I think you're close to a good metric, but you need to consider the situations where one person doesn't have a positive EV expectation, or where that expectation is provably wrong. I think those situations can empower a winning non-gambling actor.
One participant in a market can be gambling while another participant isn't gambling. In particular, casinos don't gamble.
Also for many things there exists a scale from fully random to fully skill-based. So in my opinion things can be semi-gambling with a lot of gray area.