Social networks as the virtual equivalent of the ancient agora have become a preeminent space of political discourse. They can nurture new political trends and reveal existing ones. Acting as public opinion logs they provide an insight to the evolution of political views and thus have the potential to become proxies of future election results. By combining them with traditional approaches a better image of the political milieu can be pictured and more accurate forecasts be yielded. Twitter specifically being one of the most popular social networks attracting virtually all types of actors of the public sphere and being text-oriented has become the go-to platform for opinion mining. In this context a methodology is provided to produce metrics from tweets that quantify political leanings. These metrics are used afterwards along with opinion poll results to forecast election results. The elections considered are the US presidential election of 2016. Daily metrics were produced based on sentiment measurements of tweets referring to the major political parties and written during the 6-month pre-election period. The metrics were combined with nation-wide opinion polls conducted in the same interval. A forecasting model was built on top them that predicts the outcome of the elections. The described methodology can easily be applied to other elections.