Random Picker

Introduction:

A draw stops being fair when its real rules are hidden. Equal chance, weighted chance, replacement, exclusions, guaranteed entries, and group limits describe different selection problems. Any of them can be reasonable, but the result only makes sense when everyone knows which rules applied.

Without replacement means a source row can appear at most once in a draw. It suits winner lists, classroom turns, and samples that need distinct entries. With replacement returns the selected row to the pool, so the same row may occupy several slots. That policy fits repeated trials, not a unique-winner list.

Weight

A relative chance multiplier. A weight of 3 has three times the draw influence of a weight of 1 while both rows remain eligible.

Group cap

A maximum number of selected rows sharing the same group label.

Seed

Text that fixes the pseudorandom sequence so the same eligible list and settings can replay the same draw.

Inclusion chance

The chance that a row appears at least once, which is different from its chance on any single pick.

Constraints interact. A must-include row consumes one slot and may also consume its group allowance. Exclusions remove rows before selection. Deduplication can merge repeated labels, while case-sensitive matching can keep names such as “AL” and “Al” distinct. These choices change the eligible population before randomness begins.

Weights are not percentages. Their effect depends on every other eligible weight and, without replacement, changes after each row is selected. Group caps and must-include rows make the exact inclusion chance more complicated still. A simulated odds estimate is therefore useful for comparison, but it is not a mathematical guarantee.

A replayable seed proves that a published result follows from a particular list and configuration. It does not prove that the setup was impartial, that weights were justified, or that the generator is suitable for security, gambling, or regulated allocation. Consequential draws need an agreed source list and rules before the seed is chosen.

How to Use This Tool:

Prepare the candidate list and selection policy before drawing. Keep each candidate on its own non-empty line; add optional attributes after the label only when weights or groups are part of the rules.

1. Paste the Candidate list, or load a local TXT or CSV file. Use entries such as Alice | weight=3 | group=Team A when a row needs explicit attributes.
2. Set Number of picks and choose Without replacement for distinct rows or With replacement when repeats are allowed.
3. Turn on Weighted mode only after reviewing every explicit weight and the Default weight. Leave weights off when every eligible source row should have equal influence.
4. Add exclusions, must-include labels, or a per-group maximum under Advanced. Choose case handling and deduplication before the draw because both can change which rows remain eligible.
   A must-include label is matched exactly after trimming and the selected case rule. It can still be blocked by an exclusion, a duplicate policy, or a group cap.
5. Enter a Replay seed when the draw must be reproducible, then press Pick. Copy the reported run seed together with the list and settings when the result needs an audit trail.
6. Check the selected count and any unfilled slots before using the manifest. Increase Odds simulation runs only when a steadier comparison is worth the extra local computation.

Interpreting Results:

The selected count is authoritative for the completed draw. If it is lower than the requested count, the remaining eligible rows could not satisfy the replacement and group rules. Do not treat an unfilled slot as another random outcome; change the constraints or request fewer picks.

The manifest keeps each selected row's original draw slot even when alphabetical sorting changes the display order. “Must include” identifies a forced match, while “Random draw” identifies a row chosen after forced entries were placed.

Inclusion percentages compare candidates under the current configuration across the chosen number of seeded trials. Small differences can be simulation noise. Raise the trial count and keep the list, settings, and run seed fixed before comparing two estimates.

Technical Details:

Selection begins with normalization, not randomness. Blank lines are ignored; labels are trimmed; optional weights must be from 0.001 to 1,000,000; and group text is retained as a constraint label. Exclusions are then applied. The eligible population is limited to 500 rows, and the requested draw is limited to 1 to 500 slots.

Rule Core:

1. Parse the source rows, then apply the chosen case and deduplication rules.
2. Remove exact label matches from the exclusion list.
3. Process must-include labels in their listed order. A matching row is added only if it has not already been used under no-replacement policy and its group still has capacity.
4. Fill remaining slots using the selected uniform or weighted method. Without replacement removes chosen source rows; with replacement leaves them eligible.
5. Stop when the requested count is reached or no row can pass the remaining constraints.

A zero group cap means no cap. A positive cap applies only to rows with a non-empty group label. Ungrouped rows remain eligible regardless of grouped selections.

Formula Core:

Uniform selection without replacement uses a Fisher–Yates shuffle. Weighted selection without replacement assigns each remaining row a priority key and takes the largest keys:

Here, U is a pseudorandom value between 0 and 1 and w is the row's effective weight. With replacement, weighted selection instead draws from the cumulative eligible weight on every slot, so a row may be selected repeatedly.

The inclusion estimate counts whether a candidate appeared at least once in each simulated draw:

c is the number of simulated draws containing the row and R is the run count, from 50 to 3,000. The displayed percentage keeps full calculation precision until formatting.

Reproducibility mechanism:

Seed text is converted to a 32-bit state, then a deterministic pseudorandom sequence drives the draw. Reusing the reported run seed with identical candidates and settings repeats the result. Leaving the seed blank creates a fresh local run seed; recording that generated seed is necessary for later replay.

Limitations and Privacy Notes:

The draw is a planning and selection aid, not a cryptographic lottery or proof of procedural fairness.

• Candidate text and loaded files are processed in the browser; the selection does not require server-side processing.
• Simulated inclusion percentages are estimates, especially when weights, replacement, must-include entries, and group caps interact.
• A seed makes a sequence repeatable. Anyone who can test candidate seeds before publication may be able to choose a favorable result.
• Outcomes have no monetary value and should not decide regulated, legal, financial, safety-critical, or high-stakes opportunities without an independently governed process.

References:

• Weighted Random Sampling with a Reservoir, Information Processing Letters, 2006.