Median

If you've ever felt that "the average" was being distorted by one crazy day, the median is the statistic you actually wanted.

In 30 Seconds

• The middle value — half the readings are above it, half below.
• Resists outliers — one freak value barely moves it, unlike the average.
• Often the more honest "typical" — especially for skewed or spiky market data.
• Pairs with the mean — mean and median together reveal skew.
• HIE discloses exactly how it's computed — the window and method, printed every time.

Definition

To find the median, line the values up from smallest to largest and take the one in the middle. For 3, 5, and 100, the median is 5 — the middle value — even though the mean (36) is pulled far upward by the 100. That single example captures the median's whole purpose: it reports the center of the data by *position*, not by *arithmetic*, so extreme values don't drag it around.

"Median = the middle value when the data is sorted; half the readings are higher, half lower."

The median is a kind of average — specifically a measure of the center — but a fundamentally different one from the mean. The mean asks "what's the arithmetic balance point?" The median asks "what's the middle of the pack?" When data is symmetric, the two nearly agree. When data is lopsided, they diverge — and the gap between them is itself informative.

Why It Matters

The median matters because markets are full of outliers, and outliers can distort averages. A single extraordinary day — a crash, a spike, a gap — can pull a mean far away from what a typical day actually looks like. That's why the median is often the more honest summary: it reports the middle of the pack, not the arithmetic that a few extremes can drag around. The median shrugs that off: one freak value is just one entry near the end of the sorted list, and the middle barely moves. So when you want an honest sense of "what usually happens," the median is often the truer answer.

This is why the median and the mean are best read *together* — the gap between them is often more informative than either one alone. If they're close, the data is roughly symmetric. If the mean sits well above the median, a few large positive outliers are pulling the average up (the data is right-skewed); if the mean sits below, large negative outliers are pulling it down. That comparison is one of the simplest, most honest ways to detect skew — and markets are frequently skewed.

To ground it with no math: imagine ten quiet trading days and one enormous one. The average "daily move" across those eleven days will look sizeable — inflated entirely by the one big day. The median daily move will stay small, reflecting what ten of the eleven days actually looked like. If someone told you only the average, you'd picture a more active market than really existed. The median keeps the picture honest.

How HIE Computes It

HIE computes the median as a live function — you can compare any feature (price, volume, RSI, ATR) against its own trailing median — and every answer prints the window it used. The median is measured over a strictly-trailing window, using nothing from after the current bar (no look-ahead).

In HIE's disclosure ("what ran"), a median comparison appears in this form:

VOLUME < MEDIAN(VOLUME, window=250, method=median, rolling, strictly-trailing, v1)

— feature, window, method, and version, so two people asking the same question always get the same computation. In the engine's grouping, the median is an *Aggregation* function, alongside the mean, maximum, and minimum.

Each feature carries a disclosed default window from HIE's registry (100 bars for price and RSI, 20 for volume and ATR); you can specify your own, and the window used is always printed. As with every HIE function, if your phrasing is ambiguous, HIE asks rather than guessing.

Try These HIE Questions

The median is most useful as a "compared to its own middle" question — for example, whether a feature is above or below its own recent median.

This query runs in HIE today and returns a certified answer:

• "ES volume below its median" (250-bar window chosen) — resolves to `VOLUME < MEDIAN(VOLUME, window=250, method=median, rolling, strictly-trailing, v1)`.

*(A real answer generated on 17 July 2026, research through that date; because the historical data is periodically refreshed, the exact counts can change over time.)*

Alongside the median, HIE also computes the mean, standard deviation, maximum, and minimum over a trailing window — so you can compare any feature against its center (mean or median), its spread (standard deviation), and its extremes (max/min), all disclosed the same way.

Common Mistakes

"The median and the average are basically the same."

Only when data is symmetric. When it's skewed or has outliers — common in markets — they can differ a lot, and that difference is informative. The mean is the arithmetic balance point; the median is the middle of the pack.

"The median is always better than the mean."

Not always — they answer different questions. The mean accounts for the *size* of every value (useful when magnitudes matter); the median ignores magnitude in favour of position (useful when outliers distort). Good analysis usually looks at both.

"The median predicts the typical next move."

No. The median summarises the center of *past* readings. Like every statistic here, it describes history, not the future.

"There's one true median for a market."

The median depends on the window it's measured over. A 20-bar median and a 250-bar median can differ; both are correct for their window. HIE discloses the window used.

Limitations

The median's strength — ignoring the magnitude of outliers — is also its limitation. Because it only cares about position, it discards information about *how far* the extremes reached. Two very different datasets can share a median if their middles line up, even when their tails behave completely differently. When the size of moves matters, the median alone won't capture it, and pairing it with the mean (and with a spread measure like standard deviation) gives the fuller picture.

Like every statistic here, the median depends on its window and sample, and it describes only the past window it was computed over — never the future.

How This Fits Into MPM

The median is one of HIE's aggregation functions — a more outlier-resistant measure of "typical" than the mean, and a natural partner to it for detecting skew.

You'll encounter it in:

• HIE (Historical Intelligence Engine), where the median is a live function you can compare any feature against, disclosed the same way as every other computation.
• Research Papers, where MPM deliberately leads with medians for outcome statistics, because the typical outcome is more honest than an average that a few extreme cases can distort.
• Intelligence Circle — a member-only research environment containing advanced market research, historical investigations, and trading strategies developed using the MPM research framework.

MPM's broader stance is that the honest measure of "typical" often isn't the average — which is why medians appear throughout MPM's research wherever outliers could otherwise flatter or distort the picture.