Stochastic models of electricity spot prices depend on price spikes and long-term seasonality. Therefore it is crucial to determine suitable methods for the identification of price spikes and the modeling of long-term seasonal components (LTSC). Following recent studies (Janczura and Weron, 2010; Janczura et al., 2013), we compare the proportion of observations identified as outliers for five different outlier detection methods and three approaches to long-term seasonality modeling. After removing the effects of outliers, we compare the out-of-sample forecasting performance for three categories of long-term seasonality models: dummies, Fourier series, and wavelet-based methods. We consider various combinations of each approach and perform a comprehensive backtesting comparison at different forecasting horizons for the recently liberalized Turkish electricity market.