Here’s a little secret about randomisation. You have to do thousands of trials before you can truly start making statements about statistical bias. If something is truly random, you will sometimes get apparent trends or moments of consistency. For example, take the decimal digits of π, a transcendental number with no known pattern in its digits; starting from the 762nd decimal place, there are six 9s in a row, and it’s the first time there’s a run of the same digit with a length greater than 3. Seeing such a sequence implies that the distribution of digits is not random, but when you go into the thousands or millions of digits, every arbitrary sequence is bound to appear at least once.
There’s also the “gambler’s fallacy”. You got Terran in the last 3 games… what’s the chances of it selecting Terran again? Assuming for the moment there’s no bias, it’s still 1 in 3 - the previous results have no bearing on the outcome.
Select Random a thousand times, and make a note of which race you get each time. If there’s a weighting towards Terran after that, then the hypothesis of it being biased towards Terran will hold water.