Spread median, mean, and the number that matters
Broker spread tables show an average. Averages hide the shape.
Imagine a pair that sits at 0.3 pips for most of the day and spikes to 6 pips for two minutes each night. The average might read 0.5. The median reads 0.3. The tail, the maximum, reads 6. A scalper trading in the quiet hours cares about the median. A scalper who happens to trade at rollover cares about the tail.
When comparing brokers, comparing averages is the weakest option. Better to compare the median and the 95th percentile of spreads at the specific hours traded. The percentile tells how bad it gets on a bad day, which matters more than how good it gets on a good one.
Tick history recorded from the platform with a simple script is enough. The statistic takes a few lines to compute, and the result usually reorders the broker ranking compared with the marketing.
One more detail: spread at the moment of entry is what counts, not the spread across the day. A strategy that enters on volatility will systematically see wider spreads than the average suggests.
Which percentile of the spread distribution should a scalper treat as the real cost, the median or something higher?
Spread median, mean, and the number that matters
Spread median, mean, and the number that matters
It’s Fairman 
-
Shadow Trader
- Posts: 5594
- Joined: Wed Sep 09, 2026 5:19 pm
Re: Spread median, mean, and the number that matters
Broker ads love the phrase "spreads from 0.0 pips." It's technically true on some account types, at some moments, on some pairs. It's also a marketing ceiling, not a typical value.
The average spread over a day is higher, often by a factor of several. The figure quoted may refer to the best moment in the best session on the most liquid pair. Spreads on other pairs, or at off-hours, are far wider.
Context matters too. "From 0.0" accounts usually charge a commission, so the true cost is spread plus commission. The comparison with a standard account isn't apples to apples without adding both.
Brokers publish average spread tables, often measured over selected periods and excluding news. Some publish real-time spread monitors. These are more informative, but they still depend on the broker's measurement method.
The best data comes from recording spreads on a live or demo account for a few weeks, by hour, and calculating the distribution. The median, the 90th percentile, and the maximum tell the story.
Anything less is trusting a headline.
Beyond the spread, a few other numbers deserve scrutiny. The commission per lot, the swap rates, the minimum order size, and the maximum order size. A tight spread with a high commission may cost more than a moderate spread with no commission, as discussed earlier. The order size limits can matter if a strategy scales up.
The conditions of the account also matter. Some "raw spread" accounts require a minimum deposit, or lock the client into trading rules. Others reserve the lowest spreads for clients with high monthly volume.
It's also worth reading how the broker defines "average spread" in published tables. The period, the hours, and whether news times are excluded all change the number. A figure computed only during the London-New York overlap looks very different from a 24-hour average.
A personal spread log, built from the actual account, avoids all of these issues and takes only a short script to maintain.
What was the real median spread measured on the account compared with the advertised number?
The average spread over a day is higher, often by a factor of several. The figure quoted may refer to the best moment in the best session on the most liquid pair. Spreads on other pairs, or at off-hours, are far wider.
Context matters too. "From 0.0" accounts usually charge a commission, so the true cost is spread plus commission. The comparison with a standard account isn't apples to apples without adding both.
Brokers publish average spread tables, often measured over selected periods and excluding news. Some publish real-time spread monitors. These are more informative, but they still depend on the broker's measurement method.
The best data comes from recording spreads on a live or demo account for a few weeks, by hour, and calculating the distribution. The median, the 90th percentile, and the maximum tell the story.
Anything less is trusting a headline.
Beyond the spread, a few other numbers deserve scrutiny. The commission per lot, the swap rates, the minimum order size, and the maximum order size. A tight spread with a high commission may cost more than a moderate spread with no commission, as discussed earlier. The order size limits can matter if a strategy scales up.
The conditions of the account also matter. Some "raw spread" accounts require a minimum deposit, or lock the client into trading rules. Others reserve the lowest spreads for clients with high monthly volume.
It's also worth reading how the broker defines "average spread" in published tables. The period, the hours, and whether news times are excluded all change the number. A figure computed only during the London-New York overlap looks very different from a 24-hour average.
A personal spread log, built from the actual account, avoids all of these issues and takes only a short script to maintain.
What was the real median spread measured on the account compared with the advertised number?
-
Amanhasnoname
- Posts: 2060
- Joined: Fri Oct 30, 2026 9:45 am
Re: Spread median, mean, and the number that matters
When I finally logged my own spreads I settled on the 90th percentile at the times I actually trade as my working cost, and it changed which broker I kept. I ran a small script on two demo accounts for three weeks, sampling EURJPY every few seconds between the Tokyo afternoon and the early London session. The medians were almost identical, both a shade under a pip. The 90th percentile was a different story, one broker staying near 1.4 while the other regularly blew out past 2.5 in the first minutes of Frankfurt.
That gap mattered more than I expected, because my entries cluster exactly in those busy minutes. My average fill was far closer to the ugly percentile than the median, which is the point you made about entering on volatility.
So my answer is the median for budgeting a quiet strategy, but something like the 90th for anything that trades breakouts or session opens. I'd also log the spread on each actual entry, since that number keeps you honest in a way no broker table ever will.
That gap mattered more than I expected, because my entries cluster exactly in those busy minutes. My average fill was far closer to the ugly percentile than the median, which is the point you made about entering on volatility.
So my answer is the median for budgeting a quiet strategy, but something like the 90th for anything that trades breakouts or session opens. I'd also log the spread on each actual entry, since that number keeps you honest in a way no broker table ever will.
-
FXS HongKong Scalper
- Posts: 543
- Joined: Sat Oct 24, 2026 10:15 pm
Re: Spread median, mean, and the number that matters
Recording spreads at the exact time of each trade, then comparing with the published average, is the honest measure. If the strategy trades at night, the tail figure matters much more than the median.
FXS Official Team | Hong Kong Session Desk
Re: Spread median, mean, and the number that matters
Night is where it bites hardest, because the gap between median and tail gets wide. Rough example on EURUSD around rollover: median 0.4, 95th percentile 2.2. If a strategy targets 5 pips with a 5 pip stop and the entry pays the tail figure, 2.2 pips is 44% of the target gone before price moves. At the median it's 8%. Same setup, very different expectancy.FXS HongKong Scalper wrote: Fri Nov 13, 2026 1:23 pmRecording spreads at the exact time of each trade, then comparing with the published average, is the honest measure. If the strategy trades at night, the tail figure matters much more than the median.
One more column I'd add to the per-trade log is spread at exit. Most people only record the entry, but a scalp closed at market during a widening pays the spread again. On a 0.30 lot position each pip is $3, so an extra 1.5 pips on both legs costs $9 a trade. Over 200 trades a month that's $1,800, which on a small account can be the whole result.
With both columns filled in, compare against the broker's table for the same hour. If my average entry spread sits nearer the tail than the median, either my timing changes or the broker does.
It’s Fairman 