It is known (see here for example) that, in a simple graph of odd genus $g$ with $n$ vertices and $m$ edges, the number of cycles of lenght $g$ is at most $\frac{n(m-n+1)}{g}$.

Since this can be be phrased only in terms of circuits and matroid properties (when the graph is connected, we can recover $n$ as the rank of the matroid plus one), I was wondering is this result could be extended to any matroid. More precisey, is it true that, for any matroid of size $m$, rank $r$ and odd genus $g$, the number of circuits of size $g$ is at most $\frac{(r+1)(m-r)}{g}$?

If no, are there some weaker similar bounds on the number of circuits of size $g$?

If yes, what can we say about the equality case? In the case of graphs, it is reached for the Moore graphs. Are there new non-graphical matroids reaching the equality?