<![CDATA[The 21st-century education demands high-level thinking skills, yet science teaching at elementary schools still relies on conventional methods, resulting in low scientific problem-solving abilities where students score below the minimum mastery criterion of 70. This quasi-experimental study with a non-equivalent control group design aimed to empirically examine the effectiveness of the Digital Game-Based Learning (DGBL) model assisted by Educaplay and Kahoot platforms in developing students' scientific problem-solving skills based on Polya's four thinking stages. The study population comprised 374 fifth-grade students in Gandusari District, with 21 students from SDN 1 Karanganyar selected as the experimental group and 28 students from SDN 1 Sukorame as the control group through purposive sampling. Data were collected via 20 higher-order thinking (C4–C5) multiple-choice items and analyzed using the Independent Samples T-Test and Normalized Gain (N-Gain) score via SPSS 27.0. The results indicated that both data groups were normally distributed and homogeneous. The t-test on posttest scores revealed that (12.109) exceeded (2.092) at a 0.05 significance level. The experimental group's average score rose from 60.71 to 87.14, achieving an N-Gain score of 68.19% (effective category), whereas the control group only rose from 54.10 to 63.39 with an N-Gain of 35.61% (less effective category). Conceptually, integrating dynamic environmental simulations via Educaplay and real-time formative evaluation via Kahoot successfully optimized constructivism principles. In conclusion, the DGBL model is a superior and adaptive solution that shows high effectiveness in boosting elementary students' 21st-century problem-solving competencies.]]>
