From Analysis to Probability: Modernizing Intelligence Forecasts

Everyone wants to predict the future. While not foolproof, the age of artificial intelligence and global hyperconnectivity has brought us closer to this aspiration. Imagine a source of information that covers events from politics, economics, and beyond, capable of predicting such events with astonishing accuracy, surpassing current polls and expert analysis. A prediction market, which for the purposes of this article shall be defined as a platform that allows for binary event contracts to be bought and sold that resolve based on real-world occurrences, with contract prices corresponding to event likelihood, has proven to have resounding success in making accurate predictions hours, if not days, before events actually occur.

Leading prediction markets, like Polymarket, advertise an average of 91.1% accuracy an entire month prior to contract resolution, and an average of 95.2% accuracy four hours prior to contract resolution. These statistics suggest that information on Polymarket can repeatedly and consistently provide highly probable odds of event outcomes up to a month in advance. For the purpose of this paper, accuracy will be measured in accordance with U.S. Intelligence Community Directive 203 (ICD 203).

The ability for prediction markets to create highly accurate forecasts across a spectrum of events has been proven time and time again, and so the question arises: why are these markets not more utilized by the intelligence community and government in decision-making and analysis?

The U.S. intelligence community attempted to do this once. In May 2001, the Defense Advanced Research Projects Agency opened the Policy Analysis Market as a part of the Future Markets Applied to Prediction project. Despite the success of the Policy Analysis Market, DARPA shut the project down in 2003. This was largely attributed to congressional criticism, with accusations that it was an “unbelievably stupid system,” a “tragic waste of taxpayer money,” and a “harebrained scheme.” Congress’ main qualm was a moral conundrum: accusations that it allowed people to profit off predicting terrorism.

Yet despite these concerns, research published in the Central Intelligence Agency’s academic journal Studies in Intelligence examined many different case studies and scenarios in which the Policy Analysis Market could have been incredibly useful to the intelligence community. While Intelligence Advanced Research Projects Activity (IARPA) later created the Intelligence Community Prediction Market in 2009, it too was later cancelled, meeting its end in 2020.

In this instance as well, research showed the predictions and odds found on the Intelligence Community Prediction Market were often more accurate than traditional intelligence reports. Influential think tanks, such as the Center for Strategic and International Studies (CSIS), highlighted the value of collecting forecasts from across the more than eighteen agencies that comprise the U.S. intelligence community, identifying points of disagreement among analysts, and mitigating errors by sourcing information from a large and diverse group of thinkers. The CSIS suggested that all future reports from the National Intelligence Council should include forecasts from such markets.

As intelligence agencies increasingly utilize open-source intelligence due to its abundance and value, they must include in their strategies the utilization of prediction markets. It would be naïve to rely only upon prediction markets, and they should not be the basis of any major decision-making. At the same time, they should be valued just as much as any other information collected through open-source intelligence.

Let us consider a hypothetical use case for a prediction market by the intelligence community. Two nations of strategic geopolitical importance have been locked in a state of war for decades. While press reports cite hesitation from both sides, traders on Polymarket project that a peace deal will go through with roughly seventy percent odds. When utilized in conjunction with other intelligence, such odds can help paint a much clearer picture of what will happen. This scenario is exactly what happened when the United States brokered peace between Armenia and Azerbaijan in August of 2025.

In another scenario, the intelligence community may consider adding prediction market probability values to reports or briefs delivered to policymakers in order to help them make more informed decisions. Here, prediction market scores are used as a complement to intelligence analysis, not a replacement.

While prediction markets have incredible accuracy and foresight, their limitations must be understood. As forums of trading and investment, they have the potential for manipulation. Some have voiced concerns over wealthy individuals or groups artificially inflating perceived support during elections. Bloomberg reported that a significant portion of volume on Polymarket may be inflated by artificial activity.

Ethical considerations are also central. Critics raise concerns about societal desensitization to violent events and disrespect for important ideals. At the same time, intelligence agencies already engage in practices viewed by some as immoral in order to safeguard national security. For the mild discomfort some may feel in purchasing contracts on potentially violent events, the value in predicting them ahead of time and allowing for preventative measures may be of far greater value.

Other limitations around the nature of predicting future events also arise. Western governments tend to utilize more statistical methods when analyzing extremism, while non-Western governments often prioritize behavioral and online trends. In short-term situations, a prediction market may be less useful.

With these markets, people put their money where their mouth is. In the age of the prediction market, it is imperative that governments understand the immense value they hold in predicting the future. Internal markets have failed in the past, but modern platforms feature thousands of markets with global participation and billions of dollars in volume. To have the potential for such accurate information at one’s fingertips, free of charge, should not be taken lightly.

Prediction markets should not replace intelligence analysis, but when used carefully and ethically, they can significantly enhance forecasting, support policymaking, and strengthen national security decision-making.

Bibliography

1. Office of the Director of National Intelligence. Intelligence Community Directive 203: Analytic Standards. Washington, DC: Office of the Director of National Intelligence.
2. Defense Advanced Research Projects Agency. Future Markets Applied to Prediction (FutureMAP). Arlington, VA: U.S. Department of Defense, 2001.

3. Wolfers, Justin, and Eric Zitzewitz. “Prediction Markets.” Journal of Economic Perspectives 18, no. 2 (2004): 107–126.
4. Hanson, Robin. “Decision Markets.” IEEE Intelligent Systems 14, no. 3 (1999): 16–19.
5. United States Senate. Congressional Record, statements regarding the Policy Analysis Market, July 2003.
6. Central Intelligence Agency. “Markets for Intelligence?” Studies in Intelligence 48, no. 3 (2004).
7. Central Intelligence Agency. “Using Prediction Markets to Enhance U.S. Intelligence Capabilities.” Studies in Intelligence 54, no. 4 (2010).
8. Intelligence Advanced Research Projects Activity. Intelligence Community Prediction Market (ICPM). Washington, DC: Intelligence Advanced Research Projects Activity, 2009.
9. Intelligence Advanced Research Projects Activity. ICPM Final Program Review. Washington, DC: Intelligence Advanced Research Projects Activity, 2020.
10. Center for Strategic and International Studies. Forecasting in an Era of Strategic Uncertainty. Washington, DC: Center for Strategic and International Studies.
11. National Intelligence Council. Global Trends. Washington, DC: National Intelligence Council.
12. Bloomberg News. “Polymarket Volume Raises Questions About Artificial Trading Activity.” Bloomberg.
13. Accuracy Metrics and Brier Score Methodology. Polymarket.
14. Office of the Director of National Intelligence. Open Source Intelligence Strategy. Washington, DC: Office of the Director of National Intelligence.
15. Office of the Director of National Intelligence. National Intelligence Strategy. Washington, DC: Office of the Director of National Intelligence.