unfilled while the market moves away. For an agent placing orders without supervision, these are model inputs, not accounting details.

A useful cost model separates explicit charges from execution shortfall. It also measures each component against a named reference price and records order size, venue, order type and time. Without those fields, a low reported average can hide expensive fills during the conditions that matter most.

Separate the visible costs from the hidden ones

The spread is the difference between the best ask and best bid. A market buy usually pays the ask and a market sell receives the bid, so a round trip across a stable quote costs roughly one full spread.

Commission is charged separately, often per share, contract or notional value. Some FX and CFD accounts show a wider spread with no separate commission, while exchange-traded products commonly itemise fees.

Slippage is the difference between an order's reference price and its executed price. Choose the reference deliberately.

Midpoint shortfall measures distance from the midpoint at decision time; arrival-price shortfall measures distance from the best quote when the order reaches the venue. The latter helps separate market movement during agent and network latency from the cost of crossing the book.

Do not add the full quoted spread to slippage if slippage is already measured from the decision-time midpoint. That counts some of the spread twice.

For a buy, a simple signed shortfall is fill price minus reference price; for a sell, reverse the sign. Convert the result to basis points of notional or a common account currency, then add explicit fees.

Compare order and pricing approaches

A market order favours execution certainty. It crosses the spread immediately and may sweep multiple price levels if available depth is thin. A limit order controls its worst acceptable price, but it trades fill certainty for price control. An unfilled order can miss the signal, and a fill can be adverse selection: the market continues moving against the agent shortly after execution.

Maker-taker pricing adds another decision. A passive order may earn a rebate or pay a lower fee, while an aggressive order may pay a taker fee. Rebates do not make a passive order automatically cheaper. Include the probability of a fill, the expected cost of waiting, and the risk that only part of the requested quantity executes.

• Use marketable orders when the signal decays quickly, but cap order size against visible depth and set a maximum acceptable price impact.
• Use passive limits when the signal can wait. Track fill rate and post-fill price movement, not just the price improvement on filled orders.
• Use sliced orders for larger size when immediate execution would consume several levels. Measure completion time and missed opportunity as costs too.

Venue and account pricing matter as much as order type. A fixed-spread account can simplify estimates in quiet conditions, but the spread may widen around news or low liquidity. Variable-spread access can be cheaper in liquid periods and more expensive at the exact time an agent is most active. Compare the full schedule, including minimum commissions, exchange fees and any financing charge for positions held overnight.

Typical market ranges for planning

The figures below are typical market ranges for liquid instruments in ordinary conditions, not quotes for a particular provider. They vary with venue, account tier, order size, volatility and time of day. Slippage is stated per side, and can exceed the range during news, fast markets or thin liquidity.

These ranges are starting assumptions, not substitutes for account-specific data. For scale, one pip on a standard EUR/USD lot is typically about $10, so a 0.2-pip spread costs about $2 per side at that size.

Add the round-trip commission and actual shortfall. On equities, a basis point is 0.01% of notional, which makes a few basis points material on frequent turnover.

Build the cost model into the agent

Estimate costs at the order level, then aggregate them by strategy, instrument, order type and market condition. Keep spread, explicit fees, shortfall, financing and missed fills in separate fields. This lets a developer distinguish a broker pricing change from a strategy that trades more often or sends larger orders.

For each candidate order, estimate expected cost as a function of size and current market state. A simple first version can use spread plus commission plus a slippage curve by notional relative to recent traded volume or displayed depth.

Add a fill-probability model for passive orders. Reject or resize orders when expected cost exceeds the signal's expected edge after risk and uncertainty buffers.

• Log decision timestamp, arrival quote, order submission and acknowledgement times, requested size, fills, cancellations and fees.
• Record volatility, spread, depth and recent volume at submission, then compare realised shortfall with the model by market condition.
• Set a maximum participation rate and a limit on acceptable slippage. Stop trading or switch execution mode when either threshold is breached.
• Recalibrate on rolling data, but keep a holdout period so recent costs do not simply get fitted away.

Paper trading can test whether the agent records and responds to these conditions, but simulated fills may not reproduce queue position or market impact. The agentic paper-trading environment is available from $9.99 a month; use any simulation alongside live market data and explicit assumptions about fills. Before enabling unattended orders, compare simulated costs with small, controlled live executions where permitted.

Validate the model before relying on it

Backtests often understate costs because they assume fills at bar prices, ignore queue position, or apply one fixed spread to every trade. Run sensitivity tests across wider spreads, higher fees and worse slippage.

For passive orders, vary fill probability and model adverse movement after fills. A strategy that only survives the most favourable cost assumption is not ready for unattended execution.

Review the distribution, not only the average. Track median and upper-percentile costs, rejected orders, partial fills, time to completion and missed opportunities.

Segment results by instrument, session and volatility. If costs rise sharply in one segment, an agent should have a defined response, such as reducing size, switching order type or pausing entries.

Cost assumptions are specific to the broker, venue, instrument and trading pattern. Recheck them after pricing changes and after a change in order size or latency. This advice does not apply unchanged to a strategy whose primary objective is guaranteed execution, such as an urgent hedge: in that case, execution certainty can be worth paying more than the modelled average cost.

#slippage modelling #trading execution costs #automated trading #spread and commission #market orders #limit orders #maker taker fees #order fill probability