How a daily puzzle is built: the library, the draw and the fairness rules

At midnight, your local time, Stat! serves a new set of puzzles: a case in Syndrome, a hidden disease in Traits, a board in Associations and three structures in Tangent. None is written by hand on the day: each comes from a library built for the game, assembled by rules that keep it solvable and fair. This article explains how, without spoiling any day.

Key points

• Every puzzle comes from one curated library of 250 diseases, 25 in each of ten body systems, and 349 anatomical structures.
• A program chooses each day's puzzles in advance from the game and the date, so everyone gets the same puzzles, with rules against repeats.
• Associations boards are built so every card has exactly one right group, and Tangent scores a guess by its distance from the target on the body map.

One library behind four games

Every puzzle starts from one curated library. It holds 250 diseases, 25 in each of ten body systems, and each is recorded the same way:

The findings share one vocabulary: every symptom, lab test, vital sign and imaging study in the library comes from the menu you search in Syndrome, which lists 193 symptoms, 91 lab tests, 6 vital signs and 20 imaging studies. So every finding apart from the examination signs is something you can ask about.

The library also holds 349 anatomical structures: 56 organs, 73 bones, 75 muscles, 65 blood vessels and 80 nerves. Each has a short description, a front-to-back depth and a hand-drawn place on the body map.

One date, the same puzzles for everyone

A program chooses each day's puzzles in advance, at random but reproducibly: it starts from the game and the date, so a given date always produces the same picks. Your device looks up the puzzles for its own calendar date, which is why everyone shares them and why they change at local midnight. Every pick is checked against the library before publication.

A few rules shape the draw:

• Syndrome and Traits each pick one disease from all 250, never the same disease two days running in the same game.
• Associations picks four diseases, none of them from the previous day's board.
• Tangent picks one structure for each of its three tiers, never the same structure two days running in a tier.

Even Syndrome's patient is shared: the age, sex and complaint are generated reproducibly from the date.

Syndrome: a case built from the record

The presentation comes from the hidden disease's record. The complaint is one of its listed symptoms, the age falls inside its age range, and the sex follows the record: a disease recorded for one sex always presents in that sex, and otherwise the case picks one at random. So the opening line already rules diseases out.

Each question is answered from the record:

• A symptom is present if the disease lists it, and absent if not. A general term such as cough also matches a more specific listed form.
• A lab is high or low if the record lists it, and normal if not. The wording follows how the test is reported: 26 yes-or-no tests, such as cultures and antibody tests, read positive or negative, and a urinalysis or blood smear reads abnormal or normal.
• A vital sign or imaging study is abnormal if the record lists it, and normal if not.

This is why absent findings matter. A typical disease lists about ten of the roughly 310 items on the menu, so most possible questions come back grey. Yet each grey answer rules out every disease in the library that lists that finding: in Syndrome, an absent symptom or a normal test narrows your list as reliably as a positive one. Real tests are less tidy, as our guide to reading lab results explains.

Examination signs are never on the menu. During a case they appear only through hints, which reveal a random finding you have not asked about yet. The Syndrome guide covers strategy.

Traits: six comparisons per guess

Each guess is compared with the hidden disease on six traits. System and type hold one value each, so they are either exact or a miss. Acuity, organs, diagnosis and treatment can hold several values: the same set is exact, at least one value in common is partial, and nothing in common is a miss.

How common a value is shapes what a color tells you. All but one of the 250 diseases list "clinical" among their diagnosis values, so diagnosis is almost never grey; the signal is whether it turns green. Organs is the opposite: two diseases picked at random share no organ about nine times in ten, so amber or green there is a strong lead.

Traits cannot separate every disease: in the current library, 52 of the 250 share all six values with another disease, which is why a full green row is not always a win. For strategy, see the Traits guide.

Associations: one right group for every card

A board holds four diseases with four cards each. Every card is a finding from its disease's record, and every group includes at least one examination sign.

The central rule is fairness: each card has exactly one right group. When a daily board is built, no card may be a finding that another disease on the board also lists, and no card text may appear twice; if four diseases cannot make a fair board, the program draws another four. The check also counts a symptom and the vital sign that expresses it as the same finding, so a fever card never belongs to one disease while another on the board lists an abnormal temperature. Heart rate, blood pressure, breathing rate and oxygen saturation work the same way.

Fair means fair by the library: in real medicine a card can still seem to fit two diseases, as our guide to classic clinical associations shows.

Then comes the shuffle. A board is put together one disease at a time, so unshuffled, each group's cards would sit side by side. The cards are shuffled on your own device, separately for each player, so a card's position says nothing about its group, and two friends playing the same board see different layouts. The Associations guide has tips for solving the board.

Tangent: distance measured on the map

Every Tangent structure is placed by hand on a front view of the body, using landmarks such as the sternal notch, the nipple line, the costal margin and the umbilicus (see our guide to surface anatomy), with each spinal nerve root at its vertebral level. Paired structures appear on both sides, one-sided structures on their own side, and structures at the back are drawn dashed.

Closeness is measured on those same drawings, so what looks close on the map scores close. The distance combines three things:

1. How near the two structures come on the map, between their nearest parts.
2. How spread out the target is. Part of the distance reflects how far the target's parts lie from your guess on average, so touching one end of a long vessel or nerve scores lower than running alongside it.
3. How far apart they are from front to back. A guess that covers the target on the map but lies half the body's depth in front of or behind it scores in the low 70s at best.

The skin, which covers the whole body, is a special case: closeness to it depends only on how near a guess lies to the front or back surface.

Only the exact answer scores 100. Neighbors that touch at the same depth usually score in the 90s, a typical guess lands mid-scale, and the far end of the body can score in single digits. The scale is steepest near the target, so small moves count most when you are already close. The Tangent guide lists the color bands.

Why four games

The four games come at the library from four directions: from findings to a diagnosis in Syndrome, through a disease's categories in Traits, across clusters of findings in Associations and around the body in Tangent. A disease can return in a different game or with a different presentation. Our article on learning with daily puzzles explains why mixed, spaced practice like this helps you remember.

Simplified on purpose

To make every question answerable, the library simplifies: symptoms are present or absent, labs are high, low or normal, and each disease has one system and one type. Real patients are messier. Stat! is a learning game, and nothing in it is medical advice.

Questions or corrections

Spotted a mistake in a puzzle or this article? Email stat@blottergames.com with the date, the game and what looks wrong. We check every report.