Slippage is the difference between the price at which an order was requested and the price at which it was filled. It happens because the price moved in the time between the request and the execution, or because there was not enough liquidity at the requested price to fill the whole order. It can be negative, a worse price, or positive, a better one. On average, at a fair broker, it is slightly negative, because orders are more likely to be triggered when the market is moving.
Where it appears
Market orders in fast conditions. Stop orders and stop losses, which become market orders once triggered, and which are by definition triggered when the price is moving through a level. Any order placed at the daily rollover, at the Sunday open, or around news. Limit orders do not slip, since they fill at the stated price or not at all, but they can be missed for the same reason.
Requotes and execution modes
Under instant execution, the broker either fills at the requested price or returns a requote, a new price the trader can accept or refuse. Under market execution, which most MetaTrader 5 brokers use, the order is filled at the best available price without asking, and slippage replaces the requote. An EA cannot respond to a requote in any meaningful way, so market execution is what it expects.
MetaTrader 5 allows an order to specify a maximum deviation in points, beyond which the order is rejected rather than filled. Setting it too tight causes missed fills; setting it wide accepts whatever comes. Fill policies, fill-or-kill, immediate-or-cancel, and return, govern what happens when only part of the order can be filled, and an EA has to use one the broker supports.
An EA places 300 stop orders in a year. At a broker with good execution the average slippage is 0.3 pips, about £2.40 per trade on one lot, £720 a year. At a broker with poor execution the average is 1.2 pips, £2,880 a year. If the EA's expectancy is £30 a trade, £9,000 a year, the second broker has taken a third of the edge in slippage alone. On a strategy with an expectancy of £8 a trade, the second broker took all of it.
The amount of slippage a strategy can absorb depends entirely on its average profit per trade. A daily strategy averaging 80 pips barely notices a pip of slippage. An intraday strategy averaging 6 pips is destroyed by it. Judge an EA's sensitivity by its average trade, and judge a broker's execution by running the EA on a demo and then a small live account and comparing the requested and filled prices in the history.
Backtests assume zero slippage unless the tester was told otherwise. A vendor's test should say what slippage it modelled, and a live record is the only place the real figure shows.