Research isn't applicable in learning. If you always play safe, it means you wont try things that you need to learn.
Regarding friendly games, the score doesnt matter. Why you need to win your friends? Why wouldn't you try to learn something instead? You can develop only X amount of game sense. Its one aspect you can work on, but with wider toolbox you can use it more effectively. You have to learn all the relevant tools.
Risk calculations are important, yes. But for example my hard viboras are way riskier than someone who is better than me. If I go for a win, Ill always hit it quite soft. With my skillset its the wisest choice.
Is the solution to just live with soft vibora or try to make it more aggressive by practicing it? With smart goals, youll be able to alternate the risk calculations in a quite short period of time in favor of more difficult shots.
Lets imagine someone only plays matches. If you shoot 5 hard viboras 4x week, you have succesfully done 80 repetitions in a month. If you decide to go for safe option, youll have 0 repetions of harder vibora in a month. In 2 months you have done harder vibora 160 times instead of 0.
Other player might have shot it to a wall 40 times and other player 0 times, but guess whos improving and who isnt?
The risks should have sensible goal which isnt too far away and unrealistic. If you want to learn pro level shots, your skillset should be near pro level. Idea is to take risks on the shots, you can realistically achieve and implement on your game.
At first trying to do something will be bad habit. Eventually it will transform into a valid skill, if you have done it wisely.
Currently I have a bad habit of going for topspin even thought the situation asks for other shot. The reason is that I'm practicing my topspin shots at the moment and even thought its currently more harmfull than usefull, it will be important part to add to the game.