Calendars, scanners, calculators and news aggregators

A macro calendar helps you prepare for data releases, a scanner narrows the instruments to watch, and a calculator works out position size, margin and costs. Using these tools starts with checking the data, units and assumptions. A different time zone, a revised release or a different contract specification can change the result without an obvious error. Below, we explain calendar fields and scanner filters, work through calculations and check a news channel. We also clarify their limits: the tools alone do not establish an edge or guarantee a limit on losses.

Educational illustration of tool organization; it is not a screenshot of a real account or the current interface.

Identify the comparison before interpreting the colour

A calendar row can be above the forecast and below the previous figure at the same time. There is no contradiction: the two comparisons use different reference numbers. Before treating a coloured cell as a signal, identify the statistic, period, units and version of every input. A percentage change, an index level and a change in persons cannot be substituted for each other.

Our fictional release reports a monthly employment change in persons. It is not an actual BLS release or a sample of market behaviour. Immediately before publication, a calendar has saved a consensus of +125,000 at 13:29 UTC. At 13:30 UTC, the issuer publishes an actual change of +142,000 and revises the preceding month from +180,000 to +160,000. The next table keeps these quantities separate instead of asking one colour to explain them all.

Make each subtraction answer a named question

Hypothetical release; all quantities are changes in persons
QuestionCalculationResult
Actual minus saved prerelease consensus142,000−125,000+17,000 persons
Revision to preceding month160,000−180,000−20,000 persons
Actual minus revised preceding month142,000−160,000−18,000 persons
Actual minus unrevised preceding month142,000−180,000−38,000 persons

The first row is the surprise relative to that saved consensus. The second measures the revision to an earlier estimate. The last two compare successive monthly changes, using different versions of the preceding value. They are not changes in the level of total employment. Keep the signs and units in the result; writing only “17” would leave the scale ambiguous.

BLS technical notes describe how preliminary CES estimates are revised as more sample information arrives. That primary source explains why version tracking matters; it is not the source of our invented figures. In an actual release, read the issuer’s definitions and revision notes rather than assuming every calendar provider labels the columns identically.

Preserve what was visible before the release

Suppose that at 13:31 UTC the calendar displays a consensus of +140,000. Subtracting it from the actual gives +2,000. That arithmetic is valid for those two numbers, but the later display does not replace the +125,000 snapshot as evidence of what this calendar showed before publication. It might reflect a data correction or another version; the scenario does not establish why it changed.

Save the provider, field definition, displayed timestamp and available version alongside the issuer’s release. A later clean-looking table is not automatically a faithful record of the information available earlier. Conversely, do not label the revision an error just because it changes a calculation. Record the correction and retain the earlier state, so that both the original comparison and the updated data can be reconstructed.

Let the calendar, scanner and calculator answer different questions

The fictional release is scheduled for 13:30 UTC. A viewer explicitly set to a fixed UTC+1 offset shows 14:30. This is an offset calculation, not a rule for a named country or a promise that daylight-saving changes are handled correctly. Check the actual event time and display settings before relying on a countdown. A scanner detecting a price move answers a different question from a calendar listing an event.

A calculator can verify the subtraction only after you select the right inputs. It cannot establish which forecast was known in advance or whether a surprise caused a currency move. Nor does this example determine the direction of a trade: revisions, other components, expectations and contemporaneous news may differ. Use the detailed lesson below to compare tools by their sources and limitations. The useful outcome here is a small release record whose comparisons can be repeated without silently changing the reference point.

Three exercises with worked answers

The exercise data are hypothetical. Choose an answer, then check the calculation.

1. What is the surprise relative to the saved prerelease consensus?

Actual +142,000 persons; consensus saved before release +125,000 persons.

Show the answer and explanation

C. Name the reference before subtracting. A surprise against this consensus and a change against last month are different comparisons.

  • A. That uses the revised preceding month rather than the consensus.
  • B. That uses the later displayed consensus, not the saved prerelease value.
  • C. Correct: 142,000−125,000=+17,000 persons.
  • D. That is the separate revision to the preceding month.
2. How large is the revision to the preceding month?

The preceding monthly change moves from +180,000 to +160,000 persons.

Show the answer and explanation

A. Compare two versions of the same preceding-month estimate. Neither number is the current month’s actual.

  • A. Correct: revised minus initial is 160,000−180,000=−20,000 persons.
  • B. That reverses the direction of the revision.
  • C. That compares the current actual with the old preceding value.
  • D. That is the surprise against the saved forecast.
3. What can be concluded from the later displayed consensus?

A +125,000 consensus was saved at 13:29; release at 13:30; the display shows +140,000 at 13:31 UTC.

Show the answer and explanation

D. 142,000−140,000=+2,000 persons describes the later comparison. It cannot silently replace the prerelease comparison in a historical evaluation.

  • A. A later observation alone does not establish the earlier information state.
  • B. Deleting it would remove evidence of what was displayed before publication.
  • C. A version change supplies no directional trading result.
  • D. Correct: preserve the versions and the information timing.

Continue to tool selection and source checks

The original illustrations and detailed tool discussion remain below. Keep the release record’s units, comparison labels and information times when testing any calendar or calculator.

1. A calendar entry: who owns which number

Open two macro calendars side by side on the same release. The time — provided you set the same time zone in both services — will be identical, the indicator name almost identical, and the forecast quite possibly different. That is not an error in either of them. A single calendar row looks like one piece of information, but in reality it glues together numbers originating from three different parties, each responsible for something else and each revising its data in a different way.

The actual is published by a statistical institution or a central bank: for United States non-farm payrolls that is the Bureau of Labor Statistics, for euro-area inflation Eurostat, for Polish data Statistics Poland, and for rate decisions the relevant central bank[3]. The forecast does not come from nowhere — it is the result of a survey of economists, run either by a news agency or by the calendar provider itself, with a specific cut-off date for collecting responses. The impact label, the row colour and the rounding convention are, by contrast, purely the decision of the service displaying the calendar.

One macro calendar row and the three parties responsible for its fields Six fields of a calendar row connected by lines to three owners: the actual and the previous value come from a statistical office or central bank, the forecast from an agency survey, and the time, currency and impact label from the calendar provider. Fields of a single calendar row Actual Previous Forecast Time Currency Impact Statistical office or central bank Forecast survey with a collection cut-off date Calendar provider label, colour, rounding Six fields, three owners — each with a different way of introducing corrections. Compare the period, unit and data vintage before treating a discrepancy as an error.
A calendar row looks like one piece of information, yet it combines data from three independent parties.

If two services show different actuals, first compare the indicator, reference period, unit, seasonal adjustment and data vintage. A discrepancy can reflect a revision, a different series or an update delay rather than an error alone. The appropriate institutional release and its publication date are the reference. Different forecasts can reflect different surveys or models and collection cut-off dates. Different impact ratings are providers’ classifications; check their definitions rather than treating colours as competing versions of the actual.

The criterion by which I compare calendars in this article. I do not judge which one is "better", because the answer depends on what you want to do with it — and a ranking without a stated criterion is merely the author's preference. I judge one verifiable thing: whether the calendar alone lets you reconstruct the provenance of every number in the row. The five control questions below apply to any provider, including one that appears after this text is published.

Five questions you ask a calendar before first use
Control questionWhat exactly you are checkingHow to check itWhat a "no" means
Does the entry name the primary source?The name of the institution publishing the actual and a link to its releaseExpand any high-impact entry and look for the office's nameYou cannot verify the actual without searching manually
Can you see the pre-revision value?Whether the original value appears next to "previous" when it has been revisedOpen the history of an indicator where revisions are the rule, e.g. US payrollsYour history of actuals changes silently between one visit and the next
Is the forecast methodology stated?Who runs the survey, how many economists, when collection closedLook for a methodology description, not just the numberForecast discrepancies between services cannot be explained
Is the time zone explicit and settable?Which zone the row is rendered in and whether the change persistsChange the zone in settings, log out, come back and check againThe time on screen is uncertain exactly when it matters
Can the state of an entry be exported?Export to a file or a programmatic interfaceTry saving the week view ahead of a run of releasesAutomated archiving becomes harder; a pre-release state can still be captured manually

The same set of fields is served by, among others, Forex Factory, Investing.com, Econoday and the calendar built into your charting platform. I do not give a blanket statement about free availability for them, because such a statement would not be true. The access model is a property of an individual provider, changes independently of this text, and has to be checked with each one separately, on the day you start using the tool. A single counter-example is enough to refute the generalisation: Econoday's official offering page presents its calendar as an "Online Calendar Subscription" and does not confirm that a free version exists [AS AT: 2026-08-22, econoday.com/retail-investor-solutions]. The durable rule is therefore this: check the five questions above and the access model against the current version of the service, not against a description from a year ago — including this one.

A calendar is not a source, it is an intermediate layer. If a given number is to be the basis of anything more serious than orientation within the day — a journal entry, a spreadsheet parameter, material for analysis — take it from the institution's release, and use the calendar for what it does best: knowing that something will be published at all, and at what time. What a given release means for a currency's valuation is covered separately in the article on the key macroeconomic reports, and the site's own economic calendar follows the same convention: it presents the release schedule and does not replace the primary release.
Concept illustration: Calendar: source, time zone, label
Educational illustration of tool organization; it is not a screenshot of a real account or the current interface.

2. Actual, consensus, previous, revised — four fields and four conventions

Four numeric fields look like four variants of the same quantity. In reality they differ not only in value, but also in how they came about and at what moment they stop changing.

Actual is the value published by the reporting institution. It is a fact in the sense that it can be verified against the source release — but it is not a final value if the indicator is subject to revision. For a first estimate of gross domestic product, the word "actual" means only that this is the freshest available estimate.

Consensus, described by some providers as forecast, is a survey statistic: the median or the mean of forecasts, depending on the methodology the provider has adopted. The number of economists in the panel, the day collection closes and the treatment of outliers are properties of the methodology, not of the market. That is why two honest calendars can show two different forecasts for the same release and both will be correct. If you want to compare deviations across indicators or across months, you must use the same forecast provider throughout — otherwise you are comparing two different measures.

Previous is the actual from the preceding reporting period. The key question is: in which version. Services usually show the current value, that is, the value after any revision, with no information about how it originally looked.

Revised explicitly marks a correction to an earlier reading. The provider decides how to present it. A missing separate field does not establish whether a revision occurred or whether the service omitted it; check the vintage in the primary release.

The unit and the convention are part of the number. Before you compare an actual with a forecast, check four things in the row: the unit (per cent, index point, thousands of jobs, billions in a currency), the comparison base (month on month or year on year), the annualisation method for quarterly data, and whether the series is seasonally adjusted. An inflation reading of "0.3" and one of "3.6" can describe exactly the same release — the first month on month, the second year on year. A calendar usually gives this in small print next to the indicator name, not in the value column.

A separate category is the split between headline and core values. For CPI inflation, services usually publish both series in separate rows, but the names are sometimes abbreviated so heavily that only the acronym survives in a mobile view. If you are building your own record of deviations, enter the full series name together with its unit — after three months you will not reconstruct from memory whether "CPI 0.4" referred to the full basket or the core one.

Actual, Forecast, Previous, Revised. Actuals and forecasts are comparable only after metadata alignment.
Illustrative month-on-month CPI. A missing Revised field does not establish whether a revision occurred.

What those four numbers mean for a currency's valuation, how the market weighs a surprise relative to expectations and why the reaction is sometimes counter-intuitive — that is the subject of the article on data releases in practice. Here we stop at the tool layer: you can read the row and you know what is missing from it.

3. Time, zone and data revisions — three places where a calendar misleads

The time in a calendar row is a rendering of a timestamp, not the event itself. A particular release has one publication instant, but its schedule can change; changing the display zone changes how that time in which the service displays it changes — sometimes the one the service has decided is yours, sometimes the source's zone, and sometimes both, in different views of the same service. As long as both zones change their clocks on the same date, the conversion is stable and nobody thinks about it. Twice a year it stops being stable.

A large share of American macro releases — including the employment report and the consumer price index — falls at 8:30 a.m. New York time; the Bureau of Labor Statistics schedule gives the hour "08:30 AM" against every employment report release date[3] — but this is not a rule covering every reading: business surveys, housing market data and consumer sentiment studies have their own hours, and the statement following a central bank meeting comes in the afternoon. So you take the time from the calendar row, not from memory. For an 8:30 release the conversion looks like this: in winter New York runs on UTC−5 and Warsaw on UTC+1, so the release falls at 14:30 Polish time. In summer New York moves to UTC−4 and Warsaw to UTC+2 — the difference is still six hours and the clock time stays the same. The problem appears in the window where one side has already changed its clocks and the other has not.

Assumptions of the conversion: a release time of 8:30 local New York time, target zone Warsaw, result unit — clock time in Poland. The control is a pass through universal time, performed separately for each row.

An 8:30 New York release converted to Polish time
PeriodZone in New YorkZone in WarsawDifferenceTime in Poland
WinterUTC−5UTC+16 hours14:30
SummerUTC−4UTC+26 hours14:30
Divergence windowUTC−4UTC+15 hours13:30

The universal-time control closes the calculation: 8:30 at an offset of UTC−4 is 12:30 UTC, and 12:30 UTC at an offset of UTC+1 is 13:30 in Poland. The United States moves to daylight saving time on the second Sunday of March and returns on the first Sunday of November; the European Union changes its clocks on the last Sundays of March and October. For an 08:30 New York event in 2026, the inclusive dates giving 13:30 in Warsaw are 8–28 March (21 calendar dates) and 25–31 October (7 calendar dates). On 29 March and 1 November, the same 08:30 New York event converts to 14:30 in Warsaw. These ranges describe this event time, rather than entire transition days.

Why this is not a technical curiosity. For those 28 calendar dates in this 2026 conversion, the 08:30 New York event falls one hour earlier in Warsaw than it does outside the mismatch dates. If your daily plan rests on a remembered hour instead of on a calendar row set to the right zone, for those mismatch dates you are planning next to reality. The fix is a one-off: set your local zone in the calendar, check after logging out that the setting survived, and plan by the displayed time rather than the remembered one. The source of the publication moment remains the institution's schedule — the calendar merely presents it. How time zones arrange themselves into the trading day is covered in the article on the structure of the 24-hour market.

The second source of mistakes is the latency of the row itself. A calendar updates the actual field after its system has pulled the value from the source. The gap between the institution's release and the number appearing in the service depends on the provider, the hour and the load, and is not guaranteed by anyone. If that difference matters in your process, measure it yourself — the measurement procedure is described in section eight, because it applies to calendars just as it does to news services.

The third source is revisions. Statistical institutions routinely correct earlier readings as fuller data arrive, and once a year they carry out broader methodological reviews covering long series back through time. The practical effect is that a historical series downloaded today need not be identical to the series downloaded six months ago — and that is not an error, merely a normal feature of public statistics.

If you are building your own record of deviations. Record the forecast and the previous value before the release, together with the date and time of the entry and the provider's name. After the release, add the actual. Without that step, after a few months you will be comparing today's revised historical values against forecasts that were produced in a different informational reality — and you will end up with a systematic distortion whose source is your own database, not the market. It is the same trap that in strategy research goes by the name of using information that was not available at the moment of decision.

4. "Impact" is a provider's label, not a property of the market

A red row in a calendar is not a volatility forecast. It is an editorial classification: the service has judged that a given release tends to matter for quotations and has assigned it a category. With some providers the classification is moreover attached to the indicator permanently, so the same release gets the same label regardless of the environment it appears in. Whether your calendar works that way is something you check by comparing several consecutive occurrences of the same indicator — it is a property of a specific service, not a rule for the whole category.

Providers can assign different impact ratings to the same event under their own criteria. A rating guarantees neither a large move nor its direction; low priority does not establish an absence of risk either. Research describes elevated activity or volatility around particular releases in the periods studied [4][5], rather than a certain move after the next announcement. Chaboud and co-authors studied EBS trades in EUR/USD and USD/JPY: activity also rose after releases matching expectations [6]. That is a bounded empirical finding, not a prediction from one label. The Forecast field does predict an indicator’s value, so it is incorrect to say that a calendar contains no forecasts at all.

How to test the label on your own data. A label is a claim, so it can be tested — provided you design the measurement before you carry it out. For a chosen indicator and a single currency pair, read from your own platform the range of the five-minute candle covering the release and divide it by the median range of five-minute candles from the same hour on days without a release. The ratio from a single release is a descriptive observation and settles nothing: it does not say whether the label "is backed up", because a single measurement without a pre-planned sample, without a criterion fixed in advance and without a measure of dispersion has nothing to settle. For anything more to follow from it, you must record before measuring: the number of releases in the sample, the width of the window, the way control days are selected, the decision criterion and the measure of uncertainty — and at that point it is a measurement design, not the reading of a single candle. Together with the result, record the environment: the broker's name, the account type, the symbol including its suffix, the interval, the server time zone and the number of observations — without those fields the number is not comparable with anything, not even with your own measurement from six months ago.

And that is where the tool-level part of the matter ends. How to design such a measurement, how many observations are needed for anything to follow from it, how to control for the size of the surprise and how the exchange rate behaves around data — that is material for the article on data releases in practice. The mechanics of spread widening and the conditions of order execution at such moments belong to the article on execution models.

What a calendar does not contain. Besides macro releases, volatility is affected by purely market-driven dates: contract and option expiries, month-end flows, daily benchmark fixings. A standard macro calendar does not show them, because they are not data releases. If they matter in your process, they must reach your plan from another source — for example from the instrument specification at your broker and from the calendar of the exchange on which the underlying instrument is listed.

5. Scanners: filter, condition, universe and the multiple-testing trap

A scanner does one thing: it checks a recorded condition across a given set of instruments and returns those that satisfy it. All of its usefulness comes from doing that faster than a human scrolling through charts. All of its unreliability comes from the fact that three elements of that definition usually go unstated.

Universe is the set of instruments covered by the scan. It is not obvious: a charting platform's scanner operates on the data provider's symbols, not on your broker's symbols, and those sets need not overlap. An instrument your broker does not offer will appear in the results; an instrument available at your broker under an unusual suffix may not appear at all.

A watchlist is a saved subset of the universe, not the result of a scan. A watchlist is a function of the platform or service in which you keep it — so it inherits that platform's symbol set, its time zone and its data source, not your broker's settings. It guarantees nothing beyond the chosen instruments being displayed together: on its own it checks no condition, does not police the freshness of an entry and does not remember why something ended up on it. That is why every watchlist position needs two manually added fields — the date it was added and the criterion — because after six months the instrument's name alone will not reconstruct that reason. The same applies to saved filters: a filter without a date and without a noted purpose is, after a few months, a set of parameters about which the only known fact is that it once seemed sensible. The organisation of watchlists themselves — dividing them by purpose of observation and keeping a sensible list size — is covered in the article on TradingView.

The condition is a formal statement of a rule. "Elevated volatility" is not a condition — a condition is, for example, ATR over a period of 14 on the four-hour interval above its own 20-session average. Only such a statement can be repeated, checked and challenged.

Refresh frequency decides whether the result describes the present or the state of a quarter of an hour ago. Scanners running on delayed data or recomputing the list every few minutes are entirely usable — provided you know which moment the list you are looking at refers to.

There is also a fourth element that the manuals are silent about: a scanner returns a set, not an ordering. If the tool sorts the results, the sorting is a separate decision and a separate claim. A sorted list looks like a ranking by relevance, while it is only a list ordered by one column that you chose.

The multiple-testing trap. A scanner lets you check dozens of conditions in a quarter of an hour. If you screen fifty rules over the same history and keep the one that performed best, the best result alone does not establish the condition’s quality — the risk of fitting noise increases. The more variants you test on the same sample, the greater the chance that the best of them is best by accident. The safeguard is easy to describe and dull to carry out: record the number of variants you will test before you start, and check the winner on a period you have not looked at before. The methodology of such a check belongs to the material on algorithmic trading; here it is enough to be aware that the tool makes it easier to commit this error faster than ever before.

The criterion for comparing scanners. Again I do not judge which is best — I judge whether its result can be reconstructed from it. Four questions: does the tool let you see the full definition of the condition, does it state the set of scanned instruments explicitly, does it give the moment of the last recalculation, and can the result list be exported together with the values that determined it. Those four questions assess exactly one thing: whether the tool's result can be reproduced. They do not check the correctness of the calculations themselves, the completeness of the data set or the stability of the service — those three things have to be judged separately and none of the questions above substitutes for them.

Four control questions for any scanner
QuestionWhy it mattersSymptom of a problem
Can you see the full definition of the condition?Without it you cannot tell a filter from a ready-made opinion held by the tool's authorA description such as "strong trend" with no parameters given
Do you know the set of scanned instruments?Results may contain symbols unavailable at your brokerAn instrument you do not have in your platform appears on the list
Is the recalculation moment stated?A list from a quarter of an hour ago describes a different market than the current oneNo timestamp next to the results
Can the result be exported with its values?Without the values you cannot check why an instrument made the listThe export contains only instrument names

Tools in this category include the scanner built into a charting platform, the free volatility and currency-strength summaries published by journal-style services, and a spreadsheet built by yourself on data pulled from the platform. The last option tends to be underrated, and it has one substantial advantage: you know the definition of the condition, because you wrote it. The feature scope and pricing model of ready-made tools cannot be described in one sentence for the whole category: sometimes basic filters are free while elaborate conditions, longer history and export require a paid plan, but sometimes the whole thing is paid. Check current limits with the provider rather than relying on a table in any article, including this one.

USD strength: an explicit seven-pair score. Illustrative average score with signed percentage changes.
This is a defined score. An exact reciprocal-pair return is −r/(1+r), with r expressed as a decimal.
Tick volume is not turnover. There is no centralised volume on the spot currency market, so a field labelled "Volume" in a platform usually counts price events rather than the size of turnover. An "elevated volume" filter therefore operates on a quoting-activity indicator, not on the flow of money. That does not disqualify such a filter, but it changes its interpretation — and it is another example of the same principle: the unit is part of the number.

A currency-strength measure needs an explicit definition. Here it is a defined average score across seven pairs: each percentage change is assigned a positive sign when it supports the currency being measured and a negative sign otherwise. It is not the exact return of an inverted pair. For USD, using illustrative daily changes of EUR/USD −0.30%, GBP/USD −0.20%, AUD/USD −0.45%, NZD/USD −0.35%, USD/JPY +0.50%, USD/CHF +0.25% and USD/CAD +0.15%, the contributions total 2.20 percentage points, averaging 2.20 ÷ 7 ≈ 0.3143%. The rounding check is 0.3143 × 7 = 2.2001 ≈ 2.20. An actual reciprocal exchange-rate return is −r/(1+r), with r expressed as a decimal; log returns reverse sign exactly.

The same currency computed on a different set of pairs, over a different window or on logarithmic rather than percentage changes will give a different number. That is why the readings of two different currency strength meters are not comparable until you have checked both definitions — and a ready-made tool is under no obligation to publish its definition, and sometimes does not. The idea of comparing currencies through the prism of many pairs is developed in the article on comparative currency strength analysis.

Two contexts a scanner displays that are easy to mistake for a signal: correlations between pairs and retail client positioning. If a filter returns several pairs sharing the same currency on the same side at once, those are not independent positions: they share a risk component through that one currency, while retaining separate exposures to the currencies on the other side of each pair. How much of that remains net has to be calculated — the mechanics and the method are covered in the article on correlations in practice. The share of clients on a given side of the market is a separate indicator with its own limitations and belongs to the material on retail positioning. In a scanner both fields serve as context, not as an entry condition.

6. Calculators: units, assumptions and seven fields to check

A position, pip, margin or swap calculator does no difficult mathematics. It performs multiplication and division on parameters it does not itself know — it takes them from the form or from the tool author's default settings. Correctness requires appropriate inputs, a matching formula and a correct implementation. Even a correct calculation does not predict an account’s outcome. In the Polish Financial Supervision Authority’s 2025 report, 72.2% of active clients at the domestic investment firms surveyed recorded a loss, excluding clients with neither a profit nor a loss; retail clients accounted for 99.9% of active clients [1]. These figures describe that Polish sample and period, rather than predicting a calculator user’s outcome.

The error at stake here is not that a calculator computes badly. It is that it computes correctly for an instrument with a different specification from the one you have in your platform. Contract size is sometimes non-standard, the account currency changes the conversion factor, the required margin depends on client status, and the swap rate is a parameter of a specific broker, changed without notice.

Before you trust a calculator, check 7 fields
  1. Symbol — exactly in the form your broker uses, including the suffix. The same abbreviation at two providers can denote an instrument with a different specification.
  2. Contract and pip/tick size — how many units of the base currency go into one lot. A standard value is not a guaranteed value; check it in the instrument specification together with pip/tick size and the profit calculation method.
  3. Account currency — it decides whether the result requires conversion at all, and at what rate.
  4. Conversion rate — which rate, and from which moment, the tool substitutes. On pairs where the quote currency differs from the account currency, this is often the source of discrepancies between calculators.
  5. Leverage or margin rate — the value applying to your client status and your instrument class, not the default in the form.
  6. Commission and swap — if they apply to your account type. Check which costs the tool includes and its assumptions about slippage and SL execution. A missing field does not itself establish that a cost is absent from the model.
  7. Result unit — standard lots, units of the base currency, or contracts. The same number in a different unit can imply different exposure. Also check the volume step, minimum and maximum; for a risk limit, round down rather than above budget.

Two of these fields are less obvious than the rest for a European retail reader. The first is the margin rate: for a retail client trading contracts for difference on major currency pairs, a leverage limit of 30 to 1 applies in Poland, stemming from the 2018 product intervention of the European Securities and Markets Authority and subsequently made permanent by decision no. DAS.456.2.2019 of the Polish Financial Supervision Authority of 1 August 2019[2]. For a client with experienced status, a lower initial margin is a possibility on the provider's side, not an automatic entitlement: the same decision states that providers "may therefore offer experienced clients contracts for difference with higher leverage on selected contract types, while maintaining the remaining measures protecting against excessive loss"[2] (p. 37). If your calculator holds a leverage value that your broker agreement does not confirm, the result will be systematically understated or overstated. The second is the account currency: an account may be denominated in one currency while foreign calculators default to the dollar.

How to verify a calculator's result. Not by comparing it with a second calculator — two tools with different default parameters will give two different numbers and you will not learn which is right. Verification looks different: compute one case by hand from an explicit formula, check the result the other way round, and only then compare it with the calculator. If the numbers diverge, check the seven fields, rounding, formula and implementation. Four such worked checks, with all their units and assumptions, plus a fifth added as an addendum on the break-even threshold, follow in the next section; ready-made tools for daily work are collected separately in the site's position size calculator, pip value calculator, margin calculator and swap calculator, while the full mechanics of pips, lots and margin are covered in the article on units of measurement and the article on leverage and margin.

From risk budget to volume. EUR/USD · USD · contract 100,000 · pip 0.0001 · costs excluded.
Round down to the step; check min/max and free margin. A stop does not guarantee a USD 100 loss.

7. Four worked checks and one addendum

I computed every number below separately and verified it in three relations: against my own formula, against the sum of its components or an alternative route, and backwards, by reconstructing the inputs from the result. The exchange rate values are illustrative and stated explicitly, because the outcome depends on them.

Check one: pip value

Assumptions: a standard lot equal to 100,000 units of the base currency, a pip equal to 0.0001 of the quote currency for typical non-JPY pairs (including four-decimal quotes) and 0.01 for yen pairs, an account denominated in dollars. Formula: the pip value in the quote currency is the contract size multiplied by the pip size, then converted into the account currency.

The value of one pip per standard lot, USD account
PairCalculation in the quote currencyConversion ratePip value in USDBackward check
EUR/USD100,000 × 0.0001 = 10.00 USDnot needed10.00 USD10.00 × 50 pips = 500.00 USD
EUR/GBP100,000 × 0.0001 = 10.00 GBPGBP/USD 1.270012.70 USD12.70 ÷ 1.2700 = 10.00 GBP
EUR/GBP100,000 × 0.0001 = 10.00 GBPGBP/USD 1.310013.10 USD13.10 ÷ 1.3100 = 10.00 GBP
yen pair100,000 × 0.01 = 1,000 JPYUSD/JPY 150.006.6667 USD6.6667 × 150.00 = 1,000 JPY
yen pair100,000 × 0.01 = 1,000 JPYUSD/JPY 155.006.4516 USD6.4516 × 155.00 = 1,000 JPY

The "ten dollars a pip" repeated in guides is among the displayed cases, true in the first row of the table and only for a dollar-denominated account. On EUR/GBP, a change in GBP/USD from 1.2700 to 1.3100 changes the pip value by 0.40 USD, that is by 3.15% relative to the variant at 1.2700. On yen pairs, moving from a rate of 150.00 to 155.00 changes the pip value by 3.23% relative to the variant at 150.00 — while the difference between the rates themselves is 3.33%, so the two numbers must not be confused. The result depends on the cross rate and on the contract size — it is not universal and must not be quoted as a constant.

Check two: position size

Position size = Capital × risk percentageProtective order distance in pips × pip value per lot

Assumptions: capital of 10,000 USD, risk of 1.00%, that is 100.00 USD, a protective order distance of 50 pips, a broker volume step of 0.01 lots, commission and swap omitted and flagged as omitted. Result unit: standard lots. Neither leverage nor the margin rate appears in this formula — they decide whether the computed position can be opened at all, not how large it is.

The same risk rule, three different position sizes
Instrument and ratePip valueCalculationResultBackward check
EUR/USD10.0000 USD100 ÷ (50 × 10.0000)0.20 lots0.20 × 50 × 10.0000 = 100.00 USD = 1.00% of capital
yen pair, USD/JPY 150.006.6667 USD100 ÷ (50 × 6.6667)0.30 lots0.30 × 50 × 6.6667 = 100.00 USD = 1.00% of capital
yen pair, USD/JPY 155.006.4516 USD100 ÷ (50 × 6.4516)0.31 lots0.31 × 50 × 6.4516 = 100.00 USD = 1.00% of capital

Identical capital, an identical risk percentage and an identical protective order distance give 0.20 or 0.31 lots. In this example, the difference follows from the pip value, that is in the cross rate and the contract size. This is exactly why entering a "round" volume from memory stops working the moment you change pair. The full methodology of choosing position size, limits and a risk budget belongs to the article on position sizing — here we care only about whether you can check a tool's result.

Check three: margin, margin level and the close-out threshold

Assumptions: an account of 5,000 EUR with no open result at the start, leverage of 30 to 1, EUR/USD 1.0850, GBP/USD 1.2700, from which the GBP/EUR rate is 1.2700 ÷ 1.0850 = 1.170507. Margin is computed from the notional in the instrument's base currency and converted into the account currency. The components in the table are rounded to the cent, while the totals are computed on exact values — which is why adding up only the displayed components can give a result differing by a cent from the total row.

Three positions and the effect of each on the margin level
StepPosition marginTotal margin usedFree marginMargin level
0.50 lots EUR/USD50,000 ÷ 30 = 1,666.67 EUR1,666.67 EUR3,333.33 EUR300.0%
+ 0.30 lots GBP/USD(30,000 ÷ 30) × 1.170507 = 1,170.51 EUR2,837.17 EUR2,162.83 EUR176.2%
+ 0.30 lots EUR/USD30,000 ÷ 30 = 1,000.00 EUR3,837.17 EUR1,162.83 EUR130.3%

Control: margin used plus free margin must equal capital, that is 3,837.17 plus 1,162.83 equals 5,000.00 EUR. If your broker's calculator shows a different total, check the contract size, the margin rate and the conversion rate of the GBP position in it.

The model for which we compute the thresholds — without these fields none of the numbers below means anything. Three trades on two symbols: two EUR/USD legs of 0.80 lots in total and one 0.30-lot GBP/USD position, all long and moving in parallel by the same number of pips. The result arises in dollars: 8.00 plus 3.00, together 11.00 USD per pip — and it requires conversion into the account currency, either at a fixed rate of 1.0850 or at a rate moving together with price. The table gives both variants; neither is "the practical one". The basis for the threshold is not the current "used margin" field in the terminal but the total initial required margin used as the reference for forced close-out, here 3,837.17 EUR: "The value of the margin posted, for the purposes of protection through position close-out, should be established at the moment the position is opened and remain unchanged until the position is closed"[2] (p. 39).

Adverse move needed to reach the threshold, all three positions long
Way of converting the result100% level50% level
fixed rate of 1.0850114.70 pips303.94 pips
rate moving with price113.50 pips295.66 pips
divergence of the two variants1.20 pips, that is 1.05%8.28 pips, that is 2.72%

The backward check on each of these numbers closes to 100.0% and 50.0% to within a tenth of a point. The divergence of the variants grows with the distance of the move: 1.05% at the 100% threshold and 2.72% at the 50% threshold, that is 2.6 times more. Direction is an input too: for the same three positions reversed to short, the 50% threshold at a moving rate is reached only after 312.70 pips. The conclusion for tool acceptance is the same in every variant: a 50-pip move against the positions gives a margin level of 117.1% or 117.0% and triggers nothing. If a calculator shows a different number here, the difference sits in one of the listed assumptions, not in the arithmetic.

Which threshold you are testing in the calculator. The 50% level is a requirement imposed on the provider — "standardisation of the level at which providers of contracts for difference are obliged to close out a client's open contract for difference position (at the level of 50% of the initial required margin)"[2] (p. 38) — while the 100% level is a broker parameter. The same decision permits two modes of calculation, because close-out protection "does not prevent a provider of contracts for difference from carrying out an automatic close-out on a per-position basis rather than on a whole-account basis, once the level of 50% of the initial margin for a given position is reached"[2] (p. 39) — and that is the entire difference you have to be able to detect in a tool. The whole-account basis gives, at a fixed rate, a first close-out after 303.94 pips; the per-position basis gives one after 180.83 pips on the EUR/USD legs, that is 40.50% earlier, and after 211.67 pips on GBP/USD. A calculator that knows one mode will give one of these numbers; which applies to your account, and at what margin call level, you read in the contractual document. The mechanism of forced close-out is set out in the article on execution models, the mechanics of margin and leverage in the article on leverage and margin, and the scope of the rules themselves in the section on regulation.

Check four: swap

Assumptions: an illustrative rate of −2.40 USD per day of holding a 0.30-lot position, with one triple charge per week. Unit: account currency per day. The reference quantity is the number of daily rates charged — not the number of calendar days and not the number of accounting events. Formula: cost equals the number of charged rates multiplied by the daily rate, where a triple charge counts as three rates. The number of rates coincides with the number of nights whenever the period is a multiple of a full week, and outside that case only for some opening days; below I show what it depends on.

The special case: two full weeks. Fourteen consecutive calendar nights contain exactly two Mondays, two Tuesdays and so on for every day of the week — and therefore, regardless of the day the position is opened, exactly ten nights with a charge and exactly two Wednesdays. That gives eight single charges and two triple ones: 8 × (−2.40) + 2 × 3 × (−2.40) = −33.60 USD, which is the same as 14 × (−2.40). The agreement is not a coincidence: in an ideal week without holidays a triple charge does not add cost, it merely consolidates nights that are financed anyway, so a full week is always seven rates over seven nights. For a multiple of a full week this equality holds regardless of the opening day; at other lengths it happens only for some starting days, so it must not be assumed in advance.

At every other length, the full weeks are joined by a tail of a few days, and that tail may contain a Wednesday, a weekend, both or neither — and then the result depends on the opening day. I computed the table below for all seven starting days under one explicitly stated model: a charge occurs on the night after Monday, Tuesday, Wednesday, Thursday and Friday, Wednesday is a triple charge, and Saturdays and Sundays carry no charge. The control on the model is a full week: 1 + 1 + 3 + 1 + 1 = 7 rates over seven nights.

The same −2.40 USD daily rate and the same period, seven different opening days
Opening day12 nights: rates and cost14 nights: rates and cost16 nights: rates and cost
Monday14 → −33.60 USD14 → −33.60 USD16 → −38.40 USD
Tuesday13 → −31.20 USD14 → −33.60 USD18 → −43.20 USD
Wednesday12 → −28.80 USD14 → −33.60 USD18 → −43.20 USD
Thursday10 → −24.00 USD14 → −33.60 USD16 → −38.40 USD
Friday10 → −24.00 USD14 → −33.60 USD15 → −36.00 USD
Saturday (theoretical variant)12 → −28.80 USD14 → −33.60 USD14 → −33.60 USD
Sunday (theoretical variant)13 → −31.20 USD14 → −33.60 USD15 → −36.00 USD

The middle column is constant, the other two are not. Twelve nights cost, in this model, from −33.60 to −24.00 USD, and sixteen nights from −43.20 to −33.60 USD — and if you restrict yourself to opening on a day when the market is actually running, that is Monday to Friday, the range for sixteen nights narrows to −43.20…−36.00 USD, that is from 18 to 15 rates. A single number for these periods does not exist. The commonly quoted −28.80 USD for twelve nights is true only for an opening on Wednesday — the only trading day that yields twelve rates — and −38.40 USD for sixteen nights only for an opening on Monday or Thursday.

An exact cost after the event requires opening and closing dates and times, the broker’s rollover calendar, the multiple-charge day and a settlement calendar accounting for holidays. The rates actually applied and currency conversion rules are also needed. Before entry, future rates may be unknown, so the calculator estimates a cost under stated assumptions rather than guaranteeing it. Holidays can move or alter multiple charges; a calendar of non-trading days alone cannot reconstruct the bookings.

Rates and rollover rules depend on the broker, instrument and position direction, and can change. Settlement holidays can alter the charge day or multiplier even while the market remains open. For a future period, use an explicit rate scenario and show a range of assumptions when rates are unknown. For a completed period, reconstruct the rates and conversion actually applied. The full-week model gives one value only with a constant rate, five charging days, one triple charge and no holiday changes; it does not guarantee the account’s cost.

The same swap, two different percentages. Take a profit of 60 pips on a 0.30-lot position and the same swap cost of −33.60 USD. On EUR/USD the profit is 60 × 0.30 × 10.00 = 180.00 USD, so the swap eats 18.67% of the gross result. On a yen pair at USD/JPY 150.00 the same profit in pips is 60 × 0.30 × 6.6667 = 120.00 USD, so the same swap eats 28.00%. The net-result control closes both calculations: 180.00 − 33.60 = 146.40, that is 81.33% of gross, and 120.00 − 33.60 = 86.40, that is 72.00% of gross. The sentence "the swap ate a fifth of the profit" carries no information at all without the pair and the contract size.

Addendum: the reward-to-risk calculator and its silent assumption

For a model with two fixed outcomes — a gain of R risk units or a loss of one risk unit — the break-even win rate before costs is 1/(1+R). R is the gain-to-loss ratio in the same currency; the TP-to-SL distance ratio equals it only under matching pip-value and execution assumptions. The formula does not automatically describe variable outcomes, partial exits or gaps. The next calculation additionally assumes the same cost per trade, whether it wins or loses.

Break-even share of winning trades = 11 + R

One worked check is enough to establish whether the tool has a field for cost at all. The three cost values below are illustrative — they do not come from a measurement on any account, they have no assigned broker, instrument, period or number of trials, and they should not be taken as typical: a spread of 0.20 pips, a commission of 3.50 USD per lot charged per round turn, that is for opening and closing combined, and slippage of 0.15 pips. At a pip value of 10.00 USD the commission corresponds to 0.35 pips, so the cost is 0.20 + 0.35 + 0.15 = 0.70 pips. Control by a second route: 0.20 × 10 + 3.50 + 0.15 × 10 = 7.00 USD, and 0.70 pips × 10.00 USD is likewise 7.00 USD. With the cost expressed in units of risk added, the formula becomes (1 + cost) ÷ (1 + R): at a protective order distance of 40 pips the cost is 0.70 ÷ 40 = 0.0175 risk units, so for R equal to 2 the threshold rises from 1 ÷ 3 = 33.33% to 1.0175 ÷ 3 = 33.92%, that is by 0.58 percentage points. If you substitute your own measured costs, the calculation stays the same — only the first number changes. The site's reward-to-risk calculator is a good place to run this substitution.

The conclusion concerns the tool, not the strategy: a reward-to-risk calculator answers the question about the break-even threshold before costs, and the absence of a cost field means the tool does not account for it. What follows from that threshold for the choice of strategy, for expected value and for the risk budget belongs to the material on algorithmic trading and the article on position sizing.

8. News aggregators and squawk — how to measure their latency

A news aggregator is a service that collects announcements from many sources and delivers them in a single stream. A squawk is the same idea in audio form: an editor reads headlines aloud as they appear. Both categories do the same thing — they shorten the path between an announcement and your awareness — and both are an intermediate layer, exactly like a calendar.

Separate the creation of information from its arrival at the user. An intermediary first has to receive a report, but its channel can deliver it sooner than manually refreshing the original source’s website. A headline is a summary and can omit qualifications in the full release. An audio channel allows reception without watching a screen, but still requires attention; its delay and intelligibility need checking under specific conditions. The promise of “realtime” does not itself measure any of these properties.

The reference layer — the names of services, their prices and their advertised latencies — changes often enough that there is no sense in fixing it here. I have not carried out a market review of prices and funding models, so I describe no breakdown of this category here. Instead I give the questions you ask a specific provider before comparing it with another: is the price announced in a public price list or negotiated individually in conversation; what does the service live on if you are not paying for it directly; what does the plan you are looking at include, and what is not in it. You collect the answers from the provider on the day of purchase — any number entered here would go out of date faster than the rest of the text.

Something else matters far more. Marketing in this category trades on speed, so check one thing: whether your provider makes a measurable commitment on latency and in which document it is written down. If you cannot find such a clause, you have nothing to enforce — and this is a question about your provider's contract, not a thesis about the whole category. A number without a measurement method is not a result but an advertising slogan; that applies equally to the seconds circulating in online comparisons. Below is a procedure in which the active work and sampling period depend on the method and availability of comparable releases — and which gives an answer valid exactly where you need it: on your connection, your device and your subscription plan.

Procedure for measuring the latency of an information channel
  1. Choose a reference. Distinguish a scheduled release time from the first confirmed availability of the primary release. If you only have the schedule, label the result as delay from the schedule, rather than measured infrastructure latency.
  2. Synchronize the clock. Record synchronization, measurement resolution, refresh delay and manual response time. The result is no more precise than the measurement.
  3. Record paired times. For the same release, record the reference time and its arrival in the channel being examined; record delivery failures instead of dropping those trials.
  4. Collect a series of observations. Ten trials can start a check, but are neither a universal reliability minimum nor a basis for stable estimates of rare delays. Report the sample size, median and range; a mean can be added if the effect of extremes is explained.
  5. Record the environment. Date, connection, device, application or browser version, plan and region help compare repetitions; distinguish message types.
  6. State the limitation. You measure arrival at your workstation under these conditions, including observation limitations. Without logs, you cannot separate provider, network and application delays.

The result of such a measurement is not an assessment of the provider on a general scale and should not be presented as one. It is an answer to the question of whether the difference between two channels, which you either pay for or do not, is perceptible under your conditions.

Measuring news delivery. A scheduled release time need not be its actual availability time.
Illustrative data, not a provider ranking. Record delivery failures separately rather than dropping trials.

Primary sources are available free of charge and without an intermediary. The Federal Reserve, the European Central Bank, the Bank of England, the Bank of Japan and, for a European reader, the national central bank and the national statistical office publish announcements, minutes and speeches on their own sites. The price of that channel is zero, and so is its latency relative to itself. The cost is the format: an official document, with no summary and no interpretation. So before you pay an intermediary, settle which of two things you are buying from them — faster arrival of information, or a ready-made summary. The first can be measured with the procedure above; the second you judge by reading a few headlines alongside the source announcement. You make that determination for your own provider, because the two things are often sold together and at a single price.

Derivative tools require an explicit methodology. A separate category is services converting derivative prices into the implied probabilities of central bank decisions. That is neither a forecast nor an editorial opinion, but a transformation of quotations according to a published method — and the value of such a tool depends directly on whether the method is described. Before you use such a number, check which contracts enter the calculation and what assumptions were adopted in the conversion. How the market prices rate decisions is explained in the article on central bank monetary policy.

One warning to close this section, because the category is particularly exposed to it. Messenger groups advertising themselves as "news channels" are sometimes sales channels in reality, with signal-type material appearing alongside reprinted headlines. An information service describes what has happened. If a channel describes what you are to do with your account, it is no longer a news aggregator, whatever it calls itself. Why a stream of announcements can in itself degrade the quality of decisions is described in the article on information noise.

Market behaviour at moments of sudden liquidity evaporation — when the headline appears after the move rather than before it — belongs to the material on low liquidity risk. Here a tool-level statement suffices: an information channel is an intermediate layer, so by definition it does not run ahead of the source it draws from. Whether your provider makes any commitment at all about anticipating the market is something you check in its contract, not in its marketing material.

9. Alerts as part of the set — what belongs to the platform and what to the phone

A calendar, a scanner and a calculator share one property: require checking in their basic view; some also offer configurable notifications. An alert reverses that relationship — here the tool comes to you. Two distinctions are worth knowing, but their place is elsewhere: an alert evaluated on the provider's server works regardless of whether your computer is on, while an alert evaluated locally in the terminal requires a running application; and — whichever the model — an alert is not an order, only a notification that a condition has been met. Full alert configuration on a charting platform is covered in the article on TradingView, and notification behaviour on a phone, power saving mode and the procedure for a lost connection in the article on mobile trading. One thing remains here.

A macro-release reminder. Separate the event time, display zone and reminder schedule, such as “15 minutes before”. An incorrect zone can alter the displayed time alone or also the reminder’s arrival, depending on the implementation. Compare its arrival with the correctly converted primary-source time. Repeat checks after a zone change, an application update and during the DST mismatch windows described in section three. One successful test does not reveal the service’s architecture or establish correctness on other dates.

10. An acceptance procedure for a tool before first use

Before first use, check a tool’s data provenance, units, default parameters and whether its result can be reconstructed. The procedure below organizes those questions; the time required depends on the tool and documentation available. It is not a one-off guarantee of correctness. Repeat relevant checks after changes to instrument specifications, settings, data sources or application versions.

Ten questions before trusting a tool
  1. Where does the number come from? The name of the institution, agency or data provider. If the tool does not state it, every further verification requires a manual search.
  2. In what unit is it given? Per cent, point, pip, lot, account currency. The unit is part of the number, not its decoration.
  3. What convention stands behind the value? Month on month or year on year, seasonal adjustment, annualisation, headline or core value.
  4. In which time zone? And whether the zone setting survives logging out.
  5. When was it last recalculated? A timestamp next to the result or information about the refresh frequency.
  6. Can the historical value change? For macro data the answer is "yes" and it also applies to series from years ago.
  7. Which set of instruments or events does it cover? And whether it overlaps with the set available at your broker.
  8. What parameters does it assume by default? Contract size, leverage, account currency, conversion rate — four fields in which the difference between two calculators can sit.
  9. Can you reproduce the result by hand? One case, an explicit formula, a backward check. If you cannot, you do not know what the tool is computing.
  10. What happens when the tool stops working? Write down an alternative and test it once, before it is needed.

Point nine decides the value of the other eight. A calculator whose result you cannot reproduce on paper for a single case is not a tool but an oracle. The difference reveals itself exactly when the tool gives an unexpected number — because then you can either check whether it erred or you did, or you are guessing.

It is also worth naming outright what these three categories of tool do not do. A calendar does not say which way the rate will go. A scanner does not choose trades, it narrows the set. A calculator does not judge whether an entry is worth taking — it computes a size given risk parameters you established elsewhere. None of them creates an edge on its own; they improve repeatability and reduce the number of arithmetic mistakes, and that is something entirely different from a positive expected value.

The set described in this article works at a desk and on a full screen. On a phone some of the assumptions stop holding: the calculator form looks different, notifications behave differently, an application sent to the background works differently. The article on mobile trading deals with exactly that — what can realistically be done from a smartphone and what should not be assumed about it.

FAQ — Calendars, scanners, calculators and news aggregators

Why do two calendars show a different forecast for the same release?
Because a forecast is not a market fact but a survey statistic. Providers poll different panels of economists, close collection of responses on different days, and some report the median while others report the mean. Both numbers can be correct at the same time. The practical conclusion: if you are building your own record of deviations of actual from forecast, use a single forecast provider throughout — otherwise you are comparing two different measures and the distortion enters your data, not the market.
A broker's calculator and an external calculator give a different position size. Which is right?
First separate two questions that calculators are apt to conflate. Position size follows solely from the risk budget: capital times risk percentage, divided by the protective order distance and the pip value — leverage and the margin rate do not appear in that formula. Required margin is a separate, later check: whether free margin allows such a position to be opened. So if two tools give a different volume, go through every input the formula takes: capital, risk percentage, protective order distance, symbol including suffix, contract size, account currency and conversion rate — and on top of that check whether either tool counts transaction cost into the risk budget and to which volume step it rounds the result. If only the required margin differs, the margin rate assigned to your client status is in play, and the volume stays the same. If the results still differ after aligning those fields, compute one case by hand from the formula and check it backwards — position size multiplied by the protective order distance and by the pip value must give your risk budget. For a risk limit, round the permissible volume down to the broker’s step and check the minimum and maximum. A result below the minimum does not justify rounding up beyond the budget; reject that version or change the plan. A maximum-volume cap can create a difference larger than one step, so difference size alone does not identify an error. Leverage does not change the risk formula, but available margin can restrict the volume that can be opened.
Does "high impact" in a calendar mean there will be a big move?
No. It is an editorial label from the calendar provider, at some services attached to the indicator permanently, regardless of the context in which the release appears. Two services can assign different weights to the same event and there is no objective scale that would settle the matter. The label can be tested on your own data — the worked check is in the fourth section of this article — but the result will describe the past at one data provider, not a property of the indicator.
How do I check whether a scanner is showing instruments unavailable at my broker?
Export or copy out the result list and compare it with the instrument list in your platform, paying attention to symbol suffixes. Scanners built into charting platforms operate on the data provider's symbol set, which need not overlap with your broker's offering — and the differences go both ways. Do this once when configuring the tool and record the result as your own set of scanned instruments. This is one of the four control questions for a scanner described in the fifth section.
Is it worth paying for a faster news service?
Compare cost with benefit after checking the channels on your connection and device. Record the reference, paired times, measurement errors, sample size and environment as described in section eight. Ten observations can begin that check, but do not establish a constant delay or performance in every condition. Also check whether the provider offers a measurable speed commitment. Public releases from statistical institutions and central banks are available at source; a paid service can provide convenient summaries, selection or faster access in your workflow. These are separate benefits that need separate assessment.
What is the difference between the "previous" value and "revised"?
“Previous” is the reading for the preceding period; depending on the service, it can be the original or an updated value. “Revised” explicitly marks a correction to an earlier reading. A missing revision field does not establish whether a revision occurred — check the primary release and data vintage. Institutions can revise readings and periodically recalculate series under indicator-specific rules. History downloaded today can therefore differ from an earlier vintage, and the pre-release forecast and view should be archived with a capture time and provider name.

Sources and bibliography

  1. Office of the Polish Financial Supervision Authority, "Client results on the Forex market for 2025", announcement published 27 March 2026. The share of clients who incurred a loss, among active clients, was 72.2% against 70.6% a year earlier, with the footnote under the table stating that these values "are calculated without taking into account the group of clients that showed neither a profit nor a loss"; clients' realised losses came to PLN 2,680,482,791 against PLN 740,028,152 of realised profit, which — as the announcement puts it — is "almost 4 times the combined value of clients' profits". Retail clients accounted for 99.9% of the number of active clients transacting through domestic brokerage houses and brokerage offices. knf.gov.pl [AS AT: 2026-08-22]
  2. Polish Financial Supervision Authority, decision no. DAS.456.2.2019 of 1 August 2019 establishing restrictions on the marketing, distribution and sale of contracts for difference to retail clients, making permanent ESMA's 2018 product intervention measures. A leverage limit of 30 to 1 for major currency pairs, separate limits for other instrument classes and separate rules for clients with experienced status. Text of the decision: dziennikurzedowy.knf.gov.pl (Official Journal of the KNF 2019, item 27, 47 pp.). The quotations cited in the text come from pages 37, 38 and 39 of that document. [accessed: 2026-08-22]
  3. Macroeconomic data release schedules at primary sources. Each of the addresses below was opened and checked directly on the institution's page on the access date given. Bureau of Labor Statistics, "Schedule of Releases for the Employment Situation" — the release schedule for the non-farm payrolls report; every row of the schedule gives the hour "08:30 AM": bls.gov [accessed: 2026-08-22]. Eurostat, "Release calendar" — a publication calendar kept, as the page itself states, in the "Europe/Luxembourg time zone (CET/CEST)": ec.europa.eu [accessed: 2026-08-22]. Statistics Poland, "Release calendar" — a publication schedule ordered by month, covering the regional statistical offices as well: stat.gov.pl [accessed: 2026-08-22]. Board of Governors of the Federal Reserve System, "Federal Open Market Committee" — the FOMC meeting calendar: federalreserve.gov [accessed: 2026-08-22]. European Central Bank, "Schedules for the meetings of the Governing Council and General Council of the ECB and related press conferences": ecb.europa.eu [accessed: 2026-08-22]. Narodowy Bank Polski, the "Monetary Policy Council meeting" calendar distinguishing decision-making from non-decision-making meetings: nbp.pl [accessed: 2026-08-22].
  4. Ederington L.H., Lee J.H., "How Markets Process Information: News Releases and Volatility", The Journal of Finance, 48(4), 1993, pp. 1161-1191. The classic study of the volatility response to economic data releases: windows of elevated volatility and the pace at which it normalises.
  5. Andersen T.G., Bollerslev T., Diebold F.X., Vega C., "Real-Time Price Discovery in Global Stock, Bond and Foreign Exchange Markets", Journal of International Economics, 73(2), 2007, pp. 251-277. The impact of macroeconomic data releases on the price formation process, including on the currency market.
  6. Chaboud A.P., Chernenko S.V., Howorka E., Iyer R.S., Liu D., Wright J.H., "The High-Frequency Effects of U.S. Macroeconomic Data Releases on Prices and Trading Activity in the Global Interdealer Foreign Exchange Market", Board of Governors of the Federal Reserve System, International Finance Discussion Papers, 823, 2004. A study on EBS transaction data from the interdealer spot market in EUR/USD and USD/JPY. The authors report increased activity even after releases matching expectations. federalreserve.gov [checked: 2026-10-08]

Your learning progress

Finished this lesson? Mark it yourself. Opening or scrolling a page never marks it complete.

Progress requires JavaScript and browser storage.

Saved only in this browser. No account or cross-device sync. Clearing site data can erase your progress.

Jarosław Wasiński

Editor-in-chief, ForexMechanics.com • Macro & FX analyst

Independent analyst and practitioner with over 20 years of experience in financial markets. Actively involved in Forex since 2007, with a focus on fundamental analysis, OTC market structure, and disciplined capital risk management.

  • Creator and editor-in-chief of MyBank.pl, providing reliable financial education since 2004.
  • Author of hundreds of in-depth market commentaries, structural analyses, and educational materials for traders and investors.
  • Advocate for market transparency, promoting education grounded in hard data and institutional reports (BIS, ESMA, CFTC).

All content is educational and informational only – it does not constitute investment advice or a recommendation. Remember: leveraged trading (Forex/CFD) carries a high risk of rapid capital loss.