Indian mustard (Brassica juncea L.) yield in rainfed regions remains stagnant due to restricted recombination and a narrow genetic base. This study quantified additive and non-additive genetic variance for yield traits and identified superior hybrid combinations. A 10 × 3 line × tester mating design produced 30 F1 hybrids, evaluated with 13 parents across two rabi seasons (2022–23 and 2023–24) at Nagpur, India, in a randomized block design with three replications. Twelve agronomic and quality traits were recorded. Analysis of variance revealed significant genetic variability for all traits (P < 0.01). Genotype × season interaction was significant for seed yield per plant (σ2G×S = 4.82; P < 0.01), siliquae number per plant (σ2G×S = 1245.36; P < 0.01), and 1000-seed weight (σ2G×S = 0.089; P < 0.05). The predictability ratio (σ2GCA/σ2SCA) ranged from 0.48 (days to 50% flowering) to 0.82 (point to first siliqua). Values > 0.50 for siliqua length (0.79), siliquae density (0.80), seeds per siliqua (0.80), point to first siliqua (0.82), and oil content (0.74) indicated relatively greater additive variance; values < 0.65 for seed yield (0.63), siliquae number (0.63), and branches per plant (0.56) indicated relatively greater non-additive variance. Both components were significant for all traits; classifications are therefore relative. Brijraj × Kranti produced the highest yield (25.42 g plant−1; heterobeltiosis +170.6%; useful heterosis +170.55% over Kranti), with stable heterosis across seasons (+168.3% in 2022–23; +172.1% in 2023–24). PM-31 (GCA +1.85) and RLC-3 (GCA +2.24) were superior combiners for yield. Transgressive phenotypic expression in F1 (performance exceeding the better parent) occurred in PM-31 × ACN-237 (484.60 vs. 243.47 siliquae) and DRMRQ 1-5 × Kranti (6.17 g vs. 5.40 g 1000-seed weight); because F1 is not a segregating generation, this is heterotic transgressive phenotypic expression, not classical transgressive segregation. Provisional parental groups are proposed as a working hypothesis requiring validation. Promising hybrids are candidates for multi-location evaluation, not commercial deployment. The framework offers a replicable model for self-pollinating crops facing yield stagnation.
