The survival analysis dataset the Titanic era deserves a successor to: 1,307 Broadway shows with the four columns the technique teaches on: opening date, closing date, performance count, and a Still_Running flag that IS the censoring indicator. The Phantom of the Opera is a complete observation (opened January 26, 1988, closed April 16, 2023, exactly 13,981 performances, the all-time record and the anchor of the test suite); Hamilton is right-censored, still accruing since August 6, 2015. The survival shape is textbook: the median completed run closes under a thousand performances while the long tail runs to five figures, which is Kaplan-Meier, hazard curves and censoring handling in one download. Runs were extracted from article prose, where they actually live, by Broadway-anchored sentence patterns that survived a gauntlet worth reading in the scraper: West End totals masquerading as Broadway ones, Off-Broadway premieres sharing the word Broadway, prose that elides the year (Phantom opened on 26 January, year understood), openings stated in the same sentence as preview dates, and citation lines carrying almost-right numbers. The guard is theatrical physics: eight shows a week is the standard, so any count exceeding ten per stated week is a misread and keeps its dates but loses its number (12 did). Censoring is authoritative, from the live table of all 41 Broadway houses, never inferred. Openings span every decade from the 1900s; Type separates 1,307 rows into musicals and plays; 858 category members without a recoverable Broadway opening were dropped and disclosed, and articles lead with original productions, so revival-heavy titles carry their original run. Built with a 10-test acceptance suite anchored on Phantom's crown, Hamilton's censoring and Death of a Salesman's 742.
1,307 rows in one CSV file, with 12 columns in total: Show, Type, Opening_Date, Opening_Year, Closing_Date, Closing_Year, Performances, Run_Days, Still_Running, Theatre, Decade, Show_URL.
It is built by programmatically scraping en.wikipedia.org, last pulled on 2026-10-04. Datasets are versioned; older versions stay downloadable.
Yes. Every CodeSights dataset is completely free as a CSV download; a free account is all it takes. Anyone can preview the data without signing in.
Yes. The exact Python scraper that built it is viewable on the dataset page by any signed-in member, so every number is reproducible.
Automated scraping leaves room for error and the underlying sources change over time, so no version is guaranteed accurate or complete. If a number matters, verify it against the original source.