I have done several manual and mental jobs in my life and, no matter the specific task, you can do an excellent job if you like it. Doing research is by far the most challenging job I have faced: it is multifaceted and requires a lot of hard and soft skills, which can only be acquired by growing from time to time. However, it is really stimulating, and since I have been doing it, for me, Monday has never been a tiring day - also because you don't really get away from research!

In my humble opinion, to do good research you should:

• Be passionate: without passion it is not really possible to survive in the academic world

• Work hard: I am convinced that if you work hard, something good can happen

• Keep in mind the fundamental premise: research starts from doubt and uncertainty; otherwise, it is not research

• Choose the right topic: unfortunately, as I wrote here, the career path can be really competitive and can be based primarily on quantitative measures - a hot topic can be better than a niche topic to publish more and get more citations... 

• Know the community: you should know them and they should know you; moreover, they could have a research "paradigm" (benchmarks, venues, standards, etc.) that should be followed

• Having a new idea is intuition, but doing technically sound experiments and writing quality research articles is experience: at the beginning, work with an experienced researcher who can help you learn the "culture of work"

• As I wrote here, new ideas do not come out of anywhere: start with replicating a very recent study that appeared in a top conference/journal (you will have a baseline for comparison and the future work could suggest open issues to address)

• Combining the two points above results in the following: a community working on a topic generally struggles with some open problems that are perceived as necessary to be solved (at least partially) to advance the state-of-the-art

• Especially in machine (deep) learning research, the right way is to adapt an existing method to a new problem or propose a new method that surpasses the state-of-the-art on a known problem

• Choose the right journal: it could be the one in which the articles you cite have appeared

• Avoid predatory journals: a single good article is far more beneficial to your career than many poor-quality papers

• Know your audience: it is better to take for granted what readers know than to be overly didactic

• As I wrote here, think of your article as an hourglass, going from a top-down abstraction to a bottom-up abstraction

• Use LaTeX and be conscientious with structure, grammar, punctuation, formulas, figures, and so on

• Write a convincing cover letter: it is usually underestimated but it is really important

• Do not exaggerate with self-citations

• Recognize the voluntary work of the reviewers and understand the reason for their comments

• Soften over-enthusiastic claims: your research is interesting, but it's just a small piece of a big wall and could be falsified or improved upon soon; even Einstein was cautious with his claims.