Most picked is a stat that gets used so hilariously badly and misunderstood its an achievement on itself. If you make every hero a clone of each other and nothing else matters D.va and Mercy would dwarf every other hero for how much they get picked.
Why does this matter when heroes aren’t clones of each other you might ask. Simple people jumping in gravitate towards designs they like, the rule of cool (i.e being a Cyborg ninja increases people wanting to play Genji) and other non-gameplay reasons to pick a hero. And when you are dealing with one the extreme ends D.va who is the most popular character period for OW this problem gets cranked up to 11.
More popularity equals more people practicing her more practice means more experience on a lot of people. More people practicing her means that you can find a substitute player more often. Finding a substitute player more often for d.va means its easier to plan around her and harder to plan around the other options as they don’t have as many practiced players because not as many people wanted to start learning them in comparison to the most popular overwatch character.
Bare in mind that even when a player doesn’t play d.va they still might be better at d.va than the other option simply due to time played on both heroes. Its not exactly a secret that you will do better with a hero you personally have played more with than one you haven’t messed with alot. And d.va has a huge edge at this simply due to her popularity (more people who start out picking a hero means that more people are likely to have more playtime on given hero) not actual gameplay. But this often gets ignored or discredited when it really shouldn’t be.
tldr version: Alot of the problems with d.va being “good” is actually she has a ton of players and simple experience gaps.