Z-Score

If you've ever wondered whether today's move is genuinely unusual or just feels unusual, a z-score is one of the simplest ways to answer that objectively.

In 30 Seconds

• Measures unusualness — how far a value is from its recent average.
• Uses standard deviations — +2 is unusually high, −2 unusually low, 0 is average.
• Does not predict direction — it describes the current reading, not the next move.
• Useful for historical research — it defines "unusual" in a way that holds across calm and volatile eras.
• HIE discloses exactly how it's computed — the window and method behind every z-score.

Definition

A z-score answers a simple question: how far is this value from normal, and in what units? It's defined as the distance between a value and the average, divided by the standard deviation:

The result is expressed in standard deviations. A z-score of +1 means the value is one standard deviation above the average; −1.5 means one and a half standard deviations below. Because it's scaled by the standard deviation, a z-score is unit-free — it lets you compare how unusual a $500 move in one market is against a $12 move in another, or today's volume against a quiet market's volume, on the same scale.

Two ingredients define any z-score: the average it's measured against, and the standard deviation used to scale it. Both are computed over some reference set of past values — and which past values, and how they're calculated, is exactly what has to be disclosed for a z-score to mean anything. A z-score cannot be interpreted correctly unless the reference window and calculation method are known.

Why It Matters

Raw numbers are hard to interpret because they carry no context. Is a 40-point move in the NASDAQ large? It depends entirely on what's normal lately — 40 points is trivial in a volatile week and enormous in a calm one. A z-score solves this by expressing every reading relative to its own recent behaviour, so "unusual" is measured against the market's current normal rather than an arbitrary fixed threshold.

Traders often use z-scores to study unusually large price moves, unusually high trading volume, volatility spikes, and other statistically uncommon market conditions — each expressed on the same comparable scale. This matters for historical research in particular. If you want to study "what tends to happen after an unusually large up-move," you first need a consistent, honest definition of "unusually large" — one that adapts to changing volatility rather than using a fixed point value that means different things in different eras. Z-score provides that: "a move with a z-score above +2" is a definition that stays meaningful across calm and turbulent periods alike.

To ground it with no complex math: imagine a market whose daily moves have recently averaged near zero, with a standard deviation of 20 points. A +50-point day has a z-score of +2.5 — genuinely unusual, roughly two-and-a-half standard deviations above normal. Now imagine volatility rises so the standard deviation becomes 50 points. That same +50-point day now has a z-score of +1.0 — no longer unusual at all, because the market's normal has widened. The raw number didn't change; its unusualness did. That adaptivity is the whole point.

How HIE Computes It

Two platforms can report different z-scores for the same market on the same day, because they may use different windows or calculation methods. That's why HIE discloses its methodology rather than presenting a bare number.

Z-score is a live, askable input in HIE — you can pose questions conditioned on how unusual a reading was, and HIE computes the z-score using a disclosed, deterministic method. The exact method matters, because a z-score is only as honest as the window and calculation behind it.

HIE computes z-score with a single, disclosed, deterministic method, and every answer prints the exact window it used — so nothing rests on the reader knowing the defaults. The method has three properties worth stating plainly:

• Strictly-trailing window. The average and standard deviation are measured against a value's own trailing history only — the calculation at any bar uses nothing from after that bar. That is precisely what "no look-ahead" means: a reading is judged only against history that genuinely preceded it.
• Sample standard deviation. HIE uses the sample standard deviation (dividing by n−1), the standard choice when computing statistics from a sample of history rather than a full population.
• A disclosed default window per feature, which you can override. Each feature carries its own default window from HIE's registry — for example, 100 bars for price and RSI, 20 bars for volume and ATR — and you can always specify your own (for instance, "over the last 250 bars"). Because the window depends on the feature, there is no single universal number; instead, every disclosure states the actual window used.

In HIE's own disclosure ("what ran"), the computation appears in this exact form:

ZSCORE(<FEATURE>, window=<N>, method=sample SD (ddof=1), rolling, strictly-trailing, v1)

So a question about an unusually low price over 250 bars resolves to, and is disclosed as:

ZSCORE(PRICE, window=250, method=sample SD (ddof=1), rolling, strictly-trailing, v1)

The words you ask are shown as the exact computation they became — which is the point: the definition behind your question is never hidden. (The wordings "z-score," "sigma," and "standard deviation" all resolve to this same computation.)

Interpreting a Z-Score

A rough, conventional reading of z-score magnitudes (these are general statistical conventions, not MPM findings):

• 0 to ±1 — within the normal range; roughly two-thirds of readings fall here.
• ±1 to ±2 — moderately unusual; readings this far from average are less common.
• ±2 to ±3 — unusual; only a small fraction of readings reach this far.
• Beyond ±3 — rare; extreme readings that happen infrequently.

A crucial caution: these proportions hold exactly only if the underlying values follow a normal distribution — and market data often does not. Market returns have fatter tails than a normal distribution, meaning extreme z-scores (±3 and beyond) occur more often in real markets than the textbook percentages suggest. So a z-score is an honest relative measure of unusualness, but the specific "probability" attached to a given z-score should be treated with caution for market data.

What a Z-Score Is Not

It is not a prediction. A high z-score says a reading was unusual relative to its recent history — nothing about what happens next. "Unusually high" is a description of the present reading, not a forecast of the following move.

It is not a fixed, universal threshold. Because the average and standard deviation are computed over a trailing window, the same raw value can have very different z-scores at different times. A z-score is always relative to the window it was computed over — change the window and you change the score.

It is not meaningful without its window disclosed. A z-score quoted without stating what reference period and method produced it is uninterpretable. This is why HIE discloses the computation rather than presenting a bare number.

Common Mistakes

"It hit +2, so the market must reverse."

A z-score measures how unusual a reading is — not what happens next. A +2 reading is a description of the present, not a signal that a reversal (or continuation) is coming. Treating an unusual reading as an automatic reversal cue is the most common error.

"I changed the window from 20 to 200 and got a different z-score — something's wrong."

Nothing's wrong — that's expected. The z-score is measured against whatever trailing window is used. A longer window includes calmer and more varied history, so the same value can score differently. The window is part of the definition, not a bug.

"This source says today was a +3 move and another says +1.8 — one must be wrong."

Not necessarily. Two z-scores computed over different windows or methods are simply not comparable. A z-score is only meaningful alongside the window and method that produced it — comparing two computed differently is meaningless.

"A +3 reading is basically impossible because it's so rare."

Rare doesn't mean impossible. Financial markets have fat tails, so extreme readings (±3 and beyond) occur meaningfully more often than a textbook normal distribution suggests. Treating a high z-score as "can't happen" is how people get blindsided by moves that were always possible.

Limitations

A z-score is only as meaningful as the window and distribution behind it. Over a very short trailing window, the average and standard deviation are themselves noisy, so the z-score can swing on little evidence — the same small-sample caution that applies to every statistic applies to the inputs of a z-score too.

The deeper limitation is distributional. The clean "±2 = unusual, ±3 = rare" interpretation assumes a normal distribution, and market data reliably violates that assumption with fat tails and skew. A z-score remains a fair ordering of how unusual readings are relative to each other, but converting a z-score into a precise probability of occurrence is unreliable for market data. Treat z-score as an honest relative-unusualness scale, not as a probability calculator.

Try These HIE Questions

The best way to understand z-score is to test it against history. Because z-score is a live, askable input in HIE, this concept isn't just theory — it's something you can put a historical question to directly.

Each of these runs in HIE today and returns a full certified answer, disclosing the exact z-score computation it used:

• "NQ where RSI z-score is above 2" — resolves to ZSCORE(RSI, window=250, …) > 2.
• "GC price 2 sigma below average" — resolves to ZSCORE(PRICE, window=250, …) < -2.
• "NQ z-score below -2" — HIE asks which feature you mean; choosing ATR gives ZSCORE(ATR, window=20, …) < -2.
• "GC price 2 sigma below average" with the default window — resolves to ZSCORE(PRICE, window=100, …) < -2.

Notice two things: "sigma" and "z-score" are treated as the same request, and where a question is ambiguous (which feature? which window?) HIE asks rather than guessing — then discloses exactly what it ran.

How This Fits Into MPM

Z-score is one of the statistical building blocks behind HIE's ability to condition historical research on how unusual a market reading was — turning vague ideas like "a big move" into precise, disclosed, testable definitions.

You'll encounter it in:

• HIE (Historical Intelligence Engine), where z-score is a live, askable input: you can ask historical questions conditioned on how unusual a reading was, and every answer discloses exactly how the z-score was computed — the reference window and method — so the definition behind your question is never hidden.
• Research Papers, where "unusualness" is expressed in disclosed z-score or standard-deviation units rather than raw points, so a claim holds up across calm and volatile eras.
• Intelligence Circle — a member-only research environment containing advanced market research, historical investigations, and trading strategies developed using the MPM research framework.

Z-score is live and askable in HIE today, across every entry point — chat, the Reports page, the full report, matching charts, and the exported PDF and CSV — all running the same shared parser, so identical words produce an identical computation every time.

MPM's broader stance is that a measure of "unusual" is only honest if the yardstick is disclosed — which is why HIE states the window and method behind every z-score rather than presenting an unexplained number.