Spaces:
Running
Running
Commit History
-
Simplify page language and remove posting share
570f994 -
Chart sessions by hour actually running, not just start hour; footnote the methodology; simplify wording
The Evidence 03 chart's peak/valley look wasn't a chart bug -- old-format sessions starting exactly on 6pm/8pm sat between the new format's fixed 5:30/7:30/9:30pm slots. Fixed the real problem instead of restyling it: extract.py now reads each slot's end time too (assuming 2 hours where none is stated, matching all but a handful of sessions that do state both ends) and counts a slot in every hour it's actually running. A 7:30-9:30pm slot now correctly fills in the "6pm gap" it was always covering, producing one smooth curve instead of two dual-axis bar charts. Moved the detailed start-time parsing methodology out of the Evidence 03 panel into a linked footnote under "About this data" -- useful for verifying the numbers, not for reading the page. Also did another pass removing jargon non-native speakers would trip on (oversubscribed, reconstructed, headcount, census, an unglossed "floor", parsing traps, churn, 40-fold, climate constraint) across index.html and stats.html.73f458a -
Combine session-time charts, drop "roster" wording sitewide
Evidence 03 now overlays sign-ups (bars) and slot count (line, secondary axis) on one chart instead of two stacked bar charts -- the line's continuous shape reads more clearly than two separate jagged bar charts did, and it directly shows how count and interest track together. Added the covered date range and noted the multi-slot format is mostly a 2026 pattern. "Roster" read as jargon to non-native speakers; replaced with "membership" / "member list" throughout index.html and stats.html.ee47e44 -
Fix undercounted evening session slots, correct floodlight claim, tighten growth/terms formatting
extract.py was reading one start time per poll, but most evening polls since 30 April 2026 offer three time slots (5:30/7:30/9:30pm) as separate options with their own sign-ups. Reading per-option instead of per-poll raises counted slots from 195 to 368 and surfaces a real 9pm cluster that was being silently dropped. Evidence 03 now shows both a slot-count and a sign-up- weighted view by hour. Also: the venues section wrongly claimed Sadim has no floodlights -- the export shows games running past midnight there, and a week-long park-wide lighting outage (24 Mar-1 Apr 2026) explains the confusion. Reworded that claim, and formatted the roster growth comparison and evidence-02 term definitions for easier reading.bc46f50 -
Sharpen the case-for-courts copy and add a venues section
Rewrites the opening, chart captions and a few overstated claims for clarity, and adds a reproducible venue breakdown (Evidence 04) backed by a new venues() extractor in analysis/extract.py, replacing the unimplemented "10 venues" claim and the redundant month-by-month table. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>5707607 -
Move to its own repo: add the extraction pipeline, notes and deploy skill
The Space now lives on its own with the WhatsApp export beside it in data/, which .gitignore keeps out of version control -- the export carries members' names and phone numbers and the Space is public. analysis/extract.py replaces the ad-hoc scripts the figures came from: it reproduces every number on the page from the export in one run. Two of the page's own counts move to match it (13,609 member-written messages, 943 membership events), and the About note now states plainly that the roster reconstruction lands five above the 505 WhatsApp reports rather than claiming to reconcile exactly.67b39eb
-
Rework the growth, utilisation and timing evidence
- Evidence 01: replace the days-per-100 bars (where shorter meant better) with the daily roster curve plus milestone brackets on a real time axis, and a net-members-per-month chart where taller means faster - Evidence 02: define organised session, sign-up, active member and at capacity; lay the argument out in four steps; add the capacity evidence (129 capped slots, 26 filled, 15 oversubscribed) and note that the winter poll dip is a change in coordination style, not a lull - Evidence 03: rebuild on the start time of each organised session (195 of 253) instead of every time-like string in every message, which is what produced the phantom 3-4am play; document the reading rulesb0fa454 -
Add ADPC growth analysis and the case for public courts
Verified signature72448e1 -
initial commit
Verified signaturee0b7133