The article considers the temperature distribution in steel during its continuous casting. Temperatures were measured sequentially in the steelpouring ladle (one measurement) and in the tundish (two measurements) using a platinum-platinum-rhodium thermocouple with an accuracy of ±4 °C. We have analyzed the results of 170 casts of two steel grades: 5SP and 35GS. The type of temperatures set distribution was verified on the basis of three goodness-of-fit criteria: Pearson’s χ-square criterion, λ Kolmogorov-Smirnov criterion and W Shapiro-Wilk criterion. The results obtained are consistent with the physical picture of steel casting. The metal in steel-pouring ladle is practically in a stable state and is subject only to natural cooling through the lining, top and ladle body. In the variant of analyzing a sample of temperature values in tundish at the first and second measurements, the hypothesis of normal distribution should be rejected. Here, the steel temperature depends on a number of parameters, including the feed rate and casting rate, feed time and composition of slag-forming and heat-insulating mixtures, etc. Attempts to establish the relationship between the steel temperatures of in steel-pouring ladle and tundish were unsuccessful. Considering the temperature measurement in tundish as two sequential data arrays, the first of which is an argument, and the second is a function, a linear relationship between these arrays was established. This relationship between the first and second temperature measurements in the tundish can be used to estimate the steel final temperature at thermocouple readout, including in the event of a failure. The results of the work can be used in development of a mathematical model of steel casting.