The gap between the bid and the ask on a symbol. It is paid on entry, it is the cost every trade starts behind by, and it is the one MT5 shows least clearly.
also: bid-ask spread, cost of trading, spread cost
Updated
You buy at one price and sell at another, and the difference between them is already against you the instant a position opens. Every trade starts at a small loss, and the spread is how large that loss is.
Spread is charged per trade rather than per pip of movement, so it is invisible on a strategy with hundred-pip targets and decisive on one with five-pip targets. It is also the cost a backtest models most confidently and least accurately.
| Setting | Typical value | Note |
|---|---|---|
| EUR/USD, raw account, liquid session | 0.0–0.3 pips | Plus commission, which has to be added before comparing with a no-commission account. |
| EUR/USD, marked-up account | 1.0–1.6 pips | No separate commission line; the cost is inside the quote. |
| Around a major release | 5–50 pips | Briefly, and on any account type. This is when spread filters earn their place. |
| At rollover | several times the session average | Predictable, daily, and the reason EAs should not place orders in that window. |
| Spread filter threshold | 2–3× the session average | Tight enough to skip the spikes, loose enough not to skip the strategy. |
One pip of average spread, applied to two strategies with different targets and different trade counts. The per-trade cost is the same; the verdict is not.
Spread is not expensive or cheap in itself. It is expensive relative to the size of the edge it is subtracted from.
Calculation 1.0 ÷ 5 = 20% · 1.0 ÷ 90 = 1.1% · 1.0 × 1,500 = 1,500 pips
Result The same broker is cheap for one strategy and disqualifying for the other
Judge spread against the average win, not against a table of broker averages. The comparable figure is total cost per round trip, on the pair and in the sessions the EA actually trades.
| Range | What it means |
|---|---|
| Under 2% of the average win | Immaterial. Choose the broker on execution and terms instead. |
| 2–10% | A real drag worth one broker comparison, and worth a spread filter. |
| 10–25% | Cost is now a design constraint. The strategy needs cheaper execution or larger targets. |
| Over 25% | The edge is smaller than the cost of collecting it on most accounts. |
Published backtests here state the broker and account conditions they were produced on, so the spread assumption behind a result can be read rather than guessed.
The tester's Current setting pins today's spread across the entire historical period, which for a multi-year test means one quiet afternoon's quote applied to every trade including the ones taken in a crisis. Testing at a wider fixed spread than measured live is the conservative choice, and the difference is largest for exactly the strategies that look best.
An account quoting 0.2 pips and charging $7 per lot round trip costs about 0.9 pips; an account quoting 1.2 pips with no commission costs 1.2. Ranking them by the advertised number gets the order backwards, and the error is systematic rather than occasional.
Spread is paid at the moment of the trade, not on average. A strategy whose entries cluster at session opens or around releases pays the widened quote nearly every time, so its realised cost can be several times the published average for the same symbol.
The spread is crossed twice — once entering, once leaving. Cost models that count it once understate every round trip by half, and the error compounds directly with trade count.