This study examines how children (N = 196, ages 4-10) balance fairness and self-interest when making costly third-party interventions. Using a third-party trust game, children made decisions in both punishment and compensation contexts across three degrees of unfairness. Dynamic time warping (DTW) clustering was applied to identify distinct intervention patterns across unfairness conditions, offering a novel approach to capture children's fairness behaviors as coherent trajectories rather than isolated responses. Results revealed that 4-year-old children, in low unfairness conditions (i.e., a 6:4 distribution), displayed more intervention behaviors than 10-year-olds, often going beyond what was required to restore fairness. In contrast, older children's interventions were more aligned with fairness demands. In low unfairness conditions, inhibitory control predicted intervention intensity, but its influence diminished with age. Developmental differences were evident in strategies employed across age groups, particularly in punishment contexts. Younger children tended to display less clear strategies, while older children were more likely to exhibit either fairness-oriented or self-interest-oriented strategies. Negative empathy predicted fairness-oriented decisions in punishment situations, with children exhibiting higher negative empathy more likely to be fair. However, the relationship between empathy and third-party compensation was less clear, as empathy did not predict whether participants would perform fairness-oriented or self-interest-oriented behavior. These findings suggest that with age, children's intervention strategies in third-party fairness situations become more context-sensitive, shifting from spontaneous and excessive punishment in mild unfairness conditions to more calculated and directed decisions, predominantly fairness- or self-interest-oriented. By contrast, third-party compensation remained unaffected by age in terms of orientation.