Low-volatility investing often involves sorting and selecting stocks based on retrospective risk measures, for example, the historical standard deviation of returns. In this paper, we use the volatility forecasts from a wide spectrum of volatility models to sort and select stocks and estimate portfolio weights. Our portfolios are more closely aligned with the ex-post optimal portfolio and deliver large, significant economic gains compared to traditional benchmarks after transaction costs. Importantly, we find that choosing portfolio weights by optimally combining the volatility forecasts from the different models delivers the strongest forecast and financial performance in real-time.