Check the results · Rule recognition

Keep-together not working? Check which names were accepted

Published:

Check the name-matching warning first when a keep-together rule seems ineffective. A constraints-on summary does not prove every requested name was accepted. Copy each label from the participant list into the rule, regenerate, and check both the warning and the two assignments.

This guide records a November 3, 2026 browser rehearsal in CalcBE, using invented labels. It is a rule-recognition check, not a new grouping algorithm. Random Team Split's main generator does not offer this binding control; the steps below apply to the named external tool.

Did the rule recognize the names?

We entered six labels, M01 Aster through M06 Grove, one per line, and selected two teams, seeded mode and seed 40. With M01 Aster+M03 Cedar in the keep-together field, Team A contained Aster, Cedar and Birch. Team B contained Fern, Grove and Elm. No missing-name warning appeared. These assignments came from the browser run.

The rule no longer matched. We changed only its M03 to lowercase m03; the participant list still contained uppercase M03 Cedar. CalcBE generated teams but displayed “Some keep-together names were not found and were ignored. (1)”. Its summary still reported constraints=on. Aster was now in Team A with Birch and Grove; Cedar was in Team B with Elm and Fern. The warning and the split appeared together.

The team card alone was misleading. Another recorded rule, M01 Aster+m02 Birch, gave the same missing-name warning, yet Aster and Birch appeared together in Team A. That draw also put Grove with them. Their shared team did not prove that lowercase label had been recognized.

The correction restored the earlier result. We copied the exact M03 Cedar spelling back into the rule: Team A was Aster, Cedar, Birch; Team B was Fern, Grove, Elm. The missing-name warning disappeared. Repeating the lowercase-Cedar run brought back the warning and the separated pair. We kept the same six input lines, two-team setting and seed throughout these comparisons.

What should you check before sharing?

Were the rule names recognized? Did the resulting teams meet your requirement? Check both. Check the former even when the pair happens to be together. If a warning identifies an unmatched name, correct that input first. Do not treat changing the group size as the first remedy for a spelling mismatch.

The attached observation transcript contains the actual form values and visible output for these runs. The replay file lists the browser steps and a local archive check. It transcribes browser observations, not a downloaded CalcBE CSV. We used no real participants.

CalcBE already documents binding syntax and warnings for impossible constraints. RandomWheel describes selecting names from dropdowns instead of typing rule names. Those existing controls are not our invention. This recorded comparison does not establish how other generators match names or what future versions will display.

For two genuine people with the same name, use the separate identity-label guide. That problem concerns who each roster entry represents. This page concerns whether a rule refers to an entry the selected generator actually recognizes.

Recorded comparison: exact Cedar label, accepted and together; lowercase Cedar, ignored and apart; lowercase Birch, ignored but together.
Actual synthetic-label observations, November 3, 2026. Same roster, two teams, seed 40.

Observation transcript and replay

These files transcribe visible browser observations; they are not original CalcBE exports.