This work introduces an improvement upon the existing methodology for novelty detection, which offers distribution-free false discovery rate (FDR) control guarantees. This work proposes two new algorithms for improving power in both high and low outlier proportion settings by using full-conformal e-values to quantify the evidence for novelty in a given dataset. By construction, these methods offer a strong de-randomized alternative to commonly accepted randomized algorithms. By using e-values in a multiple hypothesis testing procedure, this work proves FDR guarantees and shows that new methods outperform the power of the baseline. Moreover, this work demonstrates that the procedure can perform powerfully on small real-world datasets, showing its potential in a wide variety of applications from Physics to Medicine.