Slippage

What slippage is

An order is sent at a price the bot sees. By the time it is filled, the market may have moved, or the order may be bigger than the quantity available at that price. The fill happens at the next available price, and the difference is slippage.

Slippage can also be positive for you (price improvement), particularly with limit orders and calm markets. The problem for most bots is that it is not symmetric. Stops and market orders are the ones that slip, and they tend to be used when price is moving fastest.

What causes it

• Fast markets. News, a data release or a sudden move changes prices faster than orders can be matched. The price you saw is gone by the time the order arrives.
• Size. An order bigger than what is available at the best price fills across several price levels. The larger the order, the worse the average fill.
• Thin books. When few orders sit near the price, a modest order moves it. Quiet hours and lightly traded instruments are thinner.
• Latency. The time from the signal to the fill, including your connection, the platform and the venue, is time for the price to move.
• Stop orders becoming market orders. A stop triggers and then sends a market order. It fills at the next price, which in a gap or a spike can be well beyond the stop level. A stop is a trigger, not a promised price. Rules differ between brokers and exchanges, so check how yours work.

A worked example: what slippage does to expected value

Use the expected value formula with costs per trade:

Take a made-up strategy that wins 55 percent of the time, with an average win of 6 units and an average loss of 5 units. A unit can be a pip, a point or a tick, as long as it is the same everywhere. Before costs:

Now add slippage as a round-trip total per trade (entry plus exit), keeping everything else fixed. Spread is left out here to isolate slippage. See spread for that cost.

Calculated by us on 27 Oct 2026 from the made-up inputs above. At 1 unit the edge is nearly gone: +0.05 per trade, 95 percent of the gross figure lost. At 2 units the strategy loses 0.95 per trade and about 950 units over 1,000 trades. Nothing about the signal changed. Only the fills did.

Why bots with small targets die from it

Slippage is roughly a fixed amount per trade. A target is the thing it is measured against. The smaller the target, the larger the bite:

Scalping bots, tight grids and high-frequency EAs sit at the top of this table. A grid with 5-unit spacing gives away 20 percent of each cycle to a single unit of slippage, before spread. A grid with 20-unit spacing gives away 5 percent.

How it hits each kind of bot

• Grid bots trade many small cycles, so slippage is charged on every one of them. In a fast move through several levels it also fills several orders at worse prices at once.
• Martingale bots add size when the market is moving against them, which is when books are thinner and moves faster. The larger orders slip more, and the final stop-out is also a market order.
• DCA bots use fewer, larger trades, so slippage matters less per trade, but market buys in thin coins can still fill well above the last price.
• EAs that trade news or breakouts are entering exactly when slippage is largest.

How to measure it

• Log the signal price at the moment the bot decides, and log the fill price from the account history. The difference per order is the slippage. Average it by order type, time of day and size.
• Separate entries from exits and stops from limits. Stops usually slip more.
• Look at the tails, not only the average. A few large slips can cost more than many small ones.
• Test with slippage in the backtest. Add a fixed amount per trade, then a larger amount for stops, and see where the edge reaches zero. That number is the budget.
• Compare demo with live. A demo account often fills at or near the quote. A forward test at small live size shows the real difference.

What it does not tell you

An average slip does not describe a bad day. The worst fills come in the fast markets where a bot is most exposed. Slippage also varies by venue, instrument, size and hour, so a figure measured in one setting does not transfer. Use it to set a budget, then check that the bot's edge survives double that budget.