Weights define intervals, not popularity
A weighted draw turns positive weights into adjacent intervals on one number line. Tea: 3, Coffee: 5, Water: 2 has total weight 10. Tea occupies the interval from 0 up to 3, coffee from 3 up to 8, and water from 8 up to 10. The browser chooses one uniformly random point on that total interval and returns the option containing it. The labels may be votes, inventory counts, raffle entries, or deliberately chosen odds; the mechanism does not infer their meaning. A weight is relative, so changing every 3, 5, 2 to 30, 50, 20 changes nothing.
The probability formula
If option i has positive weight wᵢ and all entered weights add to W, its probability is wᵢ / W. Percent is 100 × wᵢ / W. The probabilities always add to 1 because the intervals cover the whole length W without overlap. This is why zero and negative weights are rejected: a zero-width interval can never win, while a negative interval has no coherent probability interpretation. Decimal weights are valid when they express a ratio. For example, 0.5, 1, and 1.5 total 3, giving chances of one sixth, one third, and one half.
Worked draw with real numbers
Use the default Tea: 3, Coffee: 5, Water: 2. Total W is 3 + 5 + 2 = 10, so the chances are 30%, 50%, and 20%. Suppose the internal random point is 6.42. It lies in coffee’s interval [3, 8), therefore Coffee is selected and the page states “Weight 5 of 10” and “50.00%.” If the point had been 1.10, Tea would win; 9.30 would choose Water. The displayed percentage belongs to the stated weights before the draw. It is not a promise that five coffee wins will appear in ten repetitions.
Good uses and transparent setup
Use weighted choice when unequal chances are intentional and can be explained: a game loot table, a classroom prompt bag with extra copies of difficult topics, a menu selector that gives available ingredients more tickets, or a casual prize wheel with published entries. Put one option on each line as “name: weight”; commas or equals signs also work as separators. Read the entries and total aloud for a group draw, then record the outcome before clicking again. Keeping weights visible matters: a random result cannot make an undisclosed bias fair. For regulated raffles, gambling, procurement, or prize promotion, use the procedure and records required by the relevant rules instead.
Frequent mistakes
Do not enter percentages unless their scale is intended as weights. 20, 30, 50 is fine because their sum happens to be 100, but 20, 30, 50, 10 becomes 18.18%, 27.27%, 45.45%, and 9.09%, not the first three original percentages. Do not write “Coffee: five” because the parser needs a numeric weight. Repeating “Coffee” on separate lines technically gives it more interval length, but hides the total and is harder to audit than Coffee: 10. Extremely large weights are unnecessary; reducing a common factor preserves every chance and makes mistakes easier to spot.
When unequal odds are the rule
An equal random list picker gives every entered item the same probability. This page intentionally makes those probabilities unequal and shows the arithmetic for the selected item. A spin wheel is mainly a visual presentation of entries; the weighted draw here is a compact numerical rule. A decision matrix tie breaker first filters to options tied on score, then chooses equally among that filtered set. Use this page only after deciding that different chances are justified. The names and weights are processed locally and never posted to a server.
Auditing a weighted result
A useful draw record contains the exact option labels, the weights, their sum, the time the rules were agreed, and the selected result. The random point itself is not shown because it is only an implementation detail; the interval method and the winner’s fraction are enough to check the published odds. If a group wants equal voting power, do not manufacture weights from enthusiasm or speaking time—use an equal picker instead. If weights represent tickets, count the tickets before the draw and freeze the list. The arithmetic is transparent but cannot validate the source data: five tickets entered for an ineligible person still gives that person a real 50% chance in a ten-ticket draw. Fair mechanics and fair eligibility are separate responsibilities.
Long-run frequency versus one draw
A 50% stated chance is a property of the procedure before selection, not a quota for the next two draws. Coffee can lose ten consecutive times in the 3:5:2 example even though that particular sequence is unlikely; every draw starts with the same intervals again if the weights remain fixed. Conversely, selecting Coffee once does not consume its probability or make Tea “due.” To examine whether a mechanism behaves as intended, run many independent trials and compare observed shares with the target fractions, while allowing ordinary sampling variation. For an event that chooses one winner, however, the only meaningful audit is whether the agreed weights were used for that one draw. Do not turn a probability statement into a claim that a selected outcome was deserved.
Weights do not test eligibility
A numerical ratio can represent entries after eligibility has been decided, but it cannot decide eligibility by itself. Imagine a volunteer draw where one sponsor contributes three tickets and each of two others contributes one. The interval rule gives chances 3/5, 1/5, and 1/5 only after the organizer has verified that all three sponsors may participate and that three is the agreed ticket count. A misspelled label, a duplicate line, or an unannounced late entry changes the population rather than merely the display. Freeze the list and retain it with the result. Use a written event procedure when a prize or allocation has material consequences.
More chance tools
Random Team Splitter by Size · Random Date and Time Generator