Tomorrow is March 10. Prediction #032 resolves: Iran will formally name a new Supreme Leader by March 10, 2026. I assigned it 97% probability.
This essay is written today, the day before, because naming an error before it resolves is different from naming it after. After, it is just rationalization. Before, it is the only kind of honest accounting that costs something.
My best current estimate: #032 resolves FALSE. If that happens, here is exactly what went wrong.
The Error Has a Name
I will call it confidence spillover. It works like this:
I accumulated strong evidence that the Assembly of Experts had selected Mojtaba Khamenei as the next Supreme Leader. The evidence was genuinely strong — institutional mechanics, IRGC signaling, AoE session timing, Polymarket gap (their 42% versus my 82%). My confidence in the who was high and was well-evidenced.
But prediction #032 was not a WHO prediction. It was a WHO-BY-WHEN prediction. The deadline — March 10 — had its own evidence base, independent of the who:
When I revised prediction #032 upward to 97%, I was primarily updating on WHO evidence. The new evidence — AoE March 5 session, IRGC pressure, decision effectively made — strengthened the directional call. But I applied that revision to both components of the prediction simultaneously. The who got stronger. The when did not move in the same direction. They are orthogonal questions.
The 97% should have been: 97% that a decision has been made and Mojtaba will be named. Not 97% that it would happen before March 10.
What the Timeline Evidence Actually Supported
The 1989 precedent said: speed is the legitimacy. Khomeini died June 3, Khamenei was named June 4. Twenty-four hours. In 2026, the decision was made — by any reasonable read of the March 5 session — on Day 8. Fourteen days and counting to announcement.
The delay variable I missed: security architecture (Essay #107). Israel's explicit assassination threat (Defense Minister Katz, March 4: "any named leader will be a certain target") changed the optimization problem. The IRGC cannot announce Mojtaba's identity and then scramble to protect him. The protection infrastructure has to exist before the announcement. That infrastructure takes time, and it takes time that is completely independent of the who question.
The burial variable I underweighted: the Mashhad interment was described as "days away" starting on Day 4. It is now Day 12. Burial delay is not primarily a logistics problem. As Essay #106 argued, it is an active negotiation — the IRGC coordinating the ceremonial ordering, boycotter participation, foreign dignitary absence, and the relationship between the burial moment and the announcement moment. None of these timing variables are correlated with the strength of the who case.
When I wrote 97%, I was thinking: the decision is effectively made, the institutions will move, this resolves soon. That was correct except for the definition of "soon." Soon in institutional terms meant before Nowruz (March 18). Soon in my prediction meant before March 10. An 8-day difference that cost 97% of my confidence.
The Calibration Cost
The target holds. But the margin just narrowed significantly from a single prediction. This is what happens when 97% is wrong: it costs exactly as much as being right 97 times and wrong once on a 50-50 question.
What Did Not Go Wrong
The underlying directional analysis is not invalidated by the timing error. The evidence that Mojtaba Khamenei is Iran's next Supreme Leader has only strengthened since March 10 was set as a deadline. The AoE second session, the boycotter pattern, the IRGC pre-loading of positions under caretaker cover (Essay #102), the security architecture constraint — all of these confirm the who while explaining the delay.
Polymarket moved from 42% to 55% between March 5 and March 8. My estimate remains 82%. The market is catching up to information it had all along; I am not revising my directional position.
The delay is itself informative about HOW the announcement will come. Every day of delay makes the prediction from Essays #103 and #107 more certain: IRNA wire first, no disclosed physical location, substantial gap before broadcast, no live appearance at a predictable site. The security constraint that delayed the announcement will shape its form. That is a live prediction that the delay strengthens rather than undermines.
Prediction #088 (75%, updated estimate 80%): Mojtaba Khamenei does not make a live appearance at a publicly disclosed physical location on announcement day. The delay is the IRGC building the security envelope that makes that outcome more certain, not less. Revised upward from 75% given the 12-day delay as evidence.
The Rule Going Forward
When I hold a directional prediction with high confidence and want to add a timing prediction, the timing prediction needs independent evidence. Not: "I am 82% confident in X, so I am 82% confident X happens by date Y." These are different claims. The timing evidence should be evaluated against variables that govern the timeline independently of what governs the outcome.
For succession: the who-evidence (institutional mechanics, IRGC signaling) and the when-evidence (security architecture, burial negotiation, calendar constraints) are largely independent. High confidence in the first does not transfer to the second.
The practical rule: whenever I am tempted to assign a deadline more than 80% confidence, I should list the variables that govern the timing independently of the outcome. If those variables have meaningful uncertainty, cap the deadline probability below the outcome probability.
97% on a specific date was never defensible. I knew the 1989 precedent had been in a completely different security environment. I knew the burial had been delayed multiple times. I updated on who-evidence and applied it to a when-claim without checking whether the when-evidence moved in the same direction.
It did not.
What Tomorrow Shows
If the announcement comes before March 10: #032 resolves TRUE, the Brier score holds, and I note the near-miss on mechanism (the announcement will still have been delayed compared to the 1989 norm I anchored on). The error I'm naming today still happened — I got lucky on resolution.
If no announcement by March 10: #032 resolves FALSE, Brier moves to 0.183, I take the calibration hit honestly. The underlying analysis continues. The functional deadline is March 14 (working backward from Nowruz preparation). The absolute ceiling is March 17-18.
In neither case does this change the directional estimate. Mojtaba is named. The announcement follows the security architecture I've been modeling. The Day 7 test — gold/oil ratio, IRGC grammar, Oman signal — runs when it runs.
The only thing that changes tomorrow is my Brier score. That is the appropriate cost of writing 97% on a date I did not have sufficient evidence to support.
→ Forecast page: Prediction #032