I have tried blowing with a straw.
If I blow under the flywheel, and the top is spinning on a concave mirror, the side of the top where I blow is sucked down, like if the air passing below it causes a local depression.
I agree with Alan who suggests this is Bernoulli effect.
The side of the flywheel that sinks down is 90 degrees later, because of gyroscopic effect, then the top walks moved by the rolling ball tip.
The effect is evident when the top spins in a concave mirror, but it doesn't work (to me at least), if the top spins on a flat surface.
I suspect that, since the bottom of the top is flat, and the mirror is concave, the air should be constrained to expand, when it is blown under the top, channeled between the two divergent surfaces, which is what causes the depression. When the air reaches the center of the mirror, the surfaces become parallel, and then convergents, so probably the air depression is lost at that point.