Weights partition one probability interval
Each positive weight occupies a segment whose length is proportional to that entry's value. For weights w₁ through wₙ, an entry's one-draw probability is wᵢ ÷ Σw. The page draws one local point from the total interval and selects the segment containing it. This assumes the weights were agreed before the draw and that every listed task is eligible. The numbers express selection frequency, not importance, urgency, or a guarantee about a short sequence.
Worked allocation: 3, 2, and 1
Use Write report: 3, Review pull request: 2, and Tidy backlog: 1. The total is 6, so the three probabilities are 3/6 = 50%, 2/6 = 33.33%, and 1/6 = 16.67%. The cumulative segments are [0,3), [3,5), and [5,6). A random point of 4.37 lands in the review segment; 0.82 lands in report. There is no gap and no overlap, which is the useful arithmetic check before a group accepts the rule.
Ratios matter more than the unit
Weights 0.5, 1.5, and 2 produce exactly the same proportions as 1, 3, and 4. A task estimated at 30 minutes can receive weight 3 beside a 10-minute task with weight 1 if expected workload is the agreed basis. That does not make one individual fairly compensated after a single draw. For equal cumulative burden among named people, maintain a record of earlier allocations; this picker starts from the full distribution each time.
Input errors have direct mathematical effects
A zero or negative weight cannot occupy a positive selection interval, so the form rejects it. Repeating “Email: 1” on two lines creates two separate intervals and doubles Email's chance; combine the lines as Email: 2 when that is intended. Decimal values are allowed, but a percentage sign is not a substitute for a shared scale. Check labels as well as totals: an omitted task has probability zero regardless of how reasonable its weight might have been.
Decide the removal rule before repeated draws
One click samples the stated distribution. A later click is another independent draw from the same list unless the group deliberately removes a completed task or changes its weight. Repeatedly clicking because an outcome feels inconvenient is not a correction to randomness. For a recurring allocation, record whether winners are removed, whether their remaining weights change, and whether a person may decline before the first selection.
The boundary of a transparent weighted draw
The browser can show the selected label, its weight, the total, and its one-draw percentage. It cannot show that the list is complete, that the weights are ethically justified, or that no one edited them after seeing a prior outcome; it keeps no server record. Publish eligibility, the weight source, and redraw conditions for consequential decisions. This method is best for low-stakes choices where unequal odds are deliberate and inspectable.
Cumulative intervals make a weighted result inspectable
For weights 4, 3, 2, and 1, the total is 10. The corresponding cumulative intervals are [0,4), [4,7), [7,9), and [9,10). A point at 6.2 selects the second entry because it lies in a segment of length three; a point at 9.7 selects the last entry. The endpoints have no practical effect on a continuously sampled point, but writing half-open intervals prevents an item from owning a gap or a shared boundary. That is the entire mechanism behind the displayed percentage.
One selection is not a quota system
A task with weight 8 beside a task with weight 2 has an 80% chance on one draw, not a promise of eight wins in the next ten draws. In ten independent selections, the 8-weight task might appear six times or ten times. If a team needs exactly eight report assignments and two review assignments, build that ten-item schedule directly or remove an item after it is selected. Probabilistic frequency and a fixed allocation are different rules with different records.
Choose what the number is allowed to mean
A weight can express estimated effort, urgency, a raffle share, or a deliberately unequal opportunity, but mixing those meanings creates false precision. For example, giving “incident response” weight 5 because it is urgent and “documentation” weight 1 because it is short says nothing coherent about fair workload. Decide on one basis, publish it beside the list, and update all entries together when the basis changes. A decimal such as 2.5 is legitimate arithmetic, not evidence of a more objective judgement.
Repeated work needs state outside this tab
After Write report is selected, leaving its weight at 3 makes it fully eligible for the next click. That is appropriate for independent chances, such as randomly choosing a prompt each day. It is inappropriate when completed work must be removed or when past assignments should reduce a person’s future share. Keep a ledger of completed tasks and revise the list before the next draw. The picker can sample stated odds clearly; it cannot remember an organisation’s workload history or validate the policy that produced the weights.