Time for US military to exit from Iran, after 1,000 troops are deployed – By week 4

The starting probabilities of this template were calculated by asking a narrow outside-view question: if the United States has already crossed the threshold of deploying at least 1,000 troops on the ground in Iran, how long have roughly comparable U.S. ground involvements historically taken to wind down?

This is a relative-time market surface, not a live calendar-date forecast. The current CSV uses a dummy anchor date only so the market can fit Antistatic's :date schema. The meaningful quantities are the elapsed-time labels: rows are weekly thresholds after deployment, while the projection_group buckets are month counts after deployment.

Once a real ≥1,000-troop deployment happens, the market should be restaged by shifting every threshold and end date so that week 1 equals deployment day + 7 days, week 2 equals deployment day + 14 days, and so on. At that point the title should also be de-conditionalized.

Reference Classes and Partial Pooling

The model uses two overlapping duration layers. The first captures limited-duration interventions such as Lebanon 1958, Haiti 1994, or other coercive footholds that end in weeks or months. The second captures major campaigns and occupations such as Korea, Vietnam, Iraq 2003, and Afghanistan, where large ground deployments can persist for years. Some cases sit between these layers. Grenada, Panama, and the Desert Storm ground offensive are counted as overlap cases rather than forcing them into only one bucket.

Overlap handling is explicit: each case gets total weight 1.0. If a case belongs to both layers, its weight is split evenly across them. That prevents short decisive invasions such as Grenada or Panama from pulling either layer too hard on their own.

The final template uses a 60% limited-duration / 40% major-campaign mixture. That is a relevance-weighted outside-view judgment rather than a raw sample frequency. Equal-weight historical counts would lean too heavily on the handful of twentieth-century occupation wars, while the actual threshold in this question is a ≥1,000-troop deployment. That threshold still leaves a real major-war tail, but it is more compatible with limited coercive occupations than with a full regime-change campaign.

Fitted Duration Distributions

Durations are modeled on log(days after deployment). Each layer borrows a small amount of information from the other to avoid overfitting: the limited-duration layer borrows 10% of its log-mean and log-sigma from the major-campaign layer, while the major-campaign layer borrows 5% from the limited-duration layer. Sigma floors are kept wide so the final curve preserves meaningful uncertainty.

Selected Conditional Probabilities

These are the template's cumulative probabilities for the involvement having ended by the stated elapsed time after deployment.

Month Bucket Summary

The projection-group buckets correspond to months after deployment. The probability shown for each bucket below is the probability by the last weekly threshold inside that bucket.

Most Influential Examples

Operational Interpretation

The curve is deliberately not a simple 'short war' forecast. The limited-duration layer keeps real mass on a fast-end outcome such as a Kharg-style seizure, a narrow strait operation, or another politically constrained coercive landing that ends in weeks or a few months. But the major-campaign layer preserves the opposite lesson of U.S. history: once a large ground deployment is politically and militarily committed, getting back out can take much longer than decision-makers expect.

Generation Script

The Python script used to produce this template is available for download: generate_market.py