Uncertainty estimation and ensembling methods go hand-in-hand. Uncertainty\nestimation is one of the main benchmarks for assessment of ensembling\nperformance. At the same time, deep learning ensembles have provided\nstate-of-the-art results in uncertainty estimation. In this work, we focus on\nin-domain uncertainty for image classification. We explore the standards for\nits quantification and point out pitfalls of existing metrics. Avoiding these\npitfalls, we perform a broad study of different ensembling techniques. To\nprovide more insight in this study, we introduce the deep ensemble equivalent\nscore (DEE) and show that many sophisticated ensembling techniques are\nequivalent to an ensemble of only few independently trained networks in terms\nof test performance.\n