"Sometimes you gotta drive your character like they're a stolen car" - Adam Koebel
One constraint on players' play is the fear of a sub-optimal move. While often the defining characteristic of the power-gamer archetype, this trait can just as easily show up with players who have a more role-playing focused play-style. I can only speak for my group, but coming from years of more tactically-heavy D&D 3.5 games, as well as computer and board games where the main goal (winning the game) relies on making optimal choices, it is often hard for some (myself included) to let go of that mindset.
5e tried to fix this a bit with the inspiration mechanic - sure, you are making a sub-optimal choice, but if it's in-character and/or meshes with your flaws/bonds you get a mechanical reward in the form of inspiration. That carrot goes away if you already have inspiration, or your flaw/bonds don't really fit the situation, or the GM isn't on the ball all the time about rewarding it (something I am certainly guilty of). Without consistency, it's not a good incentive, and by making it a mechanical reward, you are just pushing the optimization decision to be trying to calculate if the value of the reward outweighs the cost of the tactical choice.
As a GM, one (previously huge) peeve of mine is the player that starts making a move or action during their turn, and then wants to re-choose their actions because they started moving and then saw something around the corner, or realized they couldn't move as far as they wanted, or were going to make it within a ranged they needed. I used to make them stick to their original action (unless the choice was clearly from some ambiguous or incorrect information on my part), but really I don't bother forcing the issue any more: They are just going to get miffed about being made to have sub-optimal turn. If that's how they derive their fun from the session, I'm not inclined to put the brakes on it unless it's infringing on someone else's fun.
Some players just get that they, like their characters, are going to have imperfect information and will make bad choices because of it. Other players aren't going to be satisfied unless they achieve as much as mechanically possible each turn. I'd say my current group is a pretty even three-way split, with some players at opposite ends of the spectrum, and some that have a hybrid approach that puts them in the middle. Switching away from grided combat seems to help to some degree, but the tactical crunch of D&D combat is a pretty big factor in the game's appeal for many players, so even that is a trade-off.
Eventually, the decision about how to play is going to have to come from the individual player - the biggest risk you can take with your character is not making the "best" choice, but picking something else because it is fun, or it is just what the character would do, or it's just going to make the story more awesome. As a GM, the best incentive you can provide is making the results of those choices as fun as possible. If the players start to see that their risky choices can lead to more fun by picking something other than mechanical optimization, then risky becomes the best choice.