Photovoltaic (PV) backsheet polymers degrade over time when exposed to the UV, especially the loss of tensile strength in multilayer polyethylene terephthalate/ethylene vinyl acetate (PET/EVA). The traditional OLS and WLS regressions do not account for the regime dependence and nonlinearity of the kinetics. The 511 daily field measurements of a commercial PET/PET/EVA backsheet were used in this study to fit a Generalized Additive Model (GAM) with penalized thin-plate regression splines and ridge regularization. The GAM significantly outperformed benchmarks: in-sample R² = 0.9978 and RMSE = 0.80 MPa, compared to OLS (R² = 0.9249, RMSE = 4.69 MPa) and WLS (R² = 0.9202, RMSE = 4.84 MPa). The residual variance analysis revealed that the residual variance scaled with the cumulative UV dose (α = 0.5917, p < 0.001) and varied by 13.4-fold. A key finding was biphasic degradation: 0.152 MPa/(MJ/m²) in Phase 1 (UV < 196.53 MJ/m²), accelerating 3.79× to 0.575 MPa/(MJ/m²) in Phase 2. Over 511 days, tensile strength dropped 73.17 MPa (30.2% loss from 241.93 MPa initial). The results show that GAMs are a better method for modeling the reliability of PV backsheets.