Always remember the fundamental goal of writing a scientific paper: Intimidation and not communication; if you explain your ideas clearly, people will steal them whereas if they look complex they will admire you.
Do not let your creativity be hemmed by a rigid structure; a good theory will explain itself. Always remember that a scientific exposition is not a work of literature; it is a work of genius.
Show your creativity by not following the whims and fashions of previous notation. For example, it is perfectly acceptable to denote inputs by y and outputs by x, sample size by e and accuracy by m. As a side effect, your complexity results might look a lot better.
Choose your notation carefully. You can safely assume that all your readers will know about strict typing and operator precedence. It is perfectly acceptable that x denotes a number, a vector, a matrix, a string, a list, a graph, an algebra, a topological space, a ring, and a page number at the same time. Always remember that the Latin alphabet has only got 26 letters so do not squander them.
Your paper will have a more sophisticated look if you exclusively use the Greek, Gothic and Hebrew alphabet.
Do not feel obliged to have any connection between the introduction and the main body; after all smart reviewers will only read the introduction anyway.
There are circumstances in which communication is the goal of your paper. In such cases, use Shannon’s information theory results which show that redundancy can improve the communication accuracy. Do this, for example, by simply repeating whole paragraphs of your paper. This is easy with the cut-and-paste functionality of modern editors.
The most important part of your paper are the references. Do not squander a citation on someone whose favour you do not wish to curry. Similarly, it is essential liberally cite those researchers who you conjecture might be reviewers of your paper: this will be the first thing the reviewers will check.
Make sure that you do not acknowledge one of your co-authors; it gives away too much of the paper’s history.
Demonstrate your total concentration on the deep science within your paper by outrageous violation of the rules of grammar; for example, omit verbs from key sentences, deprecate the use of the definite article and utterly ignore conventions of punctuations and capitalisation.
If you managed to become a big-shot then simply assign the task of writing to your student. Make sure he gets acknowledged.
Never write an open question in your paper that you have not already answered. You will only get asked by the editor to solve it.
If you are at a loss as to what to write about, a good strategy is to find an old algorithm in the literature; rename it and obfuscate it (by change of notation – see above). If necessary introduce inefficiencies which you will be able to remove by an “improved algorithm” in a later paper. After a decent interval (a year will do) you can write another paper
Think about a good structure of your paper and try to use consistent and well known notation. Always keep in mind that you write the paper in order to communicate your idea to the readers. Adhere to the rules of grammar and avoid typographic errors.