Previous studies showed that depending on different factors, such as age, level of familiarity or group size, the number of agonistic interactions in pigs directly after mixing can demonstrate a large variation: from groups which reveal a clear decline in agonistic interactions one day after mixing to groups with highly temporal fluctuations even four days after mixing. Information of these agonistic interactions is often used to calculate dominance indices and is the basis to construct social networks. However, in order to get information about won and lost fights or about the initiator and the receiver of an agonistic interaction, these parameters rely on direct and continuous video observation. Video analysis is, however, very time-consuming, but provides the most accurate results as it measures directly the agonistic behaviour. Thus, the aim of the present study was to investigate the temporal development of agonistic interactions directly after mixing for weaned piglets, growing pigs and gilts. Furthermore, we evaluated the impact of the above described temporal variations on dominance indices as well as on social network parameters. This information is important to determine the most appropriate time span to obtain stable results for dominance indices as well as parameters derived from social network analysis. The results for growing pigs and gilts revealed a significant decrease from 0.6 to 0.2 agonistic interactions/animal/hour one day after mixing (6 h of video observation) (p < 0.05). Although the frequency of agonistic interactions showed a high fluctuation for weaned piglets, the Spearman rank correlation coefficients after two or three days demonstrated high values for dominance indices and for the social network parameters density and all centrality parameters (range: 0.74-0.98). Exceptions were the fragmentation and the size of the largest weakly connected component describing the whole network structure. Here, only low to moderate correlation coefficients were determined. Growing pigs and gilts revealed similar results. With the mentioned exceptions, stable results were obtained already two days after mixing (17 h of video observation) for weaned piglets, i.e. a third day of video observation did not add significant information. For growing pigs and gilts, also a further condensation to just one day after mixing (6 h of video observation) might be possible due to the high correlations. With this knowledge, video analysis in future studies with focus on individuals' rank position derived from dominance indices or centrality parameters can be performed more efficiently by focussing on shortened observation periods.