What is NEXTMarketReaction?
NEXTMarketReaction studies one thing: how EUR/USD has historically reacted in the first minutes after major U.S. economic releases — reports such as Non-Farm Payrolls (NFP), CPI inflation or GDP.
When a new release comes out, NEXTMarketReaction compares the result with what economists expected, and with how EUR/USD behaved after similar releases in the past. From that historical evidence, a model estimates how likely EUR/USD is to be above or below its price at the moment of the release (the reference level), at 1, 3, 5, 8 and 15 minutes after the release.
The problem it helps with
Right after a big release, prices move fast and headlines are noisy. In the moment, it is hard to judge whether a result is really a surprise, and how EUR/USD has usually responded to a surprise like it.
NEXTMarketReaction turns years of past releases into one clear, consistent reading: a probability for each time horizon, plus a separate measure of how much history supports it.
The basic idea
Every forecast follows the same path, from the economic calendar to the numbers on the release page.
Economic event scheduled
A U.S. release, such as CPI, is on the calendar with a date and a time.
Before the release
The Forecast and the Previous value are known. The Actual is not.
Release time (T0)
The Actual value is published.
Inside NEXTMarketReaction
NEXTMarketReaction reads the release data
The new Actual is detected and compared with the Forecast.
Historical model
The surprise is compared with how EUR/USD reacted after similar past releases of the same indicator.
Probability model
That historical evidence becomes calibrated probabilities.
EUR/USD reaction forecast
The result for each time horizon, shown on the release page.
Horizons: 1, 3, 5, 8 and 15 minutes after the release
UP / DOWN probability + Reliability for every horizon
Actual, Forecast and Previous
Every release on an economic calendar comes with three values. Two are known in advance; one is only known at the release.
- PreviousKnown before the release
- The value reported in the previous release, for the prior month or quarter. It is sometimes revised.
- ForecastKnown before the release
- The consensus expectation of economists, published before the release. Markets have largely prepared for it in advance.
- ActualPublished at the release
- The newly published value for the current release. Nobody knows it until it is published.
Why the Actual alone is not enough
A number such as 210K new jobs means little on its own. What matters is how it compares with what was expected. Comparing the Actual with the Forecast tells you whether the new information is different from what the market was prepared for; the Previous value adds context.
Surprise, in simple terms
Surprise = Actual − Forecast. A result above the Forecast is a positive surprise, a result below it is a negative surprise, and a result equal to the Forecast is no surprise at all.
Non-Farm Payrolls
Example (illustrative)
- Forecast
- 150K
- Actual
- 210K
- Surprise
- +60K
The release came in above expectations: a surprise of +60K.
“Above forecast” does NOT automatically mean EUR/USD must move in one specific direction. Whether a higher number is good or bad news for the U.S. dollar depends on the indicator, and markets do not always react the way history suggests. The model evaluates each surprise in the context of how EUR/USD has historically reacted to that indicator.
The model's estimate is built from the surprise — the Actual compared with the Forecast — for that specific indicator. The Previous value is shown for context, but it does not change the model forecast.
A note on words: on NEXTMarketReaction, “Forecast” always means the economists' consensus published before the release. What the model produces is called the model forecast: the UP / DOWN probabilities.
Before and after the release
Before the Actual is published, only a preliminary estimate is possible — built from pre-release information and clearly labelled. The release analysis can only exist after the Actual value is published: NEXTMarketReaction never knows the Actual in advance.
Before the release
- Event scheduled
- Forecast known
- Previous known
- Actual not available yet
- Preliminary estimate only (pre-release information)
T0 — the release
- The Actual value is published
After the Actual is detected
- NEXTMarketReaction processes the release
- The release analysis replaces the preliminary estimate
- The page updates by itself — no need to refresh
What the release page shows at each stage
- Upcoming
- The release has not happened yet. You can see the event, its scheduled time, the Forecast and the Previous value.
- Release published
- The scheduled time has passed and we are waiting for the Actual value to reach our data source.
- Preparing forecast
- The Actual is in and the forecast is being prepared.
- Prediction ready
- The forecast is available.
- Taking longer than expected
- The Actual is taking longer than usual to arrive. We keep checking automatically.
How soon the forecast appears depends on when our economic-data provider receives the Actual value. That is often shortly after publication, but no timing is guaranteed. Once a forecast is stored, it is never changed afterwards, even if the data is revised later.
Two stages: preliminary and release analysis
NEXTMarketReaction works in two separate stages. They use different information and different models, and they are never mixed.
Stage 1
Preliminary analysis — before the Actual
Uses only information available before the release: the indicator, the Forecast and the Previous value. It never uses the Actual. It is labelled Preliminary, shown with its real probability — often close to 50/50 — and may change materially once the result is published.
Stage 2
Release analysis — after the Actual
Once the Actual is published, the calibrated release model reads the surprise (Actual vs Forecast) and gives UP / DOWN probabilities with Reliability for 1, 3, 5, 8 and 15 minutes. It replaces the preliminary estimate.
In historical testing, preliminary estimates were close to 50/50; the release analysis is the main forecast. The two are shown separately so they are never confused.
How the forecast is generated
The model forecast is not a rule of thumb such as “good news for the dollar means EUR/USD falls”. It is built from what EUR/USD actually did after comparable releases in the past. When the Actual value arrives, the same steps run every time.
Compare the Actual with the Forecast
The difference is the surprise. Its direction (above or below the Forecast) and its size are measured. The size is judged against the typical surprises of that indicator, so a CPI surprise and a payrolls surprise are measured on comparable scales, and classed as small, medium or large.
Find comparable history
The model looks up past releases of the same indicator with a surprise of similar size. Because a higher number means different things for different indicators (more jobs is not the same as more jobless claims), each surprise is always read against its own indicator's history.
Check what EUR/USD actually did
For each horizon, the model uses how often EUR/USD ended above or below its release-time reference level after those comparable releases, taking into account which way their surprise went. The numbers come from these observed outcomes, not from a fixed rule: if EUR/USD more often moved the other way, the probabilities show that.
Turn it into calibrated probabilities
Those historical frequencies were calibrated in advance, so that a stated probability matches, as closely as possible, how often that outcome occurred in comparable history. The result is an UP / DOWN probability for each horizon.
Attach a Reliability level
Reliability shows how much comparable history stands behind each probability, and how specific it is. When there is little history for this exact indicator and surprise size, the model draws on a broader group of past releases and shows a lower Reliability.
If the Actual equals the Forecast, there is no surprise to read. The model forecast then shows an even 50% / 50% split: no clear directional bias.
The model is fixed. It was built from historical data in advance; it does not learn from today's market or change during the day, and the same release data always produces the same model forecast. It uses only the release data — the indicator, the Actual and the Forecast. It does not use live prices, news, sentiment or anything that happens after the release.
The reference level
Every forecast needs a reference point. NEXTMarketReaction uses the EUR/USD price at the moment of the release: the reference level (in technical terms, P0).
Each horizon asks one simple question
Will EUR/USD be above or below the reference level this many minutes after the release?
Release (T0)
- 1 minuteAbove or below?
- 5 minutesAbove or below?
- 8 minutesAbove or below?
- 15 minutesAbove or below?
What happens between the release and that moment is not part of the question — only where EUR/USD is at that minute, compared with the reference level.
How to read a forecast
Here is what a forecast looks like, reduced to its essentials: one UP / DOWN pair of probabilities per horizon.
EXAMPLE — FOR EDUCATIONAL PURPOSES
EUR/USD1 min
↑ UP 72%↓ DOWN 28%
3 min
↑ UP 70%↓ DOWN 30%
5 min
↑ UP 68%↓ DOWN 32%
8 min
↑ UP 61%↓ DOWN 39%
15 min
↑ UP 56%↓ DOWN 44%
Illustrative values only. This is not a real forecast.
What “UP 72% at 1 minute” means
According to the model, EUR/USD has an estimated 72% probability of being above its release-time reference level at the 1-minute horizon — and therefore a 28% probability of being below it.
What it does NOT mean
- that the price rises continuously for a whole minute
- that the price cannot move down first
- a 72% profit
- a 72% price increase
- a guaranteed outcome
Read the horizons together, not one percentage in isolation. In this example the directional evidence is strongest at 1 minute and fades by 15 minutes.
Directional strength
Next to the main probability, the release page shows a short label describing how far the probability leans away from an even 50/50 split.
- 80%+
- Very strong directional bias
- 70–79%
- Strong directional bias
- 60–69%
- Moderate directional bias
- 55–59%
- Slight directional bias
- Below 55%
- No clear directional bias
This label describes the model's directional bias — not the certainty of the outcome. Even a very strong bias can be wrong on any given release. Reliability is shown separately.
Probability and Reliability are different
Every horizon carries two separate pieces of information. They answer different questions and are never merged into one score.
Probability
↑ UP 74%
The model's estimated probability of each direction at that horizon. Answers: which way, and how strongly?
Reliability
Reliability: Medium
How much comparable historical evidence stands behind that estimate. Answers: how well supported is it?
The three Reliability levels
- High
- Plenty of specific history: at least 50 comparable past releases of the same indicator at that horizon, and that history supports the direction shown.
- Medium
- Less, or less specific, history — for example, fewer comparable releases, or evidence drawn from a broader group of releases. The displayed probability is kept between 25% and 75%.
- Low
- Little comparable history, or no surprise to read. The displayed probability is kept between 40% and 60%.
Why this matters
The same probability means more when it rests on more evidence. A 70% with High reliability is backed by a large body of comparable history; a 70% with Medium reliability rests on thinner or broader evidence and deserves more caution. Low reliability also limits how far from 50% the probability can go, so a Low-reliability forecast never shows a strong bias.
Time horizons: 1, 3, 5, 8 and 15 minutes
Markets can behave differently in the first minute than a quarter of an hour later, so each horizon is estimated separately. The underlying research covered every minute from 1 to 15; the product shows five of them — 1, 3, 5, 8 and 15 minutes — for a clear picture without clutter.
- 1 min
- the immediate reaction
- 5 min
- the short reaction
- 8 min
- the extended short-term reaction
- 15 min
- the later short-term state
These are rough descriptions, not fixed market behavior.
Probabilities can differ between horizons (illustrative)
- 1m
↑ UP 76%Strong directional bias
- 5m
↑ UP 69%Moderate directional bias
- 15m
↑ UP 55%Slight directional bias
This means the historical directional evidence is stronger for the immediate reaction than for the later horizon.
It does NOT mean the system is predicting a continuous path. Each horizon is a separate question about a single moment: above or below the reference level at that minute.
A complete example, step by step
Fictional example — every value is illustrative
Before 13:30 UTC
CPI is scheduled
The release page shows the event and its scheduled time, with Previous 2.8% and Forecast 2.7%. Actual: waiting. Status: Upcoming. There is no model forecast yet.
- Previous
- 2.8%
- Forecast
- 2.7%
- Actual
- –
13:30 UTC — T0
CPI is published
The Actual comes in at 2.5%, below the 2.7% Forecast: a negative surprise of 0.2 points. Inflation was lower than expected.
- Previous
- 2.8%
- Forecast
- 2.7%
- Actual
- 2.5%
NEXTMarketReaction detects the update
The status changes to Preparing forecast, then to Prediction ready. The page updates by itself.
The model looks at comparable history
It finds past CPI releases with a surprise of similar size and checks, for each horizon, how often EUR/USD ended above or below its release-time level afterwards, taking into account which way the surprise went. In this fictional example, that history favors UP.
The forecast appears
Each horizon shows its probabilities and its Reliability:
- 1m
↑ UP 70% · DOWN 30%
Reliability: High - 3m
↑ UP 68% · DOWN 32%
Reliability: High - 5m
↑ UP 66% · DOWN 34%
Reliability: High - 8m
↑ UP 61% · DOWN 39%
Reliability: Medium - 15m
↑ UP 57% · DOWN 43%
Reliability: Medium
- 1m
How to read it
At 1 minute, the model estimates a 70% probability that EUR/USD is above the reference level, with High reliability: a strong directional bias. By 15 minutes the estimate is 57% with Medium reliability: only a slight bias. The evidence is clearest for the immediate reaction.
This example is fictional. It does not describe a real release, a real forecast or a real market reaction.
Your map of the release page
When you open a release page, this is where to look. The numbers in this picture are illustrative.
- Indicator: the short name of the release.
- Currency: the release belongs to the U.S. dollar (USD) — every release NEXTMarketReaction covers is a U.S. release.
- Status: where the release is right now — Upcoming, Waiting for release data, Preparing forecast or Prediction ready.
- Event name: the economic release this page is about.
- Scheduled time: when the release is due, in your time zone (you can change it under Account).
- Release data: Actual, Forecast, Previous and the Surprise. The Actual shows “–” until it is published.
- Main forecast: the favored direction and its probability at the 1-minute horizon.
- Directional strength: how clearly the probability leans one way.
- P(UP) and P(DOWN): both probabilities for that horizon. They always add up to 100%.
- Reliability: how much historical evidence supports the estimate.
- Expiry curve: the probabilities at 1, 3, 5, 8 and 15 minutes.
- Historical context: whether the surprise was small, medium or large.
Impact: the calendar shows the data provider's own impact rating. The 11 U.S. releases with directional forecasts were chosen for their historical market relevance; the provider may still rate some of them Medium or Low (for example, Initial Jobless Claims).
Quick start: how to use NEXTMarketReaction
- Open the Dashboard.
- Find the upcoming economic release.
- Open its release page.
- Before publication, check the event information and the scheduled time.
- Wait for the Actual value to be published and processed — the page updates by itself.
- Read the directional probabilities for each available horizon.
- Check the Reliability.
- Compare horizons rather than reading one percentage in isolation.
NEXTMarketReaction is an information tool, not trading advice. How you use the information, and any risk you take, are your own decisions.
What NEXTMarketReaction is not
To avoid misunderstandings, NEXTMarketReaction is NOT:
- a guarantee of any outcome
- a prediction of the exact price path
- a prediction of the exact size of a price move
- a promise of profit
- a replacement for your own risk assessment
- advance knowledge of unpublished economic data — the Actual is only known once it is published
Markets can and do react differently from historical patterns: other news, positioning, liquidity and data revisions all play a part. A probability describes how often something happened in comparable past situations; it cannot tell you what will happen this time.
Glossary
- Actual
- The newly published value of a release.
- Forecast
- Economists' consensus expectation, published before the release.
- Previous
- The value from the prior release, sometimes revised.
- Surprise
- Actual minus Forecast: how far the result landed from expectations.
- Release
- The scheduled publication of an economic report.See:Non-Farm Payrolls (NFP)Consumer Price Index (CPI)Gross Domestic Product (GDP)
- T0
- The moment of the release.
- Reference level (P0)
- The EUR/USD price at the moment of the release. Every horizon is measured against it.
- UP probability
- The estimated probability that EUR/USD is above the reference level at a given horizon.
- DOWN probability
- The estimated probability that EUR/USD is below the reference level at a given horizon. UP and DOWN always add up to 100%.
- Horizon
- How many minutes after the release the question is asked: 1, 3, 5, 8 or 15.
- Preliminary analysis
- A pre-release estimate that uses only information available before the release — the indicator, Forecast and Previous. It never uses the Actual.
- Release analysis
- The calibrated directional model applied after the Actual is published.
- Directional forecast
- A release supported by NEXTMarketReaction's directional model: it receives UP / DOWN probabilities.
- Volatility event
- An event that may create event risk or volatility, but has no supported UP / DOWN directional model.
- Reliability
- How much comparable historical evidence supports an estimate: High, Medium or Low.
- Directional bias
- How far a probability leans away from an even 50/50 split.
- Impact rating
- The calendar provider's own rating (High / Medium / Low) of how much a release usually matters. It is not a volatility forecast, and it can differ from our choice of the 11 U.S. releases with directional forecasts.See:Non-Farm Payrolls (NFP)Consumer Price Index (CPI)Retail SalesISM Services PMI