This paper presents a detailed analysis of a set of 2 million of search engine results obtained by different settings. More specifically, we used the Terrier platform and RunGeneration interface to create different versions (run in English) by changing the indexing and retrieval models. We then evaluated each result using different information retrieval performance measures. A systematic analysis was conducted on these data to determine firstly what the parameters that have the most influence were; on the other hand what the values of these parameters most likely lead to good system performance were.