Vanish: designing a card that disappears

Vanish shows you a hand of cards, takes it away, then gives it back with one card missing. You name the card that left, from a line-up of five. There is very little else to it, which means almost every decision in the game comes down to two numbers: how many cards are on the table, and how long you get to look at them.

Two dials, one at a time

The first version of the curve moved both numbers every round. The hand grew by a card and the look got shorter, round after round, and the two of them compound faster than they look on paper. By the eighth round the game was dealing eleven cards and taking them away after about five and a half seconds. That is not so much a difficult round as a round nobody finishes, and the last third of the game had turned into watching a run end.

So the dials were separated. The Daily now grows the hand every other round: four cards for the first two rounds, five for the next two, six for the two after that, and seven for the last two. The look keeps shortening every round. One number moves while the other waits its turn, which is enough to make each round a step up from the one before without the two of them multiplying.

More puzzles offers three curves to choose from. Easy starts at three cards and grows the hand every third round. Medium starts at four and grows every second round, which is what the Daily plays. Hard starts at four and grows every round: the original escalation, kept intact, for anyone who would like to meet it.

How long the look lasts

The look is not a flat number of seconds. It is a budget per card. In the first round of a Daily you get a full second for each of the four cards, so the hand sits on screen for four seconds. Every round after that takes forty milliseconds off the per-card figure, and it never drops below 0.65 seconds however long you last.

By the eighth round the per-card look is down to about 0.72 seconds, but there are seven cards to spend it on, so the hand is actually on screen for roughly five seconds, longer than it was in round one. The round gets harder while the clock gets more generous. A flat five seconds would have been a gift with four cards and a joke with seven; a per-card budget makes each round ask for a little more than the last.

Why the third miss ends it

When the hand comes back it is one card short and reshuffled, and the vanished card is offered alongside four others. Those four are drawn at random from the rest of the deck, and nothing is done to make them resemble the answer. An earlier line-up tried to be clever about that: it guaranteed the vanished card a companion of the same rank and one of the same suit, which quietly made it the only option with both. A solver that ignored the hand entirely picked it about two thirds of the time. Uniform decoys are duller to build and far better to play, because every option is an equally plausible completion of the cards in front of you.

Five options also mean a blind guess comes off one time in five, so a wrong pick has to cost something. It costs a miss, and the third miss ends the run. Three is small enough that a late round still matters, and large enough to forgive the round where you blinked. A run that stops at round six is an ordinary result and reads as one: how far you got, how many cards you named, how long you took.

When the cards turned out to be too small

The curve was settled long before the board was. On a 375 by 812 phone, with the largest hand any curve deals, the input phase still finished 305 pixels above the bottom of the screen. The layout fitted beautifully. It had just paid for all that room in the wrong currency: card values 13 pixels tall inside 44 pixel cards, which is fine to glance at and genuinely hard to read in three quarters of a second.

The empty space at the bottom of the screen was doing no work at all. The card faces were doing all of it.

So the cards grew into the room they already had. Hand cards went from 66 to 74 pixels wide with their values at 22 pixels, the remainder row from 44 to 52 pixels at 16, and the faces in the line-up from 19 to 23. Each of those widths is the largest that still keeps four hand cards and five remainder cards on one row inside the 327 pixel content column of a 375 pixel phone. One more pixel wraps a row, and a wrapped row costs more height than the extra size is worth.

One size was held back on purpose. The remainder stays a step smaller than the line-up, because when the two rows were sized alike they read as a single block, and a row that is only there to be read started collecting taps. A few pixels of difference do the job a label would otherwise have to do, which is more or less the whole trick of this game: what you can take in at a glance, and what you cannot.