High school students have been reported to have difficulties solving complex algebraic problems. This study therefore investigated the effects of worked example instruction on students' learning outcomes in solving complex algebraic problems. The study was a quasi-experiment that involved a pre-test, an intervention, a post-test, and a delay test. The responses of 72 students (aged 14 to 15 years) were scored following the structure of the observed learning outcomes (SOLO) model and analysed using the Rasch model and regression analysis. The results indicated a significant effect of worked examples from the pre-test to the post-test; however, this effect was not completely retained at the delay test. Also, worked examples had a larger effect on the low-ability students than the high-ability students, but student gender neither influenced nor interacted with learning outcomes at the post-test and delay test. Lastly, the results revealed an interaction between the worked example effects and students' expertise level, with the high-ability students experiencing a full reversal of the worked example effect. These results are explained with respect to element interactivity and expertise reversal effects, and inform mathematics educators and teachers of the conditions of the worked example effect and the implications for classroom practices.