This article deals with standard strength tests of new float glass and focuses on the common four-point bending test. Under similar test conditions, glass specimens behave differently in terms of their fracture patterns, location of the fracture origins, ultimate loads at failure, and the tensile strengths at failure. Therefore, standards require a minimum of 30 specimens in a tested sample and present different requirements with regard to face and edge types of failure, the location of failure within the shear span, etc. This article aims at addressing some of these aspects through an experimental study of relatively large samples subjected to four-point bending and a complementary series of three-point bending tests, and by employing a stochastic theoretical model that helps to gain insight of the findings and generalize the conclusions. The article examines whether edge failure specimens should be excluded from the entire tested sample, investigates whether a failure within the shear span differs from a failure in the central constant bending moment zone of the tested specimen, examines the effect of the sample size, and discusses the effect of the specimen size on the results. The study finds that although edge failure specimens may be included in the strength evaluation, they should better be excluded. It is also found that there is no difference between shear span zone and central zone failures. The study finds that there is a considerable scatter of results when a limited size sample of 30 specimens only is tested and caution is required in the interpretation of the results of small size samples. Finally, yet importantly, the article examines the different tensile strength results obtained from different samples following different standards, because of the size effect, and discusses what the real tensile strength of glass is.