#fivethirtyeight.com

Nate Silver at Silver Bulletin:

Last Thursday night, I was working late, trying to put some of the finishing touches on our forthcoming World Cup model — and actually looking up an article I’d written for FiveThirtyEight in 2014 about my previous soccer model, SPI. Although the quality of the archive has gradually deteriorated since Disney shut down the site in 2025 (I left two years earlier in 2023), at least our text-based articles were mostly still there, or so I thought. Instead, I was auto-redirected to ABC News’s home page, which looked something like this: Sometimes weird things happen on the internet late at night, so I resisted the temptation to tweet something about it. But one of my former colleagues noticed the same thing on Friday. ABC News hasn’t made any public comment that I’m aware of — they declined to make a statement to the New York Times, which wrote about FiveThirtyEight’s disappearance. It’s possible that they have something up their sleeve, I suppose. But presumably, this was either intentional or willfully neglectful. All of the former FiveThirtyEight site from my nearly decade-long tenure at ESPN/Disney/ABC is gone.1

It’s common to read things like: “what happens on the Internet stays on the Internet”, the notion being that you can never escape your digital past. But this isn’t really true. A Pew study of a random sample of Internet links conducted in October 2023 found significant “link rot”: almost 40 percent of links that had been active 10 years earlier were broken. And that’s probably an underestimate: the study was based on the Common Crawl web archive (the same one that AI labs use to train their models), which is quite comprehensive but probably contains some bias toward more prominent sites. Another study by ahrefs found a two-thirds attrition rate for web links after 11 years.

Yes, you can still access (for now) Disney-era FiveThirtyEight content via the invaluable Internet Archive, and pre-Disney-era content from The New York Times (which I partnered with from 2010 through 2013). And obviously, we’re trying to recreate some of the most popular parts of FiveThirtyEight at Silver Bulletin. The election models and polling averages are here, and new-and-improved versions of the sports models (PELE, ELWAY, COOPER) are gradually returning too.2 Galen Druke, Clare Malone and I have even been getting the old podcast crew back together for live shows. [...]

Here are some numbers roughly in the right ballpark: during the Disney era, which lasted about 10 years, FiveThirtyEight published about 20 stories a week. Let’s say that each story took about 20 hours to produce between research, writing, graphics and editing.3 Do the math, and that works out to about 200,000 person-hours of work that ABC News just deleted. What’s probably harder to see from the outside is that none of this was ever smooth sailing: that content was produced through a lot of blood, sweat, and tears (sometimes literally on certain election nights).

Maybe the deletion of the site has me feeling wistful, but I thought I’d write about the various phases of FiveThirtyEight, including some inside-baseball stuff that I’ve never really talked about publicly. I’m just going to speak from my vantage point, not on behalf of the several dozen people who worked on the site in its various iterations. I’m not going to talk about competing editorial visions or personality clashes or any of that, though we had our fair share of newsroom drama. Instead, this is intended as more of a business school case study: a study of a large corporation, The Walt Disney Company, behaving in an incredibly neglectful way toward a smaller brand it acquired, with dozens of employees who worked exceptionally hard despite constant existential uncertainty.

A very shameful move by Disney to shut down and erase everything on FiveThirtyEight.

The data that powers this project is linked at the bottom of our poll tracker. To make the transition easier for researchers, we are providing the data in approximately the same format as 538 did. These are the known differences: • The columns poll_id, pollster_id, sponsor_ids, sponsor_candidate_id and question_id refer to the internal identifier The Times uses, rather than the 538 ID. For a map…

Throughout the campaign, it’s been noted that Clinton is more popular when she’s not a candidate. She was, generally speaking, more popular as Secretary of State, as First Lady, and New York Senator than as a presidential candidate (although some of the differences are not that large). And this discrepancy, such as it is, is often attributed to gender factors. The underlying assumption is that people like women just fine as public servants, but dislike them as ambitious candidates trying to move up the political ladder. This story isn’t as clear or neat as that – but it is a central way that people have understood Clinton’s candidacy.

Realistic count of homicides by police = about 1100 per year

…The U.S. government doesn’t track how many people are killed by the police. The FBI tracks “justifiable” police homicides, which it reports to be about 400 per year, but that tally is an undercount.

Given this vacuum, attention has recently turned to some excellent nongovernmental attempts to compile this data, including the Fatal Encounters database, the recently created Gun Violence Archive…

A Piketty problem? by The Economist