Why RTP does not describe your session
The published figure is an average over an unimaginably large number of plays. A session is a very small sample, and small samples are dominated by variance.
Suppose a game returns 96% and you stake 500 SC. The expected loss is 20 SC. That is the average of every possible way the session could go, weighted by how likely each is.
Very few actual sessions land on the average. In a typical high-volatility slot the most common single outcome is losing most of the balance, and the average is held up by a small number of large wins. The mean and the typical experience are different numbers, and the published RTP only tells you about the mean.
This is why the calculators here report an expected value and refuse to report a probability of any particular result. Working out how likely you are to finish above or below a figure requires the game's full payout distribution — every prize and its probability — which is almost never published.
It is also why 'this game is due' is false. Each play is independent. A game that has returned nothing for two hundred spins has exactly the same odds on the next spin as it did on the first, and the published RTP does not create a correction, a rebound, or a debt owed to the player.
The practical consequence: use RTP to compare games and to price a wagering requirement, which is what it is good for. Do not use it to forecast a session, because it cannot do that and nothing else on this site can either.
To see what an RTP assumption does to a wagering requirement, use the expected balance calculator. To see how widely results scatter around it, use the variance explorer.