Event suffix and remaining time prediction are sequence to sequence learning\ntasks. They have wide applications in different areas such as economics,\ndigital health, business process management and IT infrastructure monitoring.\nTimestamped event sequences contain ordered events which carry at least two\nattributes: the event's label and its timestamp. Suffix and remaining time\nprediction are about obtaining the most likely continuation of event labels and\nthe remaining time until the sequence finishes, respectively. Recent deep\nlearning-based works for such predictions are prone to potentially large\nprediction errors because of closed-loop training (i.e., the next event is\nconditioned on the ground truth of previous events) and open-loop inference\n(i.e., the next event is conditioned on previously predicted events). In this\nwork, we propose an encoder-decoder architecture for open-loop training to\nadvance the suffix and remaining time prediction of event sequences. To capture\nthe joint temporal dynamics of events, we harness the power of adversarial\nlearning techniques to boost prediction performance. We consider four real-life\ndatasets and three baselines in our experiments. The results show improvements\nup to four times compared to the state of the art in suffix and remaining time\nprediction of event sequences, specifically in the realm of business process\nexecutions. We also show that the obtained improvements of adversarial training\nare superior compared to standard training under the same experimental setup.\n