There is a large and growing movement towards the use of wearable technologies for sleep assessment. This trend is largely due to the desire for comfortable, burden free, and inexpensive technology. In tandem, given the competitive nature of professional athletes enduring high training load, sleep is often jeopardized which can result in adverse outcomes. Wearable devices hold the promise of increasing the ease of monitoring sleep in athletes which can inform health and recovery status, as well as aid performance optimization. However, wearable devices typically lack sufficient validity to assess sleep – and especially sleep stages. To address this concern, the present study aimed to validate an algorithm to detect wakefulness, light sleep, deep sleep, and REM sleep against the gold standard polysomnography (PSG), using a wearable single channel electroencephalogram (EEG). Through the single channel EEG, machine learning models were built to infer sleep staging. The model was created from training and validating EEG output and labels assigned from the PSG software. Additionally, to determine the accuracy of agreement between the devices both Random Forest and a deep learning Convolutional Neural network model were implemented. The sleep staging output was consistent with our sleep staging algorithm for the single channel EEG and more notably, the sleep versus wake agreement was strong- above 80%. Our findings show that machine learning algorithms can be used with wearable devices to accurately detect, not only the sleep versus wake cycles, but the 4 sleep stages as well. Accordingly, this technology can be applied in an athlete population for accurate assessment of full sleep architecture.