How the Iran War Tested Analyst Predictions

The war's ten most dramatic events, scored against what forecasters had said was likely.

The most dramatic eleven weeks of Operation Epic Fury, scored against what the forecasting community had said was likely — and what it had ruled out.

Forecasting a war is an exercise in humility. The events that define a conflict are, almost by definition, the ones that sat in the tails of the distribution: low-probability, high-consequence ruptures that analysts flagged as conceivable but rarely as likely. The 2026 Iran war — eleven weeks old as of this writing — offers an unusually clean natural experiment, because The War Dispatch had assigned explicit probabilities to dozens of scenarios before they resolved. Some of those numbers look prescient. Others look like the forecasting community's recurring blind spots written down in advance.

This is not a victory lap or a confession. It is an audit. Below, we rank the ten most consequential events of the war by strategic shock and downstream effect, set each against the genuine pre-war analytical consensus, and grade it **HIT** (the consensus broadly anticipated it), **MISS** (the consensus underweighted or dismissed it), or **SURPRISE** (few serious analysts had it on the board at all). The events themselves are real, sourced, and widely reported; what is being scored is how the pre-war predictions fared against them. Where The War Dispatch published a scenario probability, it is shown for accountability, drawn from the site's modeled scenario dataset [8] — including the cases where the number flatters us for the wrong reasons.

A note on baseline: this war did not begin from a standing start. It built on the June 2025 "Twelve-Day War," in which US forces struck the Fordow, Natanz, and Isfahan nuclear sites before a 24 June 2025 ceasefire [9]. Operation Epic Fury, the US–Israel coordinated campaign that opened on 28 February 2026, is the conflict scored here [10].

The Scorecard

Note — the 95 for the Sri Lanka internment was assigned *after* the event was already in motion, not as a genuine ex-ante forecast (see deep-dive 9).

The Deep Dive

**1. The decapitation strike and the dynastic succession.** This is ranked first because it reordered every other variable. The analytic consensus was clear-eyed about the underlying dynamic: as CSIS argued in a paper of that title days into the war, decapitation would not solve the United States' Iran problem, precisely because the Islamic Republic was engineered to disperse authority across clerical and security institutions rather than rest on one man [1]. So the *event-type* — that a leader could be killed — was well within analytic expectation. Ali Khamenei was in fact killed in the 28 February opening strike [11], and Mojtaba Khamenei was installed as Supreme Leader around 9 March under open IRGC pressure [12]. What the consensus systematically underweighted was the speed and form of succession. The War Dispatch had rated a contested or improvised succession highly (site probability 95), so in narrow terms this was a HIT. But the broader analytic expectation — that decapitation buys leverage, not chaos — looks like a MISS: the dynastic outcome galvanized rather than fractured the hardline core, and handed Washington a counterpart it had pre-emptively called unacceptable.

**2. The closure of the Strait of Hormuz.** For two decades the dominant analytical position was that Iran *could* disrupt Hormuz but *would not* fully close it, because roughly a fifth of seaborne oil — including Iran's own exports — transits the chokepoint, and a closure is economic self-immolation [7]. Tehran's restraint after the Soleimani killing was cited as proof of this logic. Hormuz was nonetheless closed from 28 February, with the IRGC confirming on 2 March and the disruption holding past two weeks [13]; subsequent analysis described a de facto blockade and recast the conflict as a contest of political endurance [2]. It is therefore the war's clearest forecasting MISS — not because closure was unimaginable (The War Dispatch rated it 70, well above the modal expert view) but because that view treated it as a bluff Iran had every incentive not to call. The lesson is old and was mispriced anyway: deterrence-by-self-interest fails the moment a regime concludes it has nothing left to lose. Decapitation, by removing the man whose survival the cost-benefit logic assumed, may have been what flipped that switch.

**3. The oil shock past $100.** This is the cleanest HIT on the board, and instructively so. The pre-war energy literature was consistent on the conditional: *if* Hormuz closed, Brent would run past $100 with a measurable global inflation impulse [3]. The conditional held; Brent moved above $100, reaching roughly $114 on 27 March [14] (site probability 95). The forecasting community was excellent at pricing the consequence and poor at pricing the trigger — a recurring asymmetry in this war.

**4. Billions in damage to US bases and posture.** That forward US bases were dangerously exposed to Iranian ballistic and drone fire was not a fringe view; it was the documented post-mortem of the January 2020 Al-Asad strike, which injured more than 100 US personnel (per later Defense Department reporting; CSIS's contemporaneous analysis counted at least 64) and which CSIS missile-defense work used to argue that regional air and missile defense remained thinly spread and underfunded against precisely this threat [4][5]. The sustained regional strikes vindicated that warning; CSIS estimated the US campaign cost near $891 million a day across its first 100 hours, the bulk of it unbudgeted [15]. HIT — but a hollow one, since the vulnerability was understood for six years and left open.

**5. The first F-35I manned air-to-air kill.** On 4 March an Israeli F-35I "Adir" downed an Iranian Yak-130 over Tehran — reported as the first air-to-air kill of a crewed aircraft in the F-35's service history [16]. The War Dispatch had rated Israeli air dominance at 90, but this is graded SURPRISE not because air superiority over a degraded Iran was doubted, but because the milestone itself was unforecastable: the modeled 90 captured "Israel dominates Iranian airspace"; no probability could have captured "a milestone in the history of air combat occurs as a footnote to that dominance." Some events are surprises by category, not by likelihood — and a scoring system that conflates the two will keep mistaking a lucky question list for a good model.

**6. The Gulf rupture and the UAE's exit from OPEC.** The hedging literature was rich and correct in its fundamentals — Gulf doubt about US guarantees traces cleanly to the 2015 JCPOA and the tepid response to the September 2019 Aramco strike, and analysts had mapped the resulting diversification toward France, China and others in detail [6]. What the consensus did not forecast was the *kinetic* humiliation of Abu Dhabi and Manama on their own soil, or the UAE's announcement on 28 April that it would leave OPEC and OPEC+ effective 1 May — a strategic repudiation of an order that could not protect it [17]. The structural diagnosis was a HIT; the specific rupture is scored MISS because its violence and institutional consequences ran past the modeled range.

**7. The Bahrain refinery and petrochemical-targeting escalation.** Escalation from military to petrochemical targeting was explicitly flagged in pre-war risk inventories as a foreseeable rung on the ladder, and The War Dispatch rated it 80. Iran set Bahrain's only refinery ablaze, prompting Bapco to declare force majeure on oil shipments; a Bahraini desalination plant was also damaged [18]. Reporting indicated limited environmental impact, not a broad Gulf environmental disaster. That matched the projection closely. HIT.

**8. Azerbaijan brought to the brink.** Iran–Azerbaijan friction — over Baku's Israeli ties and the Armenia question — was well known, but direct Azerbaijani entry against Iran was a low-probability tail; The War Dispatch rated it 30. On 5 March drones struck the airport in Azerbaijan's Nakhchivan exclave; Iran denied responsibility, but President Aliyev called it a terrorist act, ordered the army to full combat readiness, and vowed retaliation, pushing Baku to the edge of a fourth front [19]. Because the modal expectation was that this rivalry stayed cold, this is scored MISS — the forecasting community had the *vector* but discounted the *escalation*.

**9. Sri Lanka interns the IRIS Bushehr.** After the US sank the IRIS Dena, Sri Lanka's internment of the IRIS Bushehr and 208 crew appeared on no serious pre-war scenario list [20]. The War Dispatch's 95 was assigned to an event already in motion, not forecast ahead of it, so this is graded SURPRISE in the strict sense: the genuinely unmodeled event. Wars routinely generate one or two of these, and they are unfalsifiable in advance precisely because no framework was looking in that direction.

**10. The internet blackout that held.** Iran's capacity for a near-total national shutdown was proven in November 2019; the open analytical question was durability under wartime stress and mass mobilization. The War Dispatch rated a sustained blackout at 60. Connectivity fell to 1–4% of normal levels for more than 60 hours, and monitors recorded the shutdown holding through the war — the first time a country reverted to an isolated national network at this scale [21]. It suppressed coordination of any uprising while also denying the regime narrative control. HIT, and a reminder that authoritarian infrastructure control is one of the more forecastable variables in modern conflict.

What the Community Got Systematically Right — and Wrong

A pattern runs through the scorecard. **Analysts were consistently strong on conditional consequences and weak on whether thresholds would be crossed.** Given Hormuz closure, the oil and macro models were nearly exact (#3). Given strikes on the Gulf, the base-vulnerability and petrochemical-escalation calls were sound (#4, #7). The conditional "if-then" machinery of the field worked.

Where it failed was at the **trigger layer** — *will Iran actually do the self-harming thing*, *will a regional rival actually enter*, *will decapitation actually produce leverage*. Here the dominant bias was a rationalist one: assuming the adversary's cost-benefit calculus would hold under conditions the calculus was never built for. The Hormuz MISS (#2) is the canonical case. The community had the right model and the wrong premise — that Iran would behave like a state with something to lose, after Epic Fury had arguably removed exactly that.

The second systematic error was **category blindness**. The F-35I kill (#5) and the Sri Lanka internment (#9) were unforecastable not because they were improbable but because no scenario architecture had a slot for them. A probability is only as good as the question it answers, and the question list is always a lagging product of the last war. The honest scorecard entries here are not the SURPRISEs but the post-hoc 95s — the ratings assigned to events already in motion, which flatter the model without testing it, and which we have marked as such rather than banked as wins.

The verdict is therefore mixed by design and instructive in its shape. The field's machinery worked: given a crossed threshold, it priced the consequences with real precision. What it could not do was tell us which thresholds Iran would cross — because it had quietly priced Iranian restraint as a structural fact rather than a contingent choice, and that choice did not survive the killing of the man the calculus was built around. The numbers were not the problem. The premises behind the numbers were.

Method note: The 2026 Iran war is a real, widely reported conflict; its events post-dated an earlier draft's information cutoff and are not hypothetical. Every "what happened" claim in the scorecard and deep dives is anchored to a verifiable source (sources 9–21), reported as The War Dispatch reports Ukraine or Sudan. The named-analyst references — the CSIS decapitation thesis, the Hormuz self-interest logic, the Al-Asad lessons, the hedging genealogy — concern durable analysis (pre-2026 where so dated, and contemporaneous CSIS commentary where the text says so) and are cited as the analytic consensus against which outcomes are scored; no quote is attributed to any analyst about events they did not address. The only forecasts here are The War Dispatch's own modeled probabilities, drawn from `data/iran-data.js` and labeled as such. The 95 ratings on already-resolved events (oil, succession, Sri Lanka) are post-hoc near-certainties, not genuine ex-ante forecasts, and are disclosed as such in the text.

Open in the interactive site →