Professor Jiang Xueqin's Predictive History lectures make a lot of specific calls about wars, elections, markets and the fate of empires. This site keeps a public record of every one of them and checks each against what actually happened, so anyone can see how the forecasts are holding up.
All 180 available video transcripts from the @PredictiveHistory channel were read in full, about two million words. Captions come from YouTube's own auto-captions in the original language.
An AI model listed every forward-looking claim: anything about what will happen, with the exact quote, the moment in the video, the timeframe he gave and how confident he sounded. Re-uploads were skipped. That produced 757 calls.
Each call was then checked against news and reference sources up to Oct 16, 2026, and given one of five outcomes. Every verdict links to its sources.
Hit rate = (came true + ½ partly) ÷ resolved calls. Calls that are still open or too vague to check are left out, and so are claims that described something already underway when he said it.
It happened as predicted, within the timeframe he gave.
40 callsPart of it happened, or it happened in a different form or later than he said.
56 callsThe deadline passed without it happening, or the opposite happened.
40 callsIt can't be judged yet: the deadline is still ahead, or nothing has decided it either way.
509 callsToo vague to ever be checked, e.g. no observable outcome or no timeframe.
112 callsVerdicts are AI-assisted. They were researched and written with AI models and checked against cited sources, but mistakes are possible, especially on fast-moving stories. If you think a verdict is wrong, tell us and we'll recheck it.
Captions can mishear. Quotes come from auto-generated captions. Every call links to the exact moment in the video so you can hear the original.
"Open" isn't a miss. Most calls (509) haven't reached a verdict yet. The hit rate only counts calls that have, and it will move as more resolve.
Grouping is a judgement call. "Recurring calls" group the same idea restated across videos. Where the timing shifts between versions, each version is scored on its own terms.
The full dataset of calls, quotes, timestamps, verdicts and sources is available as predictions.json. Last updated Oct 16, 2026.
This is an independent project by Njund. It isn't affiliated with, endorsed by or connected to Professor Jiang or the Predictive History channel.