We approach the problem of temporal segmentation of cricket videos. Given simply a broadcast cricket match video, and unlike prior approaches, we have developed a new, simpler framework that does not rely on text for the extraction of 'balls' (a cricketing terminology). Our approach has produced good results on real data sets. We believe the proposed approach is generic, in the sense that the same approach will work for different forms of cricket (50 overs ODI, 20-20 (T20) and test matches. We are also able to filter videos based on user queries.