What's the level of variance for playing cash games? What I mean is what is an adequate sample size of games before you can conclude that you are not as good as your opponent?

How far are the swings in points for any given 3 hour session of playing live?

can someone pls answer this question? would be really interessted about this as well

There's really no easy answer. The swings for a three-hour session between two players reasonably close in ability would be plus or minus 20 points for the most part. A occasional session might get to +40 for someone. If one of the players is much weaker, or on tilt, or wants to start the cube on 2 or 4 or 8, then you could get 100-point swings pretty quickly.

If you're playing someone and you really don't know who's better or worse, just keep playing. If one guy eventually gets up 60 or 70 points, then he's almost certainly the better player.

Don't forget to have some fun. It is a game, after all.

From bot data, for players of the same strength, the Standard deviation of unlimited games is about 2.700.

What that mean is that over N games, there is 95% chances that the score will be within
D=2.7*1.96*sqrt(N). for N=100, D=53, for N=1000 it's D=167.

Applying that to real life:
imagine you play someone about 200 elo point weaker. your expected point per game (ppg) is 200/50*0.100=0.400.
After 20 games there are 95% chances that the score is between -16 and +32.
After 50 games there are 95% chances that the score is between -17 and +57.
After 100 games there are 95% chances that the score is between -13 and +93.
After 200 games there are 95% chances that the score is between +5 and +155.
after 500 games there are 95% chances that the score is between +233 and +567.

So if after 170 games, you are ahead you can be fraily confident that you are at least 200 Elo stronger than your opponent.
For players 100 points apart it will take around 700 games to be able to draw the same conclusion.

Your last paragraph doesn't make sense. If I'm up after 170 games, all I can conclude is that I got some good evidence that my opponent is not much stronger than I am. I'm going to be up half the time (well, almost half, on 0 occasionally) even if I'm playing a clone of myself.

You are correct, I expressed that very badly (and wrong).

What i meant is that if your expected average per game is 0.400, 170 games is the point where you have 95% confidence to be ahead in the score.