Abstract Electoral prediction from Twitter data is an appealing research topic. The article aims at inferring the results of 2019 Spanish Presidential elections analysing Tweets. It defines a specific political dictionnary to analyse the sentiment and the opinion of the messages posted during the campaign. Our predicting model compares the performance of 5 multi-linear regression algorithms and our results are compared to the ones delivered by the standard poll systems based on telephone survey. Our methodology correctly ranks the candidates and gives for the winner of the election (Sanchez) a better prediction of voting share than the national polls. This stream of studies is still in the early stage even if our findings look like very promising. Therefore, as a future line of research, we recommend to include more socio- and economic factors like sex, age, location, etc. in the objective to improve our model and results.