This paper evaluates the performance of Machine Learning (ML) methods compared to traditional statistical methods in time series forecasting. Using a subset of 1045 monthly time series from the M3 Competition, the authors compare the accuracy of popular ML methods with eight traditional statistical methods. The results show that ML methods are dominated by statistical methods in terms of both accuracy measures and across all forecasting horizons. Additionally, ML methods require significantly more computational resources than statistical methods. The paper discusses the reasons for the lower accuracy of ML models and proposes possible ways to improve their performance, such as enhancing their ability to learn from future information and reducing overfitting. The findings emphasize the need for objective and unbiased testing methods to evaluate the performance of forecasting models, particularly through large-scale competitions.This paper evaluates the performance of Machine Learning (ML) methods compared to traditional statistical methods in time series forecasting. Using a subset of 1045 monthly time series from the M3 Competition, the authors compare the accuracy of popular ML methods with eight traditional statistical methods. The results show that ML methods are dominated by statistical methods in terms of both accuracy measures and across all forecasting horizons. Additionally, ML methods require significantly more computational resources than statistical methods. The paper discusses the reasons for the lower accuracy of ML models and proposes possible ways to improve their performance, such as enhancing their ability to learn from future information and reducing overfitting. The findings emphasize the need for objective and unbiased testing methods to evaluate the performance of forecasting models, particularly through large-scale competitions.