Background. To date, little is known for pacing in ultra-endurance athletes competing in a non-stop event and in a multi-stage event, and especially, about pacing in a multi-stage event with different disciplines during the stages. Therefore, the aim of the present study was to examine the effect of age, sex and calendaryear on triathlon performance and variation of performance by events (i.e. swimming, cycling 1, cycling2 and running) in 'Ultraman Hawaii' held between 1983 and 2015. Methods. Within each sex, participants were grouped in quartiles (i.e. Q1, Q2, Q3 and Q4) with Q1 being the fastest (i.e. lowest overall time) and Q4 the slowest (i.e. highest overall time). To compare performance among events (i.e. swimming, cycling 1, cycling 2 and running), race time in each event was converted in z score and this value was used for further analysis. Results. A between-within subjects ANOVA showed a large sex×event (p=0.015, η 2 =0.014) and a medium performance group×event interaction (p=0.001, η 2 =0.012). No main effect of event on performance was observed (p=0.174, η 2 =0.007). With regards to the sex×event interaction, three female performance groups (i.e. Q2, Q3 and Q4) increased race time from swimming to cycling 1, whereas only one male performance group (Q4) revealed a similar trend. From cycling 1 to cycling 2, the two slower female groups (Q3 and Q4) and the slowest male group (Q4) increased raced time. In women, the fastest group decreased (i.e. improved) race time from swimming to cycling 1 and thereafter, maintained performance, whereas in men, the fastest group decreased race time till cycling 2 and increased it in the running. Conclusion. In summary, women pace differently than men during 'Ultraman Hawaii' where the fastest women decreased performance on day 1 and could then maintain on day 2 and 3, whereas the fastest men worsened performance on day 1 and 2 but improved on day 3.