About Pick For Me
A picker that is genuinely fair. Type or paste a list on the left (one per line, or separated by commas), or start from a sample, then pick how to choose on the right and press the big button (or S, or click the wheel).
- ONE: spin the wheel for one pick.
- SEVERAL: that many different picks, one spin each. Each name the wheel lands on is marked 1st, 2nd, 3rd… and can’t come up again in that set. The first spin is as long as you set, the rest are quicker, and pressing the button again makes the rest at once.
- NO REPEATS: everyone gets a go before anyone goes twice. Each spin takes the last winner off the wheel; when the wheel is empty a new round starts. The page shows who has gone and who is still to go.
- TEAMS: split the list into teams of a size, or into a number of teams. Sizes never differ by more than one (10 in teams of 3 makes three teams of 4, 3 and 3). This is a fair shuffle, so there is no spin: the teams are coloured round the wheel.
- ORDER: shuffle everyone into an order, numbered round the wheel. A fair shuffle, no spin.
If something is on the list twice it gets two slices of the wheel, and twice the chance; the page points that out. IS IT FAIR? makes 1,000 to 100,000 picks from your list and counts them against Expected. ASK THE BALL is a fortune-telling ball for yes-or-no questions.
Your list stays on this device. It is kept in this browser so it's there next time, and it is never sent anywhere or put in the address. FORGET wipes it.
THE STORY
Choosing by lot is very old. Ancient Athens picked most of its officials and its huge juries by lot, and from the 4th century BC used a stone allotment machine, the kleroterion: citizens' name tokens went into slots in rows, and black or white balls dropped through a tube decided which rows served. Lots were thought fairer than votes, which the rich and well-spoken could win. The wheel is old too: the Wheel of Fortune, turned by the goddess Fortuna and lifting people up and throwing them down, runs through medieval art after Boethius wrote of it around 524.
The page grew from two small Python scripts. Subject Picker.py picks a revision subject and then a topic within it (its five subjects and twenty topics are the "Revision" sample here). Picking in two steps like that is only fair because every subject has four topics: give one subject two topics and each of those would come up twice as often as a topic in a four-topic subject. The page puts everything on one wheel instead. The fortune ball comes from magic8ball.py and Magic 8-Ball.py, after the American fortune-telling toy, made in its 8-ball shape from 1950. Inside today's version a 20-sided die floats in dark liquid, with ten answers leaning yes, five leaning no and five "ask again". Its wording here is our own. One of the scripts picks with randint(0, 20), which in Python includes 20, so 1 time in 21 it reaches for a 21st answer that isn't there and stops with an error.
HOW IT WORKS
The winner is decided first, then the wheel is spun to land on it; how long it spins makes no difference. Every pick uses crypto.getRandomValues, the browser's random source for security work, rather than the quicker everyday one.
Turning a random number into a fair pick takes care. The random source gives whole numbers from 0 to 4,294,967,295. The obvious move, the remainder after dividing by the number of entries, is slightly unfair whenever the entries don't divide that range exactly: the first few entries get one more way to come up than the rest. So the page throws away the few numbers at the very top that don't make a whole round and draws again. With a big range the unfairness would be tiny; with a small one it is large, which is what BYTE REMAINDER in the fairness test shows: a single random byte (0 to 255) and its remainder. With 30 entries, 256 = 8 × 30 + 16, so entries 1 to 16 get 9 chances in 256 and entries 17 to 30 get 8: 12.5% more. With more than 256 entries, the ones past 256 can never win.
Shuffles, orders and teams use the Fisher–Yates shuffle: go down the list from the end, swapping each place with a random place at or before it. Ronald Fisher and Frank Yates gave the pencil-and-paper version in 1938; Richard Durstenfeld's 1964 version for computers is the one used here, and every order is equally likely. Several different picks are the first few steps of that shuffle (the same as drawing one at a time from the names not yet picked), each one then shown by its own spin; teams are a shuffle dealt out round the teams like cards.
The fairness test counts each entry and compares the counts with Expected (picks ÷ entries) using Pearson's chi-squared test: it says how often a fair picker would spread out at least this much by chance. A fair picker is rarely spot on; it wobbles about Expected, by a smaller share the more picks you make.
FURTHER WATCHING AND READING