Plasmopara viticola (Peronosporales: Peronosporaceae) is a highly destructive disease that can cause very serious damage under favorable conditions. Cultural practices, disease-tolerant varieties, and phytosanitary treatments have been effective tools to control this disease. However, predictive models such as decision support systems (DSS) could help reduce the need for phytosanitary treatments. In this study, we aimed to develop, adjust, and validate a customized predictive model at the plot level, comparing it with the other three existing models (Milvit, VitiMeteo-Plasmopara, and Goidanich) in Mediterranean conditions. Downy mildew symptoms were observed in the field and predicted using the four mentioned models in eight plots during the 2018 and 2019 campaigns. All tested models predicted possible daily infection cycles, which were grouped into infection blocks every 5–7 days to validate predicted data with in-field observations. In Mediterranean conditions, Milvit and VitiMeteo-Plasmopara models provided both over-prediction and under-prediction infection blocks compared to field observations. However, the Goidanich model only provided over-prediction infection blocks with very few matches to field observations. The UR-model was successfully improved to adjust it to Mediterranean conditions, and it matched observed and predicted infection blocks in 56,25% of the plots and years. However, the UR-model was more accurate (75% matching) in humid years like 2018 than in dryer years like 2019 (37,5% matching), although in both cases it was the most accurate model to be used as a decision support system.