Expressing as percentages gives an unrealistic level of precision (46%, and not 45 or 47!) and puts too much emphasis on "the prediction" and not the probability.
There is a particular potential confusion when making probabilistic predictions of election results. A reader can confuse "40% probability of winning" with "winning 40% of the popular vote". These are vastly different, but seem to have the same headline number. This was seen in the 2016 election
To counteract this, 538 and others have been expressing chances as "1 in 2" to mean "about 50%". 46% is roughly 1 in 2. It could also be roughly 3 in 7 or exactly 23 in 50, but since 46% is only an estimate of the candidates probability of winning, the fractions are kept as simple as possible. However this is just one journalist's algorithm.
538 has a mathematical model of voting. It takes as inputs the results of polling, and background information like state demographics. It output a number representing the probability of each candidate winning the election. It then rounds this number in two ways, firstly to a whole percentage, and secondly to a fraction with a small numerator and denominator. The web site then uses the fraction as the headline figure.