Money & sales
Recent completed nights — each vs. its own usual
A strip of up to 12 recent nights, each plotted against its own same-weekday usual — a Monday against Mondays, a Saturday against Saturdays. No shared baseline, no trend line: it shows the points, it does not connect them.
What it is
Recent completed nights — each vs. its own usual is a strip of up to twelve of a venue's most recent completed nights, where every dot is that one night's stored comparison against its own same-weekday usual — a Monday point measured against Mondays, a Saturday point against Saturdays. Each dot is a canonical Completed nights vs. usual (M26) reading, already computed and frozen, simply placed on a timeline by its date.
There is no shared baseline across the dots and no trend is calculated. It shows you the points; it does not connect them. The strip exists so you can see a run of recent nights at a glance and notice whether any single one sat unusually high or low against its own weekday — not to draw a line through them.
How to use it at your venue
A grounded use is to scan the strip for a single night that stands away from its own weekday usual and ask what was different about that night — a promotion, weather, a holiday, a staffing change — treating each dot as its own question rather than reading the shape of the whole run. What operators are cautioned against is the read the eye wants to make: the human mind reliably sees streaks and momentum in short, noisy sequences that carry none, a bias documented as the "law of small numbers"[1] and echoed in the "hot hand" misperception, where even trained professionals infer non-existent runs from a handful of outcomes.[2]
If the strip looks like it is climbing or falling, the suggestive next step is to keep watching for confirmation across more nights and cross-check against known causes before treating it as a shift in the business. The quality-control field reached the same discipline independently: by a standard statistical-process-control convention, a sequence is not called a genuine trend until roughly six or more consecutive points move the same way, below which the movement is read as ordinary noise.[3] This strip deliberately stops short of that judgment and leaves the trend call to a purpose-built test, if and when one exists, with a real significance threshold rather than an eyeballed sparkline.
How to read it
Read each dot on its own, against its own weekday. A higher dot means that night came in above its same-weekday usual; a lower dot, below — nothing more. The dots are never compared with each other, venues are never compared by dot height, and no line connects them, so the vertical position of one dot relative to the next is not a reading you can trust. The gaps between dots are just dates with no plottable night — closures and missing nights are empty space on the timeline, not zeros.
What it doesn't mean
⚠ What it can't tell you
A short history is noise. Higher or lower dots do not by themselves mean better, worse, or actionable, and no slope, average change, improvement, decline, anomaly, or stability measure is defined, calculated, or shown — each of those would need its own ratified definition. It cannot say why any night differed, whether nights are "trending," or anything about the empty time between dots. It never compares venues to each other, and a missing night is a state, not a zero.
How it's calculated
Loaded on demand: the venue's most recent completed nights, up to twelve, are read, and each night's already-stored comparison is placed on the timeline by its date. Nothing is recalculated. A night plots only if its stored comparison still matches the night's current figures; a night whose figures were later corrected is withdrawn from the strip and counted, never silently redrawn. The strip sequences existing frozen objects — it computes nothing new.
The exact method
Each dot is a canonical Completed nights vs. usual (M26) object produced when that night finalised, carrying its comparison against the median of the venue's past nights on the same weekday. This strip performs no arithmetic: it reads the ≤12 stored objects client-side and plots only those ratified points whose stored anchor-net still matches the night's current figures — a corrected night is withdrawn and counted, not re-plotted.
Floor: four plottable nights before a strip renders. One to three dots invite reading noise as pattern, so the floor withholds the strip; it never creates a judgment. Cap: twelve nights, which bounds the read cost and is named in the copy whenever more history exists. Each point uses a same-weekday baseline (Mondays to Mondays, Saturdays to Saturdays) — there is no shared baseline across dots and no connecting line. Version m28.points12.floor4.v1.
The research behind it
A handful of data points can look like a story even when it's just noise — our brains are wired to see streaks and trends in short runs of numbers whether or not they're really there. Quality-control engineers learned this decades ago and set a rule: don't call it a trend until you've seen enough points moving the same way in a row. This metric applies that same discipline — it shows you the raw recent nights so you can spot an unusual night, but it deliberately does not draw a trend line through too few points, because that line would be more likely to mislead than inform.
-
Belief in the Law of Small Numbers ✓ verified
Read the source ↗ -
The Hot Hand in Basketball: On the Misperception of Random Sequences ✓ verified
Read the source ↗ -
The Shewhart Control Chart — Tests for Special Causes ✓ verified
Read the source ↗
Confidence & caveats
Each dot is exactly as sound as the frozen M26 comparison beneath it — no more, no less — because this strip adds no computation of its own. What it does not carry is any directional claim: there is no trend, slope, average change, anomaly, or stability reading here, and none is implied by the arrangement of the dots. The four-night floor withholds the strip below adequacy but is not itself a verdict, and the twelve-night cap is a cost bound, not a statement that older nights don't matter. A gap is missing data, not a low night, and the strip never compares one venue to another.