Short version: I don't see what information on-centre bipolar cells are actually capturing.
Actually, the question could be extended to on-centre retinal cells as well, but I'll focus on bipolar cells for now.
On-centre cells will fire when a spot of light hits the centre of their receptive fields and doesn't touch their surround, off-centres are the reverse (light in the centre is inhibitory, surround is excitatory).
Here's an over-simplified illustration that I made, red is hyperpolarized (excited), blue is hypopolarized (inhibited). Let's pretend that horizontal cell effects have already happened, and ignoring overlapping receptive fields for simplicity.
So my question is, if off-centre cells would be useful to highlight contrasts like edges (as I understand it, though eventually edges get detected further down in the complex cells), what are on-centre bipolar cells used for?
From my perspective it looks like they'd be good for single points in light which are small enough to hit the small number of rods that encompass the centre of a receptive field but not hit the surround. But when would this ever be the case? The centre of a receptive field would be incredibly tiny given the large number of rods in say the fovea (though, granted, maybe on-centre cells become more important in the periphery?).