Our Scorecard · Live

We make predictions. We grade them. We publish the math.

No other news outlet does this. Every forecast we make gets tagged HIGH, MODERATE, or LOW confidence — and then we track whether it came true. Below: every resolved prediction we've made, scored honestly, including the ones we got wrong.

Loading live scorecard...

Why we publish this

Every news outlet makes predictions. Most bury the wrong ones. We don't. Every prediction we make gets a tracking ID, a confidence tag, a date, and an outcome. The full log is auditable. When we're wrong, we say so. When we're right, we don't overclaim.

If a news outlet isn't grading itself, it's grading you — into a version of the world that fits its priors. We'd rather grade ourselves.

How we tag confidence

HIGHWe have independent confirmation across two or more source ecosystems, the causal chain is clear, and we've stress-tested the call against the strongest counter-argument. Target accuracy: 70% or higher.

MODERATEDirectionally confident, but the evidence is single-stream, or the causal chain has a soft joint, or the counter-argument isn't fully dismissed. Target: 55% or higher.

LOWA scenario worth flagging because the downside of missing it is high, even if we don't think it's the most likely outcome. Target: under 50%. By design.

Three things we'll never do

1. Retroactively edit predictions. Once a forecast is published with an ID, it's locked. Wrong calls stay on the record.

2. Cherry-pick the wins. The scorecard above counts every resolved prediction, not just the ones that make us look good.

3. Hide behind "we said it could happen." If we tagged something HIGH and it didn't happen, it's a miss. Period.

How to use the scorecard

Before you act on anything in a BSFN briefing, check the confidence tag. If it's HIGH, treat it as actionable. If it's MODERATE, weight it but don't lean on it. If it's LOW, note it as a scenario — not a forecast.

The numbers above update daily. Every tagged prediction in every briefing links back to this page.