This study examined the relationship between complexity, accuracy and fluency (CAF), using longitudinal data. Conducted from the perspective of complex dynamic systems theory (CDST), this research adopted four CAF general measures to track the development of five Chinese undergraduates' English speech over a 15-week period. LOWESS graphs supported by correlation analyses showed that there were individual differences in CAF relations. Different trade-off effects were found in the CAF development of four individuals. One learner improved in lexical complexity at the expense of accuracy, another learner improved in accuracy at the cost of lexical complexity, one student made progress in syntactic complexity but sacrificed fluency, and one improved in accuracy and lexical complexity while compromising fluency and syntactic complexity, respectively. Furthermore, the use of moving correlations demonstrated that CAF relations changed dynamically over time, revealing eight identified change patterns. These patterns exhibited varying degrees of variability, ranging from highest to medium to low-medium to lowest. Six patterns involved state changes, transitioning from a competitive to supportive relationship, for example.