Will the US Deploy Ground Forces to Iran?

A binary probabilistic assessment, built from a four-iteration methodology, of whether American soldiers will set foot inside Iran during Operation Epic Fury. Three independent evidence lines — the expert accuracy audit, first-principles extraction from political statements, and Bayesian-discounted prediction markets — converge inside a ten-point band centred at roughly 19%.

Fifteen days into Operation Epic Fury — the most consequential military engagement in the Middle East since 2003 — one question is generating more internal Washington anxiety than almost any other: will American soldiers set foot inside Iran? The short answer, based on a systematic analysis of every direct public statement by officials, analysts, and prediction markets, is roughly one-in-five — about 19%. That is not small. It is roughly the same odds as rolling a one or two on a six-sided die, every single day, for as long as the conflict continues.

People are going to have to go and get it. Secretary of State Marco Rubio — to senators about physically securing Iran's nuclear material, classified briefing, March 2026

The risk is not what you think it is

When people hear "boots on the ground in Iran," they picture columns of Abrams tanks rolling through Persian mountain passes toward Tehran — an Iraq 2003 repeat. That scenario is essentially off the table. Iran has 86 million people, terrain that would make Afghanistan look navigable, no adjacent US military staging ground, and no Congressional authorization. The probability of that version of ground war is around 4%. Analysts across the political spectrum agree on this, and even Netanyahu himself said regime change "requires the Iranian people, not airstrikes" — admitting external military force alone cannot achieve it.

The live risk is something far more targeted: a small team of special operations forces, operating for days rather than years, sent to physically seize Iran's enriched uranium stockpile — or to take control of one of the small Iranian islands flanking the Strait of Hormuz and use it to neutralise the missile and drone launchers that have shut down the world's most important oil chokepoint. Bryan Clark, a former US Navy submariner who now directs the Hudson Institute's Center for Defense Concepts and Technology, described exactly this scenario in The Hill on March 14:

On the Iranian side there are some islands that are just offshore and you could land troops ashore there, take control of that territory and then essentially use them to engage any threats that emerge on the shoreline. Bryan Clark · Hudson Institute · The Hill · March 14, 2026

He called it operationally feasible. He is one of the most accurate military analysts in the public record.

We seem to be on a path toward deploying American troops on the ground in Iran. Sen. Richard Blumenthal (D-CT) · Senate Armed Services Committee, after a classified Iran war briefing · March 10–11, 2026

What the evidence actually says

This report is built from three independent lines of evidence that, strikingly, all land in the same zone. First, the expert analysts: the two with scoreable track records — Clark (accuracy score 0.72 / 1.00) and Brent Eastwood, a former Army infantry officer who explicitly called the SOF scenario "low-probability" (accuracy score 0.52 / 1.00) — together imply around 14–15%. Second, the politicians: applying a rigorous filter for what their statements reveal against their own interests — rather than what they want us to hear — points to 20–22%. Third, prediction markets after applying a discount for the insider trading that has plagued Polymarket's Iran contracts: 20–25%. All three arrive at roughly the same neighbourhood.

The single most important data point is Rubio's statement to senators about needing to "physically go and get" Iran's nuclear material. A Secretary of State does not say this in a classified congressional briefing as a rhetorical flourish. The second most important is David Sacks — Trump's own AI czar, one of his closest public allies — going on a podcast on March 13 to beg for an exit from the war, explicitly naming a "faction pushing Trump toward ground troops." Advisers who privately disagree don't go public against their principal unless private lobbying has already failed. That behaviour is the evidence.

There is a faction, largely but not exclusively in the Republican Party, pushing Trump toward ground troops and further escalation. David Sacks · Trump's AI & Crypto Czar · All-In Podcast · March 13, 2026 — a White House insider publicly naming the risk he is trying to prevent

What to watch

The next 72 to 144 hours are the highest-risk window of the conflict so far. Three things matter most. First: does Trump authorise the SF uranium-seizure or Hormuz island option? The NIC warned against it and the Joint Chiefs pushed back — but Trump has already launched a war without Congressional authorisation, directly threatened the new Supreme Leader, and struck Kharg Island. The institutional brakes exist; they have not been determinative. Second: does Hormuz stay closed long enough to push retail gasoline above $4.00 per gallon? At roughly $0.05–0.07 per day, that is about six days from now, and it is the number Trump's political team watches most closely. Third: does Turkey's foreign minister or China's Wang Yi produce a diplomatic framework? If they do, the pressure driving the SOF option attenuates. If they don't, the hawks gain leverage.

One number to carry forward: ~19% over the active-conflict horizon, 95% confidence interval [9%, 33%]. Wide interval, honest evidence base. The thing that should keep decision-makers up at night is not the point estimate — it's the fact that three independent, methodologically unrelated approaches all find themselves in the same 20-point zone.

Executive summary

The probability that the United States deploys active military personnel to Iranian territory — in any form, including special operations forces, coastal raids, or island seizure — is approximately 19%, with a 95% credible interval of [9%, 33%]. The probability of a full conventional invasion resembling Iraq 2003 is approximately 4%, credible interval [1%, 9%]. These are the most defensible point estimates available given the evidence base as of March 15, 2026.

Top-line findings

1. The operative risk is the SOF pathway, not a conventional invasion. A conventional ground invasion — nation of 86 million, mountainous terrain, no adjacent basing, no AUMF — is structurally precluded. The 4% is residual noise. The live risk is a limited SOF insertion: uranium seizure, Hormuz island seizure, or coastal missile strikes. These are categorically different scenarios with different feasibility profiles.
2. The most important single piece of evidence is Rubio's congressional statement. Secretary of State Rubio told the Senate Armed Services Committee in a classified briefing that securing Iran's nuclear material requires physical seizure: "People are going to have to go and get it." It is the clearest forward-looking operational statement by any senior official about a ground-presence mission.
3. The Blumenthal signal carries classified-briefing weight. Blumenthal emerged from a classified Iran war briefing and stated he was "more fearful than ever that we may be putting boots on the ground" and that the US "seems to be on a path" toward it. Even with a 35% political-incentive discount applied, this implies ~18–22% from someone who just saw the classified version of what we can only infer publicly.
4. Sacks's public intervention is a revealed-preference signal. Trump's AI czar went public to oppose ground troops, explicitly naming the internal faction pushing for them. Advisers who privately disagree do not go public against their principal unless private lobbying has already failed. The cost of going public is the evidence.
5. The expert accuracy audit dramatically reduces the usable panel. Of four analysts making direct ground-troops statements, only two have scoreable track records: Bryan Clark (E[p] = 0.72, 5 scored predictions) and Brent Eastwood (E[p] = 0.52, 4 predictions). Clark describes operational feasibility; Eastwood explicitly says "low-probability" (~12%). Expert tier: ~14–15%.
6. Three independent evidence lines converge at 19–20%. Expert tier (~14–15%), political-statement extraction (~20–22%), and prediction markets after Bayesian discount (~20–25%) are methodologically independent and converge within a 10-point band.

Question definition and scope

This assessment addresses a binary question: will active US military personnel physically enter the territory of Iran during the active hostilities of Operation Epic Fury? The operative definition — consistent with the Polymarket resolution criteria — requires active US military personnel intentionally deployed to ground positions inside Iran's internationally recognised borders. Intelligence operatives, contractors, airspace transits, and proxy forces (including PJAK Kurdish groups operating with US air support in western Iran) do not qualify.

The question decomposes into two sub-scenarios with different probability signatures.

Expert accuracy audit

The methodology requires: (1) identify only analysts who made direct, on-the-record statements about US ground forces in Iran; (2) locate and score historical predictions in the public record; (3) derive E[p] from Jeffreys Beta posteriors; (4) apply domain-shift discounts where the analyst's domain differs from the question. All 15 experts from the main panel's Appendix A are excluded — none made statements between March 1–15, 2026 directly addressing this specific question.

Bryan Clark — Hudson Institute, Center for Defense Concepts and Technology

Directness: very high. Clark is describing the operational "how," not speculating about whether. Domain: naval operations / expeditionary forces — directly applicable to this question.

Implied P(any boots) from Clark's statement: ~20% — his statement establishes operational feasibility, not a probability estimate. Treated as upper-range anchor for the expert tier.

Dr. Brent Eastwood — 19FortyFive / former US Army infantry officer

Directness: very high. The only analyst in the record to use explicit probability language ("low-probability"). Domain: military operations — directly applicable.

This would be a low-probability scenario, but it is one option for the Americans. The Pentagon could order special operations forces to patrol the coastline and blow up any missile or drone launchers. Brent Eastwood · 19FortyFive · March 12–13, 2026

"Low-probability" in military planning calibrates to 5–20%; midpoint = ~12%. Implied P(any boots): ~12%.

Additional domain voices

Political statements — first-principles extraction

Political statements are not accuracy-weighted. They are evaluated through a filter: what does each statement reveal independent of the speaker's interests? Statements that say only what the speaker would want to say carry minimal evidential weight. Statements that reveal something against the speaker's interest — or neutral to it — carry meaningful weight.

President Trump — conditional risk, open door

I don't have the yips with respect to boots on the ground — I say "probably don't need them," or "if they were necessary." … We haven't made any decision on that — nowhere near. President Trump · NY Post, March 3; The Hill, March 9–10, 2026

Trump explicitly refused to say what every other president has said — costless to say and very common. Not saying it is deliberate. The "nowhere near" six days later is a current-state constraint, not permanent foreclosure; the "if necessary" conditional from March 3 remains structurally intact. Active political management of this specific option over six days. Implied P(any boots): ~22%.

Secretary of State Marco Rubio — strongest single statement

The US needed to physically secure nuclear material in Iran. People are going to have to go and get it. Sec. Marco Rubio · Senate Armed Services Committee classified briefing · cited Al Jazeera, March 11, 2026

A Secretary of State does not make this statement in a classified congressional briefing as rhetoric. "People are going to have to go and get it" is a forward-looking operational statement made on the record. The deliberate vagueness of "people" — before senators on the Armed Services Committee — signals authorisation was not concluded, not that the option was hypothetical. Most operationally specific statement by any senior official about a ground-presence mission. Implied P(any boots): ~24%.

Secretary of Defense Pete Hegseth — explicit formal option reservation

We would be completely unwise if we did not reserve the right to take any particular option, whether it included boots on the ground or no boots on the ground. Sec. Pete Hegseth · Bloomberg / CBS / Pentagon briefing · March 2026

Hegseth could have maintained deterrence by saying "no plans" — strategic ambiguity works without naming the option explicitly. Instead he named boots on the ground and endorsed reserving it. "Completely unwise not to reserve" is a doctrine statement: the option is formally in the repertoire, not hypothetical. The open-ended "as far as we need to" links directly to Rubio's nuclear objective. Implied P(any boots): ~22%.

WH Press Sec. Karoline Leavitt — temporal constraint

Ground operations are not part of the plan right now. Karoline Leavitt · White House press briefing · cited Al Jazeera, March 11, 2026

"Not part of the plan right now" contains a temporal qualifier that is deliberate. If the answer were simply no, she would have said no. "Right now" rules out Days 1–15, not the conflict as a whole — bureaucratically precise language indicating the option was not closed, only not yet authorised for the current phase. Implied P(any boots): ~8%.

Sen. Richard Blumenthal — highest-information source

I am more fearful than ever, after this briefing, that we may be putting boots on the ground. … We seem to be on a path toward deploying American troops on the ground in Iran. Sen. Richard Blumenthal · Democracy Now, March 4; Al Jazeera, March 11, 2026

Blumenthal just saw the classified version of what we can only infer publicly. The escalation from "fearful" (March 4) to "on a path toward" (March 10–11) tracks with classified intelligence over six days. A senator on Armed Services does not make this specific, forward-tense, directional statement after a classified briefing as pure theatre without some classified basis. Applying a 35% political-incentive discount: implied P(any boots): ~19%.

David Sacks — against-interest internal signal

There is a faction, largely but not exclusively in the Republican Party, pushing Trump toward ground troops and further escalation. We've degraded Iranian capabilities massively. This is a good time to declare victory and get out. David Sacks · Trump AI & Crypto Czar · All-In Podcast · March 13, 2026

Naming the faction publicly is the against-interest signal. Sacks is risking his relationship with Trump to fight an outcome he believes is live. If the ground-troops risk were 3–5%, this political cost would be unjustified. The act of going public — naming the option, naming the faction, absorbing reputational risk — is the evidence that the faction is real, powerful, and actively pushing. Private lobbying has already failed. Implied P(any boots): ~22%.

Bloomberg — SF uranium-seizure option formally raised

Bloomberg reported on March 8–9 that the SF uranium-seizure option was "floated by Trump" and "not yet authorized." The NIC issued a classified warning against it. CJCS filed formal pushback. Execution would require a SECDEF order.

Options don't trigger NIC classified warnings and formal CJCS pushback unless they reach the level of presidential consideration. "Not yet authorized" means the authorisation question was asked. The institutional opposition being activated is the evidence the option was real. Trump has already overridden institutional caution in this conflict (no AUMF; direct Mojtaba threat). The NIC + CJCS brake is real but not determinative. Implied P(any boots): ~21%.

Prediction market analysis

Two prediction markets address the ground troops question directly. Both require Bayesian discount application consistent with the main Day 15 report — Polymarket systematically overprices Iran-related scenarios by 1.5–2.5× (ceasefire market: Poly 41% → Bayes 10–14%). For this question, the expert base is thinner, so a smaller discount (1.5×) is applied.

For the active conflict horizon (60–120 days from Day 15), the discounted market range is approximately 20–27%. Prediction markets aggregate: ~20–25%.

Three-tier weighted model

The three tiers measure different things, draw on different sources, and produce estimates through different mechanisms. Combining them is not averaging — it is a weighted synthesis that assigns greater weight to tiers more proximate to the actual decision and with higher-information sources.

Evidence map — signals contributing to the final estimate

Key uncertainties ranked by impact

Strategic conclusions

The SOF pathway is the live risk; the conventional invasion is not. Every structural constraint cited in public commentary — terrain, logistics, public opinion, AUMF absence, KC-135 attrition — applies primarily to a conventional invasion, not to a 72-hour SOF mission to a specific facility. Most public commentary fails to make the distinction. The risk is concentrated in the narrower, more operationally credible SOF scenario.

Three independent evidence lines converge at 19–20%. The expert tier (~14–15%), the political statement tier (~20–22%), and prediction markets (~20–25% discounted) are methodologically independent and converge within a 10-point band centred at 19–20%. Convergence from different methodologies increases confidence in the central estimate.

The CI width is honest, not a model failure. The 95% CI of [9%, 33%] reflects a genuinely thin direct-evidence base — only 9 scored predictions across 2 accuracy-weighted analysts, versus 113 in the main panel. The CI communicates what the evidence actually supports. Narrower CIs from prior versions were false precision.

The next 72–144 hours are the maximum risk window. Leavitt's "not part of the plan right now" is a temporal constraint, not a permanent foreclosure. The convergence of the $4.00 gasoline threshold (~6 days away), the Kharg ultimatum deadline, the UAE port threat resolution window, and the Trump–Xi summit preparation (16 days) creates a concentrated decision-pressure window. Watch CENTCOM force movements and Trump Truth Social language on nuclear sites specifically.

Diplomatic off-ramp and ground-troops option are in direct competition. If the Fidan or Wang Yi channel produces a diplomatic framework before Day 30, the political pressure driving the SOF option attenuates materially. If Hormuz stays closed and gasoline hits $5+, the hawks gain decisive leverage. These two trajectories are in direct competition over the next 6–10 days. The outcome of that competition is the single most important variable in the ground-troops probability.

Methodology evolution — four versions