# Martin-Löf randomness characterization

In his 1973 paper On the notion of a random sequence, Levin states (without proving) a characterization of Martin-Löf randomness by writing

Theorem 3. A sequence $\alpha$ is random w.r.t. the distribution $P$ in the Martin-Löf sense if and only if the probability ratio $P(\alpha_n)/ R(\alpha_n)$ is bounded below

where $R$ is the universal semicomputable (semi)measure.

I feel a bit confused how this ratio could not be bounded below, being the ratio of two positive quantities it should always be bounded below by $0$.

The ratio should be bounded below by a positive constant. Equivalently, its infimum should be positive.

• I wonder whether I should have been able to understand this by reading the text...can you point me to some proof of this statement? – Manlio Jun 15 '16 at 9:14
• I suggest checking textbooks on the area. – Yuval Filmus Jun 15 '16 at 9:16