BACKGROUND: Patients classified as POSEIDON groups 3 and 4 have diminished ovarian reserve and predicted poor ovarian response. Although GnRH antagonist, mild stimulation, and progestin-primed ovarian stimulation (PPOS) protocols are widely used, direct comparative evidence from homogeneous cohorts remains limited. METHODS: This retrospective cohort study included 592 first IVF cycles (GnRH antagonist: n = 203, mild stimulation: n = 243, PPOS: n = 146) in patients whose sole IVF indication was POSEIDON groups 3 or 4. Entropy balancing achieved near-perfect balance across nine baseline covariates. The primary outcome was cumulative live birth rate per initiated cycle. RESULTS: After entropy balancing, cumulative live birth rates showed no statistically significant differences: 32.8% for antagonist, 29.2% for mild, and 35.8% for PPOS (global P = 0.452). Cumulative clinical pregnancy rates were also similar (37.9%, 34.6%, and 39.5%; P = 0.649). Weighted logistic regression confirmed no significant protocol differences (all P > 0.05). No significant interaction was observed between protocol and POSEIDON subgroup for either outcome (P for interaction = 0.384 and 0.213). Although the antagonist protocol yielded superior embryological outcomes (more oocytes, usable embryos, and high-quality embryos, all P < 0.001), this did not translate into higher cumulative live birth or pregnancy rates. Critically, rates of no oocytes retrieved (6.0%, 6.4%, and 7.6%; P = 0.888) and cancellation due to no usable embryos (16.5%, 19.3%, and 16.2%; P = 0.786) were nearly identical across groups, suggesting a potential threshold effect that warrants further investigation, as protocol choice did not significantly affect the probability of obtaining at least one embryo in this exploratory analysis. PPOS had the lowest premature LH surge rate (1.5%, global P = 0.047). CONCLUSIONS: In POSEIDON groups 3 and 4 patients, the three protocols showed no statistically significant differences in cumulative live birth and clinical pregnancy rates despite differences in embryo yield. This exploratory finding suggests that embryo obtainability may be more strongly associated with intrinsic patient characteristics than with protocol choice, although this requires confirmation in future prospective studies.