The E-index of a scientist's portfolio \(\mathcal P = \{c_1, ..., c_N \}\), including
N papers totaling \(C_{tot}\) citations is
$$E (\mathcal P) = -\frac{1}{N}\sum_{i=1}^N c_i log \frac{c_i}{C_{tot}}$$
which is just the product of the average number of citations \(C_{avg}\) and the Shannon entropy of the citation distribution. Therefore, to have a large E-index one needs to have a high value of \(C_{avg}\),
i.e., high average impact, and a high value of the citation entropy, which corresponds to a portfolio with consistent quality, as opposed to having isolated big hits standing out of a bulk of low-impact works. In our paper we have shown that the E-index is more capable at identifying future Nobelists than current metrics, like the H-index. For more technical details, see our
GitHub repository.