Jiang Xueqin Wrong Predictions: Every Miss We've Tracked
As of 2026-09-25, six of the 400 predictions in this tracker are scored wrong. That is the whole answer to the headline question, and it is deliberately unglamorous: 18 calls are confirmed, six are wrong, 24 have resolved at all, and 303 remain pending. Headline accuracy sits at 75% — a figure resting on a small denominator that will move as the backlog clears.
The six misses are not buried. They are listed individually, with their IDs, source lectures, and expired timelines, on the wrong-status page. What follows is how a call earns that label, why the misses matter more to credibility than the 18 hits, and why nobody should treat 75% as settled.
What "wrong" means in this tracker
Every prediction enters the system with three things attached: the claim, the timeline Jiang gave it, and the lecture it came from. Scoring happens only when the timeline runs out.
- Confirmed — the predicted event occurred within the stated window.
- Wrong — the window closed without the event, or later reporting contradicted the claim.
- Not yet / unverifiable — the window is still open, or the claim is too vague or too private to adjudicate.
Nothing is scored wrong because it looks unlikely. Nothing is scored confirmed because it looks prescient. A call with a five-year window and a call with a two-week window get identical treatment: wait for the clock.
That rule matters here. The 303 pending entries are not a graveyard of quiet failures; they are mostly predictions whose windows simply have not closed. Some are dated 2027, 2028, and beyond.
Why the misses drive trust, not damage it
Publishing misses is the cheapest available proof that the confirmed column means anything. A tracker that only surfaces hits is a highlight reel, and highlight reels are worthless as evidence. Anyone searching Jiang Xueqin failed predictions elsewhere should ask a simpler question first: does that source list the calls that did not work?
The counter-intuitive part is that six wrong entries raise the value of everything else. A hit rate computed over a set that includes public, dated, falsifiable misses is worth more than a perfect record computed over a curated set — because the second number is not measuring anything.
The hits that give the misses context
The confirmed column is where the asymmetry shows up. Jiang called the 2024 presidential outcome (P001) and JD Vance's place on the ticket (P002b) months ahead. He called a US war with Iran (P011), the bombing of Iranian nuclear sites (P012), Iran's closure of the Strait of Hormuz (P026), and Iranian strikes on Gulf states (P027) — all now marked confirmed. So are a US attack on Venezuela (P082) and Trump-era trade wars against Canada, Mexico, and the European Union (P325).
Several confirmed calls are unusually specific: a decapitation strike killing Ayatollah Khamenei (P088), the destruction of the Iranian navy (P105), Mojtaba Khamenei's succession (P041), and the judgment that nuclear weapons would not be used in the Iran war (P016). Others are structural: the US-Iran war lasting weeks or years (P018), escalation against civilian infrastructure (P020), China staying out of the conflict (P071), Putin visiting China after Trump's departure (P315), and AI companions arriving as a growth strategy (P129).
A forecaster who lands a Strait of Hormuz closure and still misses six calls is more useful to read than one with no misses on record. The misses tell you where the model bends.
The denominator problem
The caveat that matters most is statistical rather than editorial. Only 24 of 400 predictions have resolved. That leaves 303 pending and a 75% headline figure built on a sample small enough that a handful of resolutions could shift it by double digits.
A rough 95% confidence interval on 18 of 24 runs from about 58% to 92%. Treat any single figure — ours included — as a snapshot, not a verdict. The tracker draws on 180 source lectures, 179 of them transcribed: a large input pool and a still-small output pool.
There is also an asymmetry in how the backlog clears. Long-dated predictions take years to score, and long-dated predictions are the hardest kind to get right. If the pending 303 break against Jiang at a higher rate than the first 24 did, the headline number will fall — and this page will say so when it regenerates.
What this page looks like next time
Every three weeks, the numbers above get rewritten. If the wrong count rises, the list at the wrong-status filter grows and this post reflects it. That is the entire editorial policy: log the miss, keep the ID, keep the source lecture, and let readers see both columns.
The honest read on predictive history misses is not that they exist. It is that they are countable, dated, and published directly beside the wins.
Browse all tracked predictions · Accuracy analysis · All blog posts