This paper presents a box-fitting algorithm for detecting periodic transits in stellar photometric time series. The algorithm searches for signals characterized by a periodic alternation between two discrete levels, with much less time spent at the lower level. It is shown that the crucial parameter is the effective signal-to-noise ratio — the expected depth of the transit divided by the standard deviation of the measured photometric average within the transit. When this parameter exceeds the value of 6, a significant detection of the transit is expected. The algorithm performs better than other methods, especially for low signal-to-noise ratios. The algorithm assumes a strictly periodic signal with two discrete values, H and L, and aims to find the best model with estimators of five parameters — P0, q, L, H, and t0. It fits a step function to the folded time series with parameters for the levels in specific intervals. The algorithm minimizes the expression for the deviation of the fit and calculates the Box-fitting Least Squares (BLS) frequency spectrum based on the signal residue. The algorithm is tested on various time series and shows high efficiency in detecting periodic transits. It is compared with other methods such as the W-R method, L-K method, DFT, and Fourier-sum-fitting Least Squares method. The BLS method is found to be more efficient in detecting transits, especially in cases with low signal-to-noise ratios. The computational efficiency of the BLS method is also demonstrated, with execution times being significantly lower than those of DFT for the same number of frequency steps. The significance of the detection depends primarily on the effective signal-to-noise ratio of the transit. The signal is the stellar brightness within the transit relative to the brightness outside the transit, and the noise is the expected scatter of the measured average of the stellar brightness inside the transit. The effective signal-to-noise ratio should exceed 6 for a significant detection. The algorithm is efficient in detecting transits, especially when the signal-to-noise ratio is low, and when many measurements are accumulated. The algorithm is also computationally efficient, with execution times being significantly lower than those of DFT for the same number of frequency steps.This paper presents a box-fitting algorithm for detecting periodic transits in stellar photometric time series. The algorithm searches for signals characterized by a periodic alternation between two discrete levels, with much less time spent at the lower level. It is shown that the crucial parameter is the effective signal-to-noise ratio — the expected depth of the transit divided by the standard deviation of the measured photometric average within the transit. When this parameter exceeds the value of 6, a significant detection of the transit is expected. The algorithm performs better than other methods, especially for low signal-to-noise ratios. The algorithm assumes a strictly periodic signal with two discrete values, H and L, and aims to find the best model with estimators of five parameters — P0, q, L, H, and t0. It fits a step function to the folded time series with parameters for the levels in specific intervals. The algorithm minimizes the expression for the deviation of the fit and calculates the Box-fitting Least Squares (BLS) frequency spectrum based on the signal residue. The algorithm is tested on various time series and shows high efficiency in detecting periodic transits. It is compared with other methods such as the W-R method, L-K method, DFT, and Fourier-sum-fitting Least Squares method. The BLS method is found to be more efficient in detecting transits, especially in cases with low signal-to-noise ratios. The computational efficiency of the BLS method is also demonstrated, with execution times being significantly lower than those of DFT for the same number of frequency steps. The significance of the detection depends primarily on the effective signal-to-noise ratio of the transit. The signal is the stellar brightness within the transit relative to the brightness outside the transit, and the noise is the expected scatter of the measured average of the stellar brightness inside the transit. The effective signal-to-noise ratio should exceed 6 for a significant detection. The algorithm is efficient in detecting transits, especially when the signal-to-noise ratio is low, and when many measurements are accumulated. The algorithm is also computationally efficient, with execution times being significantly lower than those of DFT for the same number of frequency steps.