We propose to CHEAT approach to the MorphoChallenge contest: Combinatory Hybrid Elementary Analysis of Text. The idea is: acquire results from a number of other candidate systems; CHEAT will read in the output files of each of the other systems, and then line-by-line select the "majority vote" analysis - the analysis which most systems have gone for. If there is a tie, take the result produced by the system with the highest F-measure; if the other systems´ output files are ordered best-first, then this is achieved by simply taking the first of the tied results. To justify our approach, we need to show that this really is unsupervised learning, as defined on the MorphoChallenge website; arguably the CHEAT approach involves super-sized unsupervised learning, as it combines three different layers of unsupervised learning.