Two-dimensional diffusion calculations, as commonly used, are unreliable for excess reactivity computations of small nuclear systems with significant leakage in the untreated dimension. Observing that the errors are mainly systematic, similarity between nuclear systems is defined and a theory concerning effective buckling is developed that eliminates systematic errors. Excess core reactivities of the materials-test-reactor-type swimming-pool research reactor IRR-1 are then computed with an accuracy better than ±0.5% with 99% confidence, and, in fact, the differences between the computed and measured values are <0.15% for the cases computed.