If you want to define some distance on knots, you should have a (local) transformation/move on knots or or knot diagrams such that any two knots can be connected by finitely many of them.
As mentioned by Ryan, the most important local move is crossing change, and the corresponding distance is called the Gordian distance.
Besides of this, $\sharp$-operation (H. Murakami, Some metrics on classical knots. Math. Ann. 270, 35-45, 1985), $\triangle$-operation (H. Murakami, Y. Nakanishi, On a certain move generating link-homology. Math. Ann. 284, 75-89, 1989) and $n$-gon move (H. Aida, Unknotting operation for Polygonal type. Tokyo J. Math. Vol. 15, No. 1, 111-121, 1992) all are unknotting operations, hence can be used to define a distance between knots. Recently, Ayaka Shimizu proved that region crossing change is also an unknotting operation for knots (A. Shimizu, Region crossing change is an unknotting operation. J. Math. Soc. Japan. Volume 66, Number 3 (2014), 693-708). While this move depends on the choice of the diagram, hence the distance defined by region crossing change are restricted on minimal diagrams.
On the other hand, there also exist some local operations that we do not know whether it can unknot every knot or not, for example the 4-move (see problem 1.59 on Kirby's list for more details and Dabkowski and Przytycki's paper: Unexpected connections between Burnside groups and knot theory. Proc Natl Acad Sci U S A. 2004 Dec 14;101(50):17357-60 for some update).