Background Spinal cord injury leading to paraplegia affects the mobility and physiological well-being of one in a thousand people. Powered exoskeletons can temporarily restore the ability to walk. Their relevance in daily life is still limited because of low performance beyond ground that is even. CYBATHLON is an international competition promoting improvements in assistive technology. In this article, we present the latest design and results of testing of TWIICE One version 2018, one of the competing devices in the 2020 race. Methods A person with a motor-complete spinal cord injury at thoracic level T10 participated as race pilot. Training ahead of the race took place over one week at a rate of 2 h per day. The time to perform each of the seven tasks of the competition was recorded together with the number of repetitions. Performance is compared over the training period and against the 2016 race results. Results Progression was observed in all tasks and accounted for by both user training and technology improvements. Final competition rank was second out of seven participating teams, with a record time of 4 ' 40". This represents an average improvement of 40% with respect to comparable obstacles of the 2016 race, explaining the two ranks of improvement since then. Conclusion These results help understand which features had a positive impact on the real-life performance of the device. Understanding how design affects performance is key information to create devices that really improve the life of people living with paraplegia.