Conventional location fingerprint techniques usually require a prebuilt training set of fingerprints sampled at known locations, so that locations of future fingerprints can be determined by comparing to this set. For good accuracy, the training set should be large enough to appropriately cover the area. However, it is not always feasible to obtain a quality training set in practice, and so recent studies have resorted to utilizing fingerprints that are available but without location information. This paper investigates how these so-called unlabeled fingerprints can be useful for location tracking of a mobile device as it is moving. Specifically, we propose a fingerprint-based tracking approach based on Hodrick-Prescott filtering and substantiate its potential via an evaluation study.