Worldwide Forecasting Contest Supports Efficient Market Hypothethis

The M6 Financial Forecasting Contest has drawn to a close, with a strong result for the Efficient Markets Hypothesis.

WEST PALM BEACH, FLORIDA, UNITED STATES, February 14, 2023 /EINPresswire.com/ — The M6 Financial Forecasting Competition has concluded. This year-long contest was the first major machine-learning competition to directly address the Efficient Markets Hypothesis. The organizers hoped it would “provide empirical evidence about how investors can improve the accuracy of their forecasts, mitigate the uncertainty involved in these forecasts, and exploit their findings to build robust, profitable portfolios”.

The contest challenged participants in two separate ways. First, they were required to provide monthly probabilities that asset returns would lie in a given performance quintile. Second, they were required to provide a portfolio that could be updated at most once a month. With the contest leaderboard now frozen, three student entries have found their way into the top ten. One participant, Miguel Perez Michaus, won prizes in two categories. Methodologies used by the winners are yet to be released.

One top-ten entry has attracted considerable reaction on social media because it contained no forecast model. The microprediction contest entry was revealed to be a mere proxy for the intelligence present in the financial options markets. It contained no forecasting model in-and-of itself. In an article on the his options-driven entry, author Peter Cotton also revealed that this entry had finished in the top quartile of forecasting performance 5 out of 5 times (the pilot stage of the contest, and in all four quarters). It had beaten 96% of all participants overall.

Cotton also presented evidence that the few who beat the options benchmark (which finished within 0.0002 forecasting points of 4th place) were probably lucky (those on the podium had much higher Sharpe ratios). The conclusion drawn by Cotton: not a single team amongst the 168 assembled from around the world was able to prove they could outperform the options market. The Efficient Markets Hypothesis is intact.

A secondary observation from this experiment: the set of methods developed in mathematical finance for exploiting information already present in markets are as crucial as they have ever been, and should complement advances in modeling per se. In a recent book published by MIT Press, Cotton has argued that companies should be thinking hard about using new, lightweight markets in conjunction with existing data science pipelines.

Peter Cotton
+1 646-637-1338
email us here
Microprediction LLC
Visit us on social media:
Twitter
LinkedIn