Forecasting by swarm

January 15, 2018 08:04 AM

Artificial Swarm Intelligence (Swarm AI) is a combination of real-time human input and artificial intelligence algorithms enabling an emergent intelligence to form that can provide optimized insights. Modeled after swarms in nature, the process enables human groups to converge upon optimized solutions by combining their knowledge, wisdom, insights, and intuitions using AI algorithms (see “Winners by Unanimous Decision,” MT March 2017). 

For this challenge, Unanimous fielded a Swarm AI system comprised of “market enthusiasts,” effectively building an “artificial expert” that should be able to make significantly more accurate predictions than the individuals could have made on their own. The results are not a simple aggregation or weighted average of the responses provided by the individual, but a real-time system, with feedback loops, that converges upon solutions that maximize the collective confidence of the population.   

Highlights of accurate predictions

Many prior studies have shown that swarm-based forecasting can significantly amplify the predictive ability of human populations. A recent study performed by researchers at Unanimous AI and Oxford University required participants to predict a set of 50 professional soccer matches in the English Premier League over five straight weeks. Individuals averaged 55% accuracy in these predictions, which is reasonable considering outcomes can be wins, losses, and ties. When working together as closed-loop swarms, participants jump to 72% accuracy.  During that same period, the BBC published predictions from an AI-based forecasting system called SAM (Sports Analytics Machine) which achieved 64% accuracy. This study, and others, have demonstrated great promise for swarm-based predictions over other human and machine-based forecasting methods.  

In a study predicting financial markets, researchers at Unanimous AI and Oxford demonstrated that a group of “enthusiast traders” could amplify their predictive accuracy by 26% in forecasting markets (SPX, gold, and crude oil) across six months of weekly forecasts.  Perhaps of greater interest is that the Swarm went 10-1 predicting last year’s NFL playoff games against the spread, topping it off by predicting the exact final score — New England 34, Atlanta 28 — of the Super Bowl. 

Who Participated in MT Survey? 

Unanimous recruited 70 participants to form an Artificial Swarm Intelligence and respond to the questions MODERN TRADER put to its panelists. The 70 participants were recruited from MODERN TRADER readers, followers of Unanimous’ financial intelligence programs, and qualified participants recruited from a paid panel service.  All participants were screened to ensure they at least had “enthusiast” level knowledge of financial markets and speculative trading (see “Swarming the S&P, right”).

While the methodology to answer the survey questions required selecting from multiple choice options is pretty easy to understand, those that required specific predictions were a little more complicated. To answer the question: What will be the biggest fundamental factors affecting the markets in 2018? The swarm considered the relative impact of six factors that could impact markets in 2018: Technical corrections, the value of the dollar, foreign economies, U.S. interest rates, U.S. tax structure, and political instability. Through an iterative evaluation process in which the swarm sequentially identified the factor expected to be least impactful, the swarm ultimately decided that political stability/instability would have the greatest effect on U.S. markets in 2018. 

Following in order of importance was: U.S. tax structure, interest rates, foreign economics, the dollar and technical correction.  


About the Author

Joe Rosenbaum is Community and Communications Officer for Unanimous A.I. @unanimousai