from David Fay (see original here):

Being a successful geneticist (not to mention scientist) requires a high level of foresight, diligence, and commitment. Half measures and vague notions will seldom suffice. Unlike coursework, there is generally no partial credit in the real world of science. One faulty link in the chain of logic and experimental execution usually leads to zero results. The three keys to success in genetics are as follows: (1) understand from the start exactly what you are doing and what you expect to happen at each step; (2) notice if things do not go as expected; and (3) always take the patented "sledgehammer" approach. The bottom line is that to be an effective C. elegans geneticist you must consistently get things to work the first time. Failure to do so will vastly reduce your progress. In this sense, C. elegans genetics is not substantially different from many other scientific disciplines. Given the time required for worms to develop, however, one can waste significant time and effort before discovering that the experiment has failed. Try hard to prevent this from happening to you.

To ensure the first point - thoroughly understanding your experiment from beginning to end - it will almost always be necessary to draw out the entire set of crosses, taking into account and quantifying all possible outcomes. This is particularly true when you are just starting out in genetics, and you will want to do this before picking a single worm. Remember this: if your basic strategy is flawed, then all the experimental diligence in the world won't save you. Each genetic situation will have unique considerations. By drawing out the entire genetic flowchart, complete with all possibilities, one can nearly always guarantee a good result. Avoid at all costs a faulty scheme. DRAW IT OUT!

With respect to the second point, it is essential that you quickly and consistently note any inconsistencies between the expected results and those actually obtained. This requires looking hard at your plates over the entire course of the genetic procedure. Continually ask yourself if the observed plate phenotypes make sense and if the approximate ratios are in line with your expectations. Do not sweep any significant inconsistencies under the rug! This is a red flag and may be telling you that either one of your starting strains is not as advertised or that there is a fundamental error in your experimental design. Both situations are your responsibility to avoid. Bad or incorrectly described strains can generally (though not always) be detected by a careful examination of the strain before beginning the experimental process. Rather than investing weeks or months of your time in trying to work with a questionable strain, obtain (or generate) a correct version of the strain from some other source, or possibly come up with an alternative strategy for your experiment. Sometimes it may be difficult or impossible to know if a strain is definitively correct. To some extent we must operate on faith, and we are usually safe in doing so. It is always advisable, however, to have multiple pieces of corroborating data before moving on to subsequent steps, particularly when it comes to genetic mapping.

Finally, always take the "sledgehammer" approach. The bottom line is that it usually takes only a couple of extra minutes to pick a few more animals or to set up additional plates for matings. Contrast this to the days or weeks that can be lost if sufficient animals were not picked to isolate the necessary genotype or generate sufficient numbers of crossprogeny. Plates are cheap, but your time is precious.