Inside Our Spread Sampling Protocol: Cadence, Venues, and Outlier Handling

A brokerage's advertised EUR/USD spread of 0.7 pips, while appealing on a promotional banner, often bears little resemblance to the actual cost incurred during a volatile market entry. This discrepancy is not merely an inconvenience; it represents a significant, often invisible, erosion of trading capital. At FX Auditor, our primary mandate is to strip away marketing hyperbole and present verifiable figures. We do this by implementing a stringent, multi-faceted spread sampling protocol designed to capture the granular reality of broker pricing, not just their marketing claims.

The industry frequently publishes "average spreads," a metric that is, by its very nature, prone to misinterpretation. An average spread might look attractive over a 24-hour period, yet fail to account for the crucial seconds during a news release when the spread widens to three or four times its typical value. This is the precise moment when many traders execute their entries or exits, transforming a seemingly low-cost environment into an unexpectedly expensive one. Our methodology prioritises capturing these critical moments, ensuring our assessments reflect genuine trading conditions.

Our approach begins with the fundamental understanding that spread is a dynamic variable, influenced by liquidity, volatility, and the broker's internal risk management. It is not a static number. To truly audit a broker's pricing, one must move beyond simple screenshots or self-reported data. We establish direct, high-frequency data feeds from live trading environments, allowing us to build a full picture of spread behaviour across various assets and market conditions. This groundwork is essential for any credible assessment of trading costs.

The timing and frequency of spread data collection are perhaps the most critical aspects of accurate auditing. A spread value sampled once per minute, or even every ten seconds, can completely miss fleeting but significant widenings that occur during high-impact market events. Consider the US Non-Farm Payrolls announcement: spreads on major pairs can flash from 1.0 pip to 10.0 pips and back within a three-second window. A lower sampling cadence will simply gloss over these spikes, presenting a misleadingly stable picture.

Our protocol mandates tick-by-tick data capture where feasible, or at a minimum, one-second interval sampling for core currency pairs such as EUR/USD, GBP/USD, and USD/JPY. This intense frequency provides the necessary granularity to detect micro-spikes and transient liquidity issues. We deploy automated agents directly on live trading terminals, often utilising APIs where available, to mirror the experience of an actual trader. This process runs continuously, 24 hours a day, five days a week, ensuring we cover all major trading sessions – Sydney, Tokyo, London, and New York.

We meticulously tag each data point with a precise timestamp, allowing for correlation with economic calendars. This enables us to differentiate between typical market behaviour and the predictable, albeit extreme, volatility surrounding scheduled news events. Without this temporal precision, any analysis of spread performance remains superficial, failing to inform traders about the real-world costs they will encounter when trading around critical macroeconomic releases.

The source of spread data is as important as its temporal resolution. Many brokers operate across multiple platforms and often in different data centres, each potentially offering slightly varied pricing. For instance, a broker might provide one set of spreads on their proprietary web platform and another, subtly different, set on their MetaTrader 4 (MT4) server, particularly if these are hosted in distinct geographical locations or managed by different liquidity providers. Our methodology accounts for this complexity by sampling from the actual trading environments where clients transact.

We establish dedicated, funded live trading accounts with each broker under review. This is non-negotiable. Data obtained from demo accounts, however convenient, frequently fails to replicate the real-time pricing and execution conditions of a live environment. Demo accounts may draw from aggregated feeds, lack the same server priority, or simply not be subject to the same liquidity constraints as production systems. A broker like Pepperstone, offering MT4, MT5, and TradingView, may exhibit nuances across these platforms that only live account data can reveal.

Crucially, we aim to connect to the broker's primary trading servers, often located in key financial data centres such as Equinix LD4 in London or NY4 in New York. This direct connection minimises the impact of intermediary routing and ensures we are capturing the spreads as close to the source as possible. Relying on aggregated data from third-party vendors or public websites simply introduces an unacceptable layer of abstraction, obfuscating the true operational spread.

Many brokers, including prominent ones like XM or AvaTrade, proudly display "average spreads" on their websites. While mathematically correct, an average spread is often a poor indicator of the actual trading cost. An average is highly susceptible to outliers: a single one-minute period where EUR/USD widens to 15 pips due to an unexpected geopolitical event can significantly inflate the 24-hour average, making the typical spread appear worse than it is. In contrast, a broker might strategically calculate averages during periods of extreme liquidity, such as the London-New York overlap, to present an artificially tight figure, while their spreads widen dramatically during Asian session hours.

To counter this statistical ambiguity, FX Auditor prioritises strong statistical measures. The median spread offers a more reliable representation of the "typical" cost of trading, as it is unaffected by extreme outliers. If a broker's EUR/USD spread is 0.7 pips for 99% of the time, but flashes to 15 pips for 1%, the median will still be 0.7 pips, whereas the average will be considerably higher, misrepresenting the regular experience.

Beyond the median, we examine percentile spreads. For instance, knowing the 90th percentile spread for EUR/USD tells a trader that 90% of the time, the spread was at or below that figure. This provides a fuller understanding of potential worst-case, yet still frequent, scenarios. A broker consistently offering a 90th percentile spread of 2.0 pips on EUR/USD, even if their median is 0.7, warrants closer inspection compared to one with a median of 0.8 but a 90th percentile of 1.2 pips. These strong metrics offer actionable insight, unlike the often-misleading single average figure.

Even with high-frequency data, raw spread information can contain anomalous readings that do not reflect genuine market conditions but rather data transmission errors, fleeting liquidity gaps, or even manipulative tactics. For instance, a single tick showing EUR/USD at 50 pips for a millisecond, then immediately reverting to 0.8 pips, is statistically an outlier and, if unaddressed, would severely distort any calculated average or percentile. Handling these outliers is a critical step in producing credible spread analysis.

Our protocol employs a multi-stage outlier detection and mitigation process. First, we establish logical boundaries for spreads on each currency pair, based on historical observations and expected maximums under extreme conditions. For EUR/USD, a spread exceeding 10 pips outside major news events would typically be flagged. Second, we implement statistical methods such as the Interquartile Range (IQR) rule, where any data point falling significantly outside 1.5 times the IQR from the first or third quartile is considered an outlier. This is the part most guides skip, often leading to skewed results.

Once identified, outliers are not simply deleted indiscriminately. Instead, we typically employ Winsorization, where extreme values are 'capped' or 'floored' at a specific percentile (e.g., the 99th or 1st percentile). This process retains the data point but limits its distorting influence, ensuring the statistical integrity of the dataset while acknowledging that even extreme values hold some informational content about market behaviour. For data errors, however, direct removal is warranted. This meticulous approach ensures that our reported spreads reflect typical, tradable conditions, not statistical flukes.

The financial markets operate on microseconds, and even a spread widening that lasts only a few seconds can have a material impact on a trader's profitability, particularly for those employing high-frequency or scalping strategies. Many traders focus on the "visible" spread, the number displayed on their platform for extended periods. However, the true cost can be dictated by fleeting "micro-spikes" that occur faster than human perception, often during periods of rapid price movement or thin liquidity.

Consider a trader attempting to execute a market order at a specific price. If, in the millisecond between the order being sent and executed, the broker's spread widens from 0.8 pips to 3.0 pips, the trader will experience higher slippage and a less favourable entry price. This increased cost, even if momentary, accumulates over numerous trades. Capturing these micro-spikes requires an extremely high data granularity, ideally tick-by-tick, where every single price update (bid and ask) is recorded.

Our collection agents are configured to log every bid/ask price change, not just snapshots at arbitrary intervals. This ensures that even a one-second widening is fully documented. The sheer volume of this data presents computational challenges, but it is a necessary burden for forensic accuracy. Without this level of detail, any assessment of spread competitiveness remains incomplete, failing to account for the hidden costs of trading within a dynamic, high-speed environment.

A broker's internal execution model fundamentally dictates the spreads they offer and how those spreads behave under varying market conditions. Broadly, these models fall into three categories: Market Maker, Straight-Through Processing (STP), and Electronic Communication Network (ECN). Understanding the distinction is vital for interpreting spread data.

Market Makers, like some iterations of Plus500 or eToro, typically internalise client orders. They quote both bid and ask prices and act as the counterparty to their clients' trades. This allows them to offer seemingly fixed or very tight variable spreads during calm periods, as they control their own pricing. However, during high volatility or low liquidity, Market Makers often widen spreads more aggressively and frequently than other models to manage their own risk, as they are taking on client positions. This can result in unexpected costs for traders when market movements are most pronounced.

STP brokers, by contrast, route client orders directly to external liquidity providers (LPs), without internalising them. This model typically features variable spreads that fluctuate with market conditions, as the broker passes on the best available bid and ask prices from their pool of LPs, often with a small markup. While this generally leads to more transparent and less manipulated spreads, the spreads can still widen significantly if the LPs themselves face liquidity issues.

ECN brokers, such as IC Markets or FxPro with their raw spread accounts, aim to connect traders directly to an interbank network, aggregating quotes from multiple LPs. They typically charge a commission per lot traded, offering "raw" spreads that can go as low as zero pips during peak liquidity. This model theoretically provides the tightest spreads, as the broker's profit comes from the commission, not the spread itself. However, even ECN spreads can widen during very thin market conditions, and the combined cost of raw spread plus commission must be considered. Our analysis always accounts for the specific execution model of each broker, as it shapes the context of their spread performance.

Regulatory bodies, while not directly setting spread values, exert significant influence over broker operating models, which in turn impacts pricing. The ESMA product intervention on CFDs in 2018 is a prime example. This intervention capped leverage for retail clients at 1:30 for major currency pairs, 1:20 for minors, and even lower for other assets. While the primary aim was investor protection through reduced risk exposure, it had indirect consequences for broker spread strategies.

With reduced leverage, the potential profit per client trade for brokers also diminishes, as clients are trading smaller notional values. To compensate for this, some brokers might opt to slightly widen their spreads or introduce other fees to maintain profitability. However, the increased regulatory scrutiny, particularly from bodies like the FCA in the UK or ASIC in Australia, can also push brokers like OANDA or Pepperstone to maintain highly competitive spreads to attract volume in a more transparent and compliant environment. The FCA's stringent oversight and requirement for 'best execution' policies compel brokers to continuously seek the best available prices for clients.

The regulatory environment creates a complex dynamic. Brokers operating under strict regimes (e.g., FCA-regulated entities like FxPro or Exness in certain jurisdictions) are under constant pressure to demonstrate fair pricing and transparent practices. Those regulated in less stringent jurisdictions might face fewer direct constraints on their spread behaviour. Our analysis takes into account the regulatory framework each broker operates under, recognising that compliance costs and market pressures from regulators play a role in their ultimate spread offering.

Conducting a rigorous, reproducible spread audit requires a precise, multi-step protocol. Our process is designed to be as objective and thorough as possible, mimicking the real-world experience of a trader while employing scientific data collection and analysis methods.

1. Broker Selection and Account Establishment: We select a diverse range of brokers based on market share, regulatory standing (e.g., FCA, ASIC, CySEC, NFA registrations checked via the Financial Services Register or NFA BASIC), and reported client base. For each selected broker – say, FOREX.com, AvaTrade, and Exness – we open a dedicated, funded live trading account. This ensures we are capturing data from their genuine production environments, not demo accounts or marketing simulations.
2. Platform Integration and Automated Capture: Our technical team integrates custom data collection agents directly with the broker's trading platform. For MetaTrader 4/5, this involves Expert Advisors (EAs) that log every tick (bid/ask price change) to a local database. For brokers offering proprietary APIs, such as OANDA, we develop direct API connections for even finer granularity and efficiency. This automation runs 24/5 from multiple geographical locations to mitigate localised network issues.
3. Data Synchronisation and Storage: All collected tick data, comprising instrument, timestamp (to the millisecond), bid price, and ask price, is securely transmitted to our central database. This data is then meticulously timestamp-synchronised to account for any minor discrepancies between different broker feeds or data collection points, ensuring an 'apples-to-apples' comparison.
4. Initial Cleaning and Validation: Before analysis, the raw data undergoes an initial cleaning phase. This includes removing duplicate entries, correcting any obvious data corruption (e.g., non-numeric values), and validating timestamps for chronological integrity. We also cross-reference with major news announcements from sources like the US Bureau of Labor Statistics for Employment Situations or the ECB for euro reference rates to contextualise price movements.
5. Outlier Detection and Treatment: As detailed previously, this crucial step involves identifying and addressing extreme spread values that fall outside statistically defined thresholds. We employ Winsorization, capping values at the 99.5th percentile for each currency pair to mitigate the distorting effect of transient spikes without discarding valuable information entirely.
6. Statistical Analysis and Reporting: The cleaned data is then subjected to rigorous statistical analysis. We calculate not only the average spread but, more importantly, the median spread and various percentile spreads (e.g., 25th, 75th, 90th percentiles). This provides a full view of typical, better-than-average, and worse-than-average spread conditions. This analysis is performed across different trading sessions and market volatility levels.

This multi-stage protocol ensures that our spread analysis is based on factual, verifiable data, reflecting the genuine trading conditions clients encounter.

While granular spread data forms the bedrock of our analysis, a full broker audit extends beyond mere pricing. The overall reliability and operational integrity of a broker are equally critical to a trader's success. A broker might offer tight spreads but fall short in other areas, negating any perceived cost advantage.

One often-overlooked factor is withdrawal efficiency. We regularly initiate withdrawals from our live accounts to assess the actual processing times. For instance, a withdrawal from a Pepperstone account to a UK bank typically clears within 1-2 business days. In contrast, some brokers, particularly those regulated in less stringent jurisdictions, might exhibit delays of 5-7 business days, or even longer, often citing 'compliance reviews' as a reason. This is a practical consequence that impacts a trader's capital accessibility. The desk will often ask twice for documentation they already possess.

Customer support responsiveness and effectiveness are also weighed. Can a trader reliably contact support during peak market hours? Are issues resolved promptly and competently? Our team periodically interacts with broker support channels, assessing their knowledge, politeness, and resolution speed. The regulatory environment is highly important. Brokers like FxPro, regulated by the FCA, CySEC, and FSCA, offer a higher degree of investor protection compared to those with fewer or less reputable licences. We consult registers such as the FCA Financial Services Register, ASIC's professional registers, and NFA BASIC to verify regulatory standing and any disciplinary actions. A broker's regulatory status provides a foundational layer of trust, without which even the tightest spreads are of little comfort.

The current state of spread reporting in the retail forex industry remains fragmented and, at times, deliberately opaque. While some progress has been made, particularly among well-regulated entities, a standardised methodology for reporting real-time trading costs is conspicuously absent. This lack of uniformity allows for continued marketing distortions and makes genuine comparative analysis unnecessarily complex for the average trader.

The industry would benefit significantly from a consensus on key metrics beyond the simple "average spread." Adoption of median spreads and specific percentile spreads (e.g., 25th, 75th, 90th) as standard reporting metrics, alongside clear disclosure of sampling cadence and outlier handling procedures, would increase transparency considerably. Regulatory bodies, such as ESMA or the FCA, could play a more active role in mandating these reporting standards, much as they have intervened on leverage and negative balance protection.

Looking forward, advancements in technology could offer further solutions. Distributed ledger technology, while still maturing for this specific application, holds the theoretical promise of immutable, verifiable spread records. Imagine a system where every tick from a broker is cryptographically recorded on a public or semi-public ledger, making retrospective auditing trivial and unassailable. While such a system is not yet commonplace, the direction of travel should be towards greater, not lesser, transparency. Until then, meticulous independent auditing remains the most reliable means of discerning true trading costs from marketing fiction.