Laser-induced breakdown spectroscopy (LIBS) provides rapid elemental fingerprints, but multiclass classification often requires both interpretable variable selection and a transparent classification rule. In this study, we propose a combined framework that links one-to-the-others wavelength-dependent interclass distance (OTTO-WDID) with sequential threshold-rule classification. OTTO-WDID evaluates each wavelength by comparing one target class with all remaining classes, thereby identifying class-specific emission lines suitable for one-to-the-others decisions. A local baseline-drift correction was applied using adjacent non-emission wavelengths to estimate the emission-line-specific contribution to the observed interclass distance. The framework was applied to LIBS spectra of five edible salt products. OTTO-WDID identified the emission peaks of Ba II at 455.4 nm, Si I at 288.2 nm, Ca I at 612.2 nm, Sr II at 407.8 nm, and Li I at 670.8 nm as class-specific variables for the corresponding sample classes (S1 – S5, respectively). These variables included weak trace-element lines, showing that highly discriminative information is not necessarily associated with the strongest emission peaks. The selected variables were then used to construct a sequential threshold-rule classifier, in which each step separates one target class from the remaining classes using a single class-specific variable and threshold. For the five-class problem, only four active threshold rules were required because the final class was assigned as the residual class. The least reliable Ba II variable was excluded from the active rule sequence, and S1 was assigned as the residual class. Using the sequence S2 → S3 → S4 → S5 → S1, the classifier achieved 100% accuracy in leave-one-out cross-validation. The results demonstrate that OTTO-WDID and sequential threshold-rule classification provide a simple, interpretable, and compact pipeline for multiclass LIBS modeling.