The development of helmet-type magnetoencephalography (MEG) systems that do not require liquid helium (e.g., OPM-MEG) has sparked growing interest in steady-state visual evoked field (SSVEF)-based brain-computer interfaces (BCIs). Unlike electroencephalography (EEG), MEG records less distorted signals with a high spatial resolution, covering the entire head without requiring cumbersome electrode attachment. However, conventional algorithms, such as the filter bank-driven multivariate synchronization index (FBMSI), are prone to misclassification in ambiguous cases where the differences between synchronization indices (S indices) are minimal. Additionally, these algorithms fail to fully exploit high spatial resolution and whole-head coverage of MEG. To address these limitations, this study proposes a novel, calibration-free SSVEF classification algorithm termed Spatial Distribution Analysis (SDA). The SDA algorithm utilizes the center of gravity of the S index distribution in the MEG channel space to enhance classification accuracy. Experimental evaluations with 20 participants using a helmet-type SQUID MEG system demonstrated that the proposed SDA algorithm achieved significantly higher classification accuracy and information transfer rate (ITR) across all window sizes. Notably, the largest improvements of 5.76 % in accuracy and 4.87 bits/min in ITR were reported for a window size of 2.5 s. Furthermore, the generalizability of the SDA algorithm was validated on an OPM-MEG dataset, showing performance improvements across all window sizes. The SDA algorithm also mitigated misclassification due to adjacent stimuli and showed short time delay of 0.0907 s, enough to be used for real-time BCIs. These findings highlight the potential of SDA algorithm to enhance the overall performance of SSVEF-based BCI.