▤ Does my horoscope actually describe me? Four lines a day, one of them yours

Every horoscope sounds a little like you. Umbriel's free Decoy Day puts that to the test: four short lines about today, one built from the real sky against your own chart, three built from the same sky against charts that belong to nobody. In the evening you say which one fitted, and only then does the real one step forward with the contact it came from. Over time it shows your rate against the 25% four lines give you by luck, with the error bar drawn.

• What you'll needYour birth date, time and place. Nothing else, and no account.
• What you getFour unlabelled lines every morning; an evening reveal naming the real one, the transiting planet behind it and its orb to the degree and minute; and, once thirty rounds are answered, a band showing your rate with its 95% interval against the quarter chance alone gives
• CostFree, no account required

The oldest objection to a horoscope is the fairest one: it sounds like you because it would sound like anyone. Warm, a little specific, impossible to be wrong about. Nobody has ever been able to check it, because there has never been anything to compare it against.

Decoy Day gives you the comparison. Each morning it works out the tightest contact the real sky is making to your own chart today, and writes it as one short line. Then it does exactly the same thing for three invented charts under exactly the same sky, and writes those three lines with the same function, in the same words, at the same length. Four lines sit on the page, unlabelled, in an order shuffled before you see it.

The decoys are not filler and they are not vague. Every one of them is a real transit, from a real planet, that is genuinely happening in the sky at that hour. The only thing that makes yours yours is the chart underneath it. That is the whole design: if the reading is doing nothing, all four should feel about equally plausible, and a quarter of your taps will land on the real one.

A test like this is only worth having if it cannot be gamed, so the things that could give it away are measured rather than hoped for. Over 150 consecutive rounds the real line is the tightest of the four about a quarter of the time, sits in each of the four positions about a quarter of the time, and never repeats within three rounds. Only the five fast movers transit, because a slow planet sits inside orb for weeks and your line would come back while the strangers' lines kept changing. If you have no birth time and therefore no Ascendant, the invented charts are read without one too.

The answer waits until the evening, so the day gets to happen before you judge it. Then you tap the line that fitted, or "none of them", which is a real answer and is counted on its own rather than as a miss. The real line is named, with the transiting body, the aspect and the orb in degrees and minutes, so you can check it against the transit page yourself.

And the result is shown as a band, never a bare number. Below thirty rounds nothing is drawn at all, because under thirty the honest range around any rate is wider than fifteen points either side, which would tell you about the arithmetic rather than about you. Above it you get your rate, the 95% interval around it, and a dashed rule at the 25% four lines give you with your eyes shut. If the quarter sits inside your interval, the page says so plainly: this is chance, as far as that many rounds can see, and that is the ordinary result rather than a mark against anybody.

The sky is computed once, for local noon of that day, by the same VSOP87 and ELP-2000 engine that draws every chart on this site, so the round is the same round whether you open it at breakfast or at lunch. Against that one sky, each of four charts is asked the same question: which of the Moon, Sun, Mercury, Venus and Mars is closest to an exact conjunction, opposition, square, trine or sextile with one of the fourteen points Umbriel writes about, and how close. The answer becomes one line in the app's single reading vocabulary, the same table that writes the daily sky and the weekly read, so no line can be more specific, longer or more flattering than another. THE DECOY CHARTS ARE INVENTED, and invented by a stream seeded from the date and your own chart, so they are fixed for you on that day and cannot be turned over by reloading. They are not anybody's birth data and they are never a saved person. THE TELLS ARE MEASURED, NOT ASSUMED. Over 150 consecutive rounds driven through the real code: your own line is the tightest of the four on 37 of them, where a quarter is 37 or 38; it falls in the four positions 29, 40, 42 and 39 times; no line shown in the last three rounds is used again in either role; and your own contact comes round again within ten rounds 4 times against a stranger's 16, out of the same ten chances. Only the five fast movers transit for exactly this reason: with the slow planets included, your own contact returns 28 times in 150 rounds against a stranger's 7, and a returning face is an answer. The position is shuffled and the real slot written down before anything is drawn. THE ARITHMETIC. Under the assumption that a reader cannot tell, the chance of tapping the real line is exactly one in four, because which of the four is real is independent of everything on the screen. The rate is your hits over the rounds where you tapped one of the four; rounds answered "none of them" are counted separately and left out of it rather than counted as misses. The interval drawn is the Wilson score interval at 95%, written out rather than approximated, because the usual normal approximation is badly wrong near a quarter at small samples. The odds against chance come from the binomial tail, the same one the Jury prints for its clocks, so there is one implementation of it here and not two. The band opens at thirty rounds because at thirty the interval is about fifteen points either side of the rate, and below that it is wider than the thing it is measuring.