Abstract Accurate decisions regarding exposure to and hedging against market risk, both of which are crucial for electricity producers and consumers, depend on a correct assessment of electricity price dynamics. This paper proposes a comprehensive empirical methodology to model price variations in electricity markets based on hydrothermal power generation. This proposal combines advances in time-series econometrics related to regime modeling, conditional heteroskedasticity, and extreme events. The study considers stylized facts of the series describing electricity prices, which are especially relevant for accurate risk pricing. More importantly, the proposed methodology enables the description and characterization of asymmetries at almost every level: intraday patterns, seasonal components, spikes, volatility regimes and extreme values; and the estimation of the effects of such asymmetries on traditional risk measures, such as VaR and CVaR. Different tail behaviors of positive and negative electricity returns are documented, which are relevant for assessing the risks of selling and buying strategies. In addition, asymmetries in risk that are conditional on the time at which the transaction occurs are found.