Avoiding overconfidence: Evidence from the M6 financial competition

Spyros Makridakis, Evangelos Spiliotis, Maria Michailidis

Research output: Contribution to journalArticlepeer-review

Abstract

The M6 competition aimed to identify methods that can accurately forecast asset returns and exploit such forecasts to make efficient investments. Specifically, the forecasting track of the competition required participants to estimate the probability that each of the 100 selected assets would be ranked within the first, second, third, fourth, or fifth quintile with regards to their relative percentage returns. Overall, less than 25% of the teams managed to estimate the probabilities more precisely than a benchmark that assumed equal probabilities for all quintiles. Moreover, those that did so reported inconsistent performance across the 12 submission points and minor forecast accuracy improvements. We identify price volatility as a key driver of forecast deterioration and show that avoiding overconfidence by assuming similar probabilities for symmetric quintiles can improve both forecast accuracy and portfolio efficiency. Interestingly, our findings hold true even when simple methods are employed to estimate the base predictions and investment weights.

Original languageEnglish
JournalInternational Journal of Forecasting
DOIs
Publication statusAccepted/In press - 2024

Keywords

  • Forecast accuracy
  • Forecasting competitions
  • M6
  • Price volatility
  • Probabilistic forecasting

Fingerprint

Dive into the research topics of 'Avoiding overconfidence: Evidence from the M6 financial competition'. Together they form a unique fingerprint.

Cite this