The non-elicitability of ES have proved to be a big challenge. On top of all, the formal definition of backtestability and revealing its connection to elicitability have brought another challenge. In a breakthrough paper, it has been pointed out by Acerbi − Székely (2017) that ES, the risk measure which already got a foothold in banking regulation, is not backtestable. In the same paper, however, the authors have shown the possibility for ES to admit a so-called ridge backtest which requires the prediction of VaR, which is elicitable, to be involved. It has been built on the idea that ES is jointly elicitable with VaR, shown by Fissler − Ziegel (2016). An appealing property of a ridge backtest is that it has a limited sensitivity to the prediction of VaR and the direction of the bias is known. Furthermore, building on the concept of a sharp backtest, in the most recent contribution Acerbi – Székely (2019) have shown the possibility of introducing the notion of ‘realised ES’ which can serve as an ex-post measure of tail risk. The significance of this finding is that it makes possible to compare the predicted (ex-ante) and realised (ex-post) value of ES, and eventually − in the wording of Acerbi − Székely (2019) − to perform “apples-with-apples comparisons”.1