Meet the league
The matchups layer of a US community impact platform: a season long table where 12 cities compete through real basketball fixtures, and every result moves the standings. The league had just been reshaped to the owner exact city set: Denver, Portland, Houston, New Orleans, Phoenix, Los Angeles, Philadelphia, New York, Detroit, Chicago, Washington and Atlanta. One former city came out entirely.
The challenge
A standings table built on fixtures nobody verified is fiction with good design. The city change made it worse: the outgoing city had 13 recorded finals and 30 game sightings sitting inside the ledger, mixed into the same data the live table reads. Ship the new city set on the old ledger and the table would quietly score games for a city that no longer exists in the league.
The solution
Rebuild the fixture list, then check every game
The fixture list was rebuilt for the 12 cities and 14 teams, and then every single fixture was cross checked against Basketball-Reference, the reference source for the schedule: 237 fixtures, 237 matching. At build time the set held 23 finals locked, with the rest live, provisional or upcoming, and the hero counts on the page were read from the build itself rather than typed in by hand.
Clean the ledger, archive do not delete
The outgoing city 30 game sightings were pruned from the ledger, taking it from 267 entries to exactly 237. Its 13 finals were not deleted: they were preserved in a finals archive, intact, where they no longer count in the standings but remain on record. A re-run audit after the cleanup came back clean: 0 mismatches, 0 schedule changes, 0 missing games.
Re-base the QA suite for the new league
The automated checks were rebuilt for the new city set and run green: 112 checks in the matchups suite and 23 in the attribution suite, 135 passing, 0 failed, including checks on multi game days across the season. The finished pages were deployed and verified live on both domains by checksum, with the removed city name appearing zero times anywhere on the shipped pages.
The results
- 237 of 237 fixtures matching Basketball-Reference
- 12 cities, 14 teams, one league table
- 135 QA checks passing, 0 failed
- 30 stale sightings pruned, 13 old finals preserved in an archive
- 0 mismatches in the final audit
Later heartbeat audits the same week still read 237 of 237 against the reference source as results accumulated, which is what a verified fixture base is for: the table can move every day without anyone re-asking whether the games underneath it are real.
The return: trust compounds. Every standing, streak and headline the platform publishes on top of this table inherits the verification underneath it, and the next city change now has a proven procedure instead of a hopeful one.
FAQ
Why verify fixtures that come from a schedule feed?
Because feeds drift, ledgers accumulate stale sightings, and a city change silently invalidates part of the set. The cross check is what turns a list of games into a table you can publish standings on.
What happened to the removed city results?
They were archived, not deleted. Its 13 finals sit in a preserved ledger where they no longer affect the standings, so the history exists without distorting the current league.
How is the table kept honest after launch?
Heartbeat audits re-run the same cross check against the reference source. The audits later in launch week still read 237 of 237 as finals accumulated.
What do the QA suites check?
Fixture integrity and attribution: counts per day, locked finals, city membership, and that every published number on the page can be traced back to the verified fixture set.
