1) OBJECTIVE: The main objective of the large-scale quality analysis of S3D movies is to gain a better understanding of how quality control was performed in different movies. Also several novel quality metrics are presented, including channel swap detection, evaluation of temporal shifts between stereoscopic views and depth continuity. 2) METHOD: The main technical obstacle that we had to overcome was an enormous amount of computation and disc space required by such an analysis. Evaluation of one movie could take up to 4 weeks and required over 40GB for the source Blu-ray only. To maximize the efficiency we had to rewrite all of our metrics to exploit the multicore architecture of contemporary CPUs. We have also developed a system that efficiently distributes the computations across the cluster of up to 17 computers working in parallel. It enabled us to finish the evaluation of 105 movies in about 6 months. 3) RESULTS: An evaluation of 105 S3D movies’ technical quality has been conducted that span over 50 years of the stereoscopic cinema history. Our main observations are as follows: According to our measurements, “Avatar” in fact had a superior technical quality compared to the most S3D movies of the previous decade. So it is not surprising that it was positively received by the viewers. S3D quality improvement over the years is fairly obvious from the conducted evaluation, e.g. the results of average-quality movies from 2010 correspond to the results of the 2014 movies with nearly the worst technical quality. A more important conclusion from the analysis, however, is that it gradually becomes possible to produce low-budget movies with excellent technical quality, that was previously within reach only for high-budget blockbusters. We hope that new objective quality metrics like the channel mismatch metric will find their applications in production pipelines. It can further decrease the number of viewers experiencing discomfort and give a start to the new surge of S3D popularity. 4) CONCLUSION: Objective S3D quality metrics make it easier to find problematic frames or entire shots in movies, that could potentially lead to discomfort of a significant fraction of the audience. Our analysis have already revealed thousands of such scenes in real S3D movies. But to directly estimate this discomfort subjective evaluations are necessary. We have organized several of such evaluations with the help of volunteers, that were asked to watch some of the scenes with the worst technical quality according to our analysis. These experiments allow us to further improve the metrics and to develop a universal metric that could directly predict a percentage of the audience experiencing a noticeable discomfort. It is already clear that the development of such universal metric is a very challenging problem, so we are looking for collaboration. It is also clear to us that the majority of problems could be fixed in post-production with minimal user intervention, if not entirely automatically. Some of these techniques are not widely employed just because the problem itself is not considered important enough to require correction. We hope our work could help shed the light on the problem and more attention will be drawn to correcting the S3D production issues. © 2016, Society for Imaging Science and Technology (IS&T).