I often see the question "Which has more variance, limit or No limit?"
Hopefully this post should clear things up.

The swings in NL for a given blind structure are much greater. This is why NL requires a much larger bankroll for a given blind structure.

e.g. About $3000 for 200NL compared to about $1200 for 2/4 limit, both with blinds of 1/2.

By the mathematical definition of variance
"The average of the squares of the differences from the mean"
NL has much higher variance when looking at it from a pure $$$ perspective.

However, a skilled player's edge is much greater in NL.
A "good" win rate in NL is 10BB (20x the bb) per 100 hands.
A "good' win rate in limit is 3BB/100 hands.

So looking back at our 1/2 blind structure, A skilled player would be averaging $40/100 hands in NL, and $12/100 hands in limit.

Standard Deviation is the square root of variance.
A good standard deviation for limit is 15BB/100 hands
For NL, it is about 30BB/100 hands

So, while standard deviation is higher in NL, when you compare it to your winrate it is actually lower.

Looking at it from another perspective:

To get approximately the same winrate as our 2/4 limit game you could play 50NL.
Your winrate would be around $10/100 hands with a standard deviation of $30/100 hands,
compared to $12/100 at 2/4 with a standard deviation of $60/100 hands.

This is why the bankroll requirements for the 50NL game are only about $750, compared to our $1200 for 2/4 limit.