I have been discussing the following topic with my work teammates;

We were talking about statistics and probabilities under roulette. What they say is that given an amount of time (it could be high or short), if we take examples of something that happened on some lapsus over that time (let's say we take examples numbers given in a roulette by 1 hour and we will play 2), that examples could help us to determine what would happen next 1 hour.

If this is rejected, they may say that statistics could help them maybe with given more time (let's say we can get examples of 1 year).

Because they see it globally and not separately like me (each bet is different one independent of previous others). For example, blacks or reds. The possibility is 50% and 50% (without 0/green) Then in 2 hours the probability would be the same, so, if we take 1 hour and see that 60% is black and 40% is red, then we can use this to bet for red, because it is difficult to have too much black ones still increasing. We would not have ~70%, it is really difficult to have something like that.

Let's say the best scenario for them, 80% are black in that given amount of time (1 hour), what chances do we have to have 90% of blacks in total (so blacks still increase) within the next one hour (to complete the proposed time of 2 hours)?

Could Bayes' Theorem be used in this kind of scenarios (roulette), so previous results could help to predict future ones?

I know that some roulettes have imperfections. Do you think some of this could be useful anyway or, at contrary, it may help?

If you say it is a fallacy (which is what I think), please explain what parts are being misunderstood!

Thank you! :wave::wave::wave::wave: