Human mind works in "A cause B", and a major part why people are bad with true randomness is that neural networks are meant to approximate hidden formulas (those "F: A -> B" in nature). And when there is a random, they are stuck as they try to found causation and pattern, and there is no for it.
This is another cognitive bias.
You really should keep track your success and fails as humans perceive fails more seriously than successes (up to 2 times), so even if you have absolutely uniform distribution which gives you 50%/50% chances, in a long run you'll feel that you fail more than succeed due to difference in mental perception.
In nature it's more important to punish for failure, that to reward for success. Thus discrepancy in strength.
It actually can, as you can get an unpredictable seed if you apply the same chaos theory. For instance game can take a first several bytes of data from last TCP packet it got at the moment of seed generation. As there lots of such packets, its not really possible to determine seed and thus number will be random.
Most of randomness in real world occur due to the same reasons. There are explicit formulas for many parts, even human behavior is algorithmic by its nature. Problem arise due to many formulas sensitive to very high orders/small parameters (so called butterfly effect) making proper tracking almost impossible.
