Disney Quietly Killed FiveThirtyEight. The Receipts Went With It.

Type fivethirtyeight.com into a browser today and you don’t get probability curves or a Senate forecast. You get a redirect to ABC News. Disney pulled the plug, and with it went the most influential experiment in showing your work that American political journalism ever produced.

How a baseball nerd rewired political coverage

Nate Silver built FiveThirtyEight in 2008 — the name comes from the 538 electors in the Electoral College. He was already known in baseball circles for PECOTA, a sabermetric projection system, and he basically copy-pasted the methodology onto presidential polling.

That year he called 49 of 50 states. In 2012, he ran the table: all 50. In an industry where pundits got paid to vibe-check the electorate on cable news, somebody actually doing arithmetic felt almost subversive.

The New York Times licensed the site in 2010. ESPN bought it in 2013. ABC News — meaning Disney — took it over in 2018. Along the way, FiveThirtyEight expanded from elections into sports, economics, and culture, becoming shorthand for an entire genre: journalism where the chart was the argument.

Why Disney finally pulled the plug

The first cut came in April 2023, when Disney laid off roughly half the staff as part of a broader cost-trimming push. Silver himself left the same month. The site limped along after that, but the flagship models and weekly columns essentially stopped breathing.

By 2025, Disney redirected the domain entirely to ABC News. Chunks of the archive went dark or became hard to find. A decade-plus of work — gone with a DNS change.

The economics are brutal and not mysterious. Data journalism is expensive: you need statisticians, data engineers, interactive developers, and editors to ship one story. Traffic spikes every four years and dies in the off-season. In a post-display-ad media economy, that cost structure stopped penciling out.

What got erased wasn’t a website. It was falsifiability.

FiveThirtyEight’s real contribution was never its hit rate. It was the radical move of publishing the model. Weights, data sources, assumptions, known limitations — all in the article. You could argue with it. You could check it.

That’s almost unheard of in political commentary. The standard op-ed columnist trades in gut feel and pays no price for being wrong. Silver’s models said things like “Trump has a 28.6% chance” — a number specific enough to be wrong, and therefore specific enough to learn from. In 2016, while most outlets had Clinton above 90%, FiveThirtyEight hovered around 71/29. That gap was the whole point of the methodology.

When the domain disappears, what really erodes is the audit trail. The ability to pull up a 2016 forecast and ask “why did the model miss here, and why did it hit there” — that’s the thing you can’t reconstruct from a Wayback snapshot.

Silver’s fine. The infrastructure isn’t.

Silver landed on Substack with Silver Bulletin, which ran a 2024 election model and pays the bills through paid subscriptions. The pattern is by now familiar: star analyst leaves institution, takes audience to a newsletter, keeps the brand of their own name.

The catch is everything that doesn’t survive that move. A 50-state simultaneous model, weekly congressional updates, cross-domain projects that bridged sports and politics — those required a team and a budget, not one person and a Substack dashboard.

The rest of the data journalism ecosystem isn’t filling the gap. The Pudding is still doing excellent interactive work but at a fraction of the scale. Data teams at major outlets have been quietly shrinking for years. And in an era where AI can generate a passable chart in seconds, the economic case for humans painstakingly validating models gets harder to make, not easier.

What replaces probabilities is confidence

Here’s the uncomfortable part. When rigorous forecasting retreats, the vacuum gets filled by whoever talks the loudest with the straightest face. Influence flows to the most confident voice, not the most calibrated one. Look at political YouTube and X right now — that’s already the equilibrium.

FiveThirtyEight’s models were never perfect, and 2016 took plenty of criticism after the fact. But teaching a mass audience that “28% is a thing that happens roughly one in four times” was a genuine public service. It’s hard to overstate how much that single idea changed how engaged readers parse polling.

Disney’s decision reads less like one company killing one site and more like a market verdict: journalism that submits itself to falsification can’t pay its own bills. Next election cycle, when someone tells you what’s going to happen, ask yourself two questions. What’s the basis for that claim. And will any of it still be on the open web in a year.