Data Changelog — Every Correction, Dated

The real interest rate series was not read from any data source. It was produced by a random number generator on every cold start while the site presented it as live data. Because the value was cached for an hour, it looked stable within a session but the entire three-year history changed whenever the process restarted. It also fed the growth probability figure on the dashboard, which could move by itself with no market change behind it.

The indicator now reads the Federal Reserve series DFII10 — the 10-year TIPS constant maturity yield. The corrected reading on the day of the fix was 2.43%, against roughly 0.5% shown by the generated series.

Yahoo Finance pads its calendar with days that carry no closing price. Those days were being recorded as a price of zero rather than skipped, which dropped the Buffett Indicator series vertically to zero — including on the most recent days, so the headline value read 0% and the status badge read POSITIVE. Ten such days appeared in the three-year window.

Days without a close are skipped. The corrected reading was 234%, with a NEGATIVE status. The site-wide crash risk moved from 1/3 (MODERATE) to 2/3 (HIGH) as a result. The corrected value matches the 234.3% GuruFocus published for the same date.

Axis ticks were generated in 0.5 steps but rendered without decimals, so an axis running 1.5 / 2.0 / 2.5 / 3.0 displayed as 2 / 2 / 3 / 3. Gridlines were not where their labels claimed.

Non-integer ticks keep one decimal place.

When the Federal Reserve API key was missing or the API returned an error, the site silently substituted a generated sample series and logged only a warning. Readers had no way to tell a real reading from a placeholder.

Data source failures now surface as errors instead of being papered over with generated values.