A floor, an average and a typical spread are three different numbers, and Australian pricing pages use all three. Across the ASIC-regulated brokers tested for this article, the figure appeared in more than one of those forms and the form was often unlabelled. A page quoting “from 0.0 pips” is stating a minimum, not what a position typically costs. Before comparing two brokers on price, establish which kind of number each one is showing.
The figures below come from CompareForexBrokers, which recorded spreads across the ASIC-regulated brokers it covers between May and June 2026.
Three numbers that are not the same number
A floor is a minimum. It is the tightest quote the broker is prepared to publish, and on a raw account it is often zero. It says nothing about the distribution behind it.
An average depends entirely on the window it was measured over. A figure sampled across the London and New York overlap, the deepest hours in the market, will look different from one sampled at rollover when liquidity providers step back. Nothing obliges a broker to say which window it used.
A typical or indicative spread is a description rather than a measurement, and it carries no defined method at all.
None of this requires anybody to lie. A well-labelled floor and a generously chosen window produce flattering numbers on their own.
What the testing found
The capture ran on IceFX SpreadMonitor connected to each broker’s MT4 or MT5 server, on rolling 24-hour windows starting at 2pm Brisbane time. That window takes in the end of the New York session, the Sydney open, Tokyo, the London session and the London to New York overlap, so a full liquidity cycle sits inside every average rather than one favourable hour.
The measurement runs on a London VPS. That is deliberate and worth stating plainly: it holds the recording environment constant across brokers so the numbers compare to each other, and a spread is a broker-side quote rather than a function of where the trader sits. Network latency is a separate question and a separate test.
Results are published as an eight-pair average covering EUR/USD, AUD/USD, GBP/USD, USD/JPY, EUR/GBP, EUR/JPY, AUD/JPY and USD/CHF.
Across the raw accounts tested, that eight-pair average ran from 0.15 pips at the tightest to 0.68 at the widest. On standard accounts it ran from 0.83 to 1.97. Both ranges sit inside a single regulator, on brokers holding the same licence class, quoting the same instruments.
The outliers stay in. Where a rollover spike pushed a maximum out to several pips, that reading remains in the underlying data rather than being trimmed, because the maximum is what a stop-loss meets during news.
What the gap costs
The dispersion is the finding, and it reads more clearly in money than in pips.
On the eight-pair average, the widest standard account in the tested set sat 1.14 pips above the tightest. One pip on a standard lot of a US dollar quoted pair is about US$10, so a trader placing 50 standard lots a month is looking at roughly US$570 a month between the two ends of the same regulated market. Nothing about the licence explains that gap, because every broker in the set holds the same one.
Raw accounts compress the range without removing it. The eight-pair spread average ran from 0.15 to 0.68, a difference of 0.53 pips before commission is counted, and commission varies independently of the spread. A broker with a tight spread and an expensive commission can finish behind one with a wider spread and no commission at all.
That is the practical reason the two numbers have to be added together before anything is compared, and the reason a leaderboard built on advertised spreads alone can put the brokers in the wrong order.
“A published spread is whatever the broker chose to publish,” says Justin Grossbard, co-founder of CompareForexBrokers. “It might be a minimum, it might be an average, and the page often doesn’t say which. What I want to know is what the account was quoting at nine on a Tuesday morning, so we record it and publish the method beside the number.”
The commission is bigger than the spread
On a raw account the headline number is the smaller half of the cost, and this is where the spread alone gives the most incomplete picture of what a trade actually costs.
Across the tested raw accounts, AUD/USD spreads sat around 0.1 to 0.16 pips. Commission added between 0.39 and 0.58 pips on top. The commission is three to four times the size of the spread printed beside it.
A trap sits underneath that. Commission is usually quoted per side, so the round-trip cost is double the advertised figure, but at least one Australian broker quotes a single round-turn charge debited at open. Doubling that number to make it comparable produces a cost that does not exist.
Commission is also published in several different forms across the tested set: per side, per round turn, per million of notional value, and in some cases as a rebate rather than a charge. Several brokers quote it in US dollars on an otherwise Australian pricing page. Where the basis is not stated, a per-lot figure cannot be derived from the pricing page at all, which makes a like-for-like comparison impossible without opening the account.
Cost sensitivity also scales with frequency. A discretionary trader placing two positions a week barely feels a 0.2 pip difference. An automated system placing forty feels it every week, which is why the all-in figure matters more the more often the strategy trades.
The hours that get missed
Rollover lands at 5pm New York, which is 7am on the Australian east coast in southern winter. Every broker widens briefly while swap is recalculated. In the tested set the tighter raw accounts recovered within about 30 seconds. Some took several minutes.
Overnight pricing is worth checking separately. Some brokers apply a different spread outside main trading hours and define that window in GMT rather than local time, which can place part of it inside Australian working hours. The window and its time zone usually sit in a footnote on the pricing page rather than in the headline table, so it is worth reading the footnotes before funding an account rather than after.
The ASIC floor
Cost decides the comparison in Australia partly because the baseline protections do not vary.
ASIC’s product intervention order took effect on 29 March 2021 and was extended in April 2022 to run to 23 May 2027. It caps retail leverage between 30:1 and 2:1 by asset class, standardises margin close-out at 50%, requires negative balance protection, and bans inducements such as credits, rebates and gifts.
The measured effect was large. In the order’s first six months ASIC recorded a 91% fall in aggregate net retail losses, from an average of $372 million per quarter to $33 million. Negative balance occurrences fell 88%, margin close-outs fell 87%, and loss-making accounts fell 51%. The analysis behind the extension drew on data from more than 60 CFD issuers.
Client money sits in trust accounts under the Corporations Act, and Australia has no statutory compensation scheme covering the failure of a CFD issuer. These protections apply across ASIC-regulated providers, but they do not eliminate broker or counterparty risk, and they say nothing about a firm’s financial strength or operational controls. What they set is a floor that every licensed provider shares, which is why the comparison has to happen somewhere else.
Three checks before comparing on price
Establish what kind of number you are reading. Floor, average or typical, and over what window. If the methodology is not stated, do not assume the figure represents an average or a typical trading cost. BabyPips’ spread explainer covers how the quote is built in the first place.
Add the commission back, and check the basis first. Per side doubles; round turn does not. On a raw account the commission is usually the larger half of the total.
Compare at the session you actually trade. A broker that wins the London and New York overlap can lose the Sydney morning, and an average across both hides it. The forex market hours tool shows where your own trading hours sit against the four sessions.
The advertised number is what the broker chose to show. The tested number is what the server quoted. When the two disagree, only one of them appears on a statement.
