Tag: statistics

Posted on by

Ancient Web: Trenches on the Web Put World War I Casualty Figures Into a Table People Kept Citing

A good historical table can travel farther than the website that made it. The Trenches on the Web World War I casualty page is a case in point: a compact country-by-country table that has been cited and copied for years.

View the casualty figures page.

The page organizes wartime numbers into categories such as troops mobilized, deaths, wounded, and missing or prisoners. The Russian figures, for example, have often been reproduced from the table with roughly 12 million mobilized, 1.7 million dead, 4.95 million wounded and 2.5 million missing or captured.

The attraction is obvious. World War I statistics are enormous, and a table lets a reader compare countries without digging through a shelf of demographic studies.

The danger is equally obvious: casualty numbers are not perfectly settled facts.

Different sources use different definitions. Some count only military deaths; others include deaths from wounds or disease differently. Missing soldiers may later be reclassified. Prisoners overlap with categories in some datasets. Borders and successor states complicate national totals. Civilian losses are a separate nightmare.

That does not make an old reference table useless. It tells you how to use it. Treat the page as a fast comparative index and a clue to the figures circulating in historical literature, then check important numbers against modern scholarship and primary statistical sources.

The page’s long afterlife is itself worth noticing. Teachers, forum discussions, research papers and other history sites have linked back to it because simple, readable reference pages were one of the old web’s greatest strengths. Somebody did the clerical work once, and thousands of later readers benefited.

Today, the same query often produces automated snippets with no visible methodology. The old table is at least inspectable as a whole. You can see which columns exist and where ambiguity might enter.

Open the Trenches on the Web casualty table with the right mindset: useful reference, not sacred scripture. History gets more accurate when the table is the beginning of the question rather than the end.

Posted on by

Ancient Web: ICYouSee Was Online in 1994 and Turned the Titanic Into Web Data

ICYouSee belongs to the generation of websites that existed before most people had decided what a website was supposed to look like. The site, created and maintained by John R. Henderson, went online in December 1994. One of its most durable projects used the Titanic passenger list as something the Web could organize, compare and explore.

Source: ICYouSee

The Titanic material first appeared in 1998 and broke passenger information into demographic tables: nationality, travel class, survival, deaths and lifeboat occupancy. Instead of treating the disaster only as a narrative, the site made the passenger list legible as data.

That sounds ordinary now. In 1998, it was a much more interesting use of a browser.

Before the Titanic Dataset Became a Machine-Learning Cliché

Today, a version of Titanic passenger data is famous because it became a standard beginner dataset for statistics and machine learning. Students routinely predict survival with decision trees or logistic regression before they know much about the historical source.

ICYouSee represents an earlier stage of that transformation. The purpose was not a Kaggle exercise. The tables were meant to help readers inspect the people aboard the ship and compare who survived.

Later statistical literature discussing the many versions of the Titanic dataset specifically points back to Henderson’s ICYouSee page and dates the site’s launch to 1994 and the Titanic page to June 1998.

HTML 3.2, Background Images and a Cat GIF

Surviving descriptions of the site also preserve its technical character: HTML 3.2-era markup, a background image, old font and align attributes, a small cat graphic and a navigation structure built before responsive design became an expectation.

That is not merely visual nostalgia. It tells us what a serious educational site looked like when the Web’s conventions were still unsettled.

ICYouSee mixed historical material, personal pages, finding tools and educational resources under one independent domain. Its Titanic work is especially valuable because it sits at the intersection of Internet history and data history: a historical passenger list becoming a browsable dataset years before “data science” became a mainstream label.

The live site is now unreliable, which makes the surviving references to its structure and publication dates more important, not less.

Original site: http://icyousee.org/

Posted on by

Ancient Web: Evan Miller Made Statistics Readable for Programmers

Some personal websites become useful enough that people stop thinking of them as personal websites.

Visit Evan Miller’s site

EvanMiller.org is one of those places.

The homepage is a dense index of technical writing, software, mathematical notes, ranking methods, programming experiments, and statistical tools. It includes material on A/B testing, confidence intervals, rating systems, probability, programming languages, Nginx, Erlang, data formats, and open-source software.

One article, Statistical Formulas for Programmers, collected classical statistical methods into a compact working reference aimed directly at developers. Another series explains ranking systems for stars, votes, hotness, and news items. The site also hosts sample-size calculators and other tools for experiment design.

A technical notebook that escaped its owner

The interesting historical pattern here is how a personal site can become infrastructure without ever turning into a publication company.

Miller’s writing circulated because programmers linked directly to individual pages. Search results, Hacker News discussions, blog posts, documentation, and forum answers sent readers into a site that still looked structurally like one person’s corner of the Web.

That architecture is important.

A useful article stays where it was published. Years later, somebody can still link to the same explanation instead of pointing at a screenshot of a thread or a deleted social post.

Miller’s homepage even keeps a project graveyard alongside software that still works. That is exactly the kind of honest chronology personal sites are good at. Dead projects do not need to vanish merely because they are no longer maintained.

The site has continued evolving rather than being frozen as a museum piece. Newer work sits beside old Nginx notes, old Erlang references, statistics articles, ranking formulas, and programming-language reviews. The result is a visible technical career expressed as hyperlinks.

CacheRat’s 1,967 Ancient Web Domains research list includes many pages that demonstrate the same principle: the Web works best when useful knowledge has a stable address.

EvanMiller.org is what happens when that address remains useful for a very long time.

Browse Evan Miller’s technical archive

Posted on by

Ancient Web: Mcubed Turned Sports History Into a Hand-Built Statistics Machine

Mcubed.net looks less like a modern sports site than a very determined spreadsheet escaped into HTML.

Visit Mcubed.net

That is a compliment.

The site organizes historical results across the NFL, MLB, NBA, NHL, college football, college basketball, golf, tennis, the Olympics, NASCAR, soccer, and bowling. Individual sections break those sports into franchise records, streaks, standings, championship history, series records, tournament results, rankings, and best-or-worst seasons.

There are no giant video carousels blocking the numbers.

The numbers are the point.

The Web as a queryable notebook

One of the underappreciated uses of the personal Web was turning somebody’s private obsession into public infrastructure.

Mcubed does exactly that.

Want to compare college football teams across eras? Track a franchise’s winning streaks? Look up NCAA championships by school? Browse World Cup history? Find the most frequent playoff matchups?

The site has a page for it.

That structure feels old-fashioned now because modern sports publishing tends to bury historical data underneath news, betting products, video, and account systems. Mcubed instead behaves like a reference desk run by one person who kept adding drawers.

The historical value is not merely the statistics themselves. Much of the same data exists elsewhere.

The interesting part is the independent organization of it.

The categories reveal what a sports-history obsessive considered useful enough to preserve and compare. The pages are stable, linkable, and largely free of the churn that makes commercial sports sites difficult to use as long-term references.

CacheRat’s 1,967 Ancient Web Domains research list contains a surprising number of these one-person databases.

They are reminders that the Web was once full of people who solved a problem for themselves and then simply left the solution online for everybody else.

Mcubed’s slogan is that it takes sports histories to the third power.

The site has apparently taken that literally.

Dig through the sports-history tables at Mcubed