Anything that can go wrong, will.
Accordingly, it can only properly be applied to situations in which there is a reasonable expectation of success, unless something goes wrong. The odds of winning the lottery are sufficiently low that Murphy's Law doesn't come into play-- you have no reasonable expectation of picking the winning lottery numbers, so there's no need for anything to "go wrong" in order for you to not win.
Of course, Murphy's Law has a long and distinguished history in science-- as an experimental physicist whose basement lab is prone to flooding, I have more experience than I would like with Murphy's Law and its various offshoots. The really interesting cases for science, though, are the occasional violations of Murphy's Law: cases where experiments turned out to work better than they had any reason to expect.
The most famous example in my area of physics is probably the "Sisyphus cooling" effect in laser cooling, the explanation of which got Claude Cohen-Tannoudji his share of the 1997 Nobel Prize. The full explanation is a little complicated, but it's a very clear example of something working better than it had any right to.
The basic physics the lets you cool atoms with laser light (explanation from the early days of this blog: part 1 part 2) was worked out using idealized two-level atoms. Of course, as Bill Phillips was famously misquoted as saying, "There are no two-level atoms, and sodium is not one of them." Real atoms have many more than two levels-- there are an infinite number of possible bound states for an electron around an atom, though most of them aren't useful-- and even the two "levels" being considered in most laser cooling experiments are in reality collections of sub-levels each having the same energy.
This sublevel structure was known, and exploited to make a magneto-optical trap (MOT) using the behavior of the excited state sublevels in a magnetic field. The ground-state sublevels had never really been considered carefully, though, and if asked, most physicists at the time would've expected them to be a complicating factor that would reduce the effectiveness of the laser cooling mechanism.
In the late 80's, though, some weird things started showing up in experiments with laser-cooled atoms at NIST. To figure out what was going on, they went back and measured the temperature of their atoms clouds very carefully, and were stunned to discover that it was much lower than they expected. In fact, the temperature was something like one-sixth of the minimum temperature they should've been able to reach using laser cooling as it was then understood.
It turns out that, contrary to expectations and in blatant violation of Murphy's Law, the ground-state sublevels they had been ignoring made a completely new cooling mechanism possible. This cooling, it turns out, worked best when the pairs of laser beams doing the cooling had opposite polarization, a condition which also happens to be essential for the operation of a MOT. So, completely by accident, they had set up conditions that got their atoms much, much colder than they had ever expected to see. The much lower temperatures possible with Sisyphus cooling are a big part of what makes laser cooling such a revolutionary technique for atomic physics-- while you could do a lot of interesting physics with atoms at the Doppler cooling limit (the minimum temperature possible without Sisyphus cooling), the Sisyphus cooling mechanism brought a huge and even more fascinating regime of physics into play.
There are some other famous examples of experiments that turned out better than the experimenters had any right to expect. The Stern-Gerlach experiment, which is now famous as the first demonstration of electron spin only worked by accident. The silver atoms they use happen to have a ground state whose only angular momentum comes from a single unpaired electron spin. Most other atoms in the periodic table would've given a much more complicated signal that would've been harder to interpret. They also get bonus Murphy-violation points for initially interpreting their data incorrectly, and for the involvement of a cheap cigar in their success.
The other classic Murphy violation in physics is the Davisson-Germer experiment demonstrating wave behavior in electrons scattered from nickel, which only worked as well as it did because they broke their vacuum system, and then melted their nickel target in the course of repairing the damage from the break. That's pretty hard to top, really.
I'm sure there are others, though. So, what's your favorite example of a Murphy's Law violation in science?
But you see, your examples CONFIRM the Murhpy's law.
After all, the expected results of these experiments should have been negative. But even this went wrong!
I agree with Alex. Not only did the experiments not match up with theory, but the result of game-changing physics coming out of these examples just goes to show that some people can't even screw up properly.
Mathematicians can prove that Murphy's law is a fact.
R. Vakil: Murphy's law in algebraic geometry: badly-behaved deformation spaces, Invent. Math. 164(2006), no. 3, 569-590
NB for those who are not in mathematics: Inventiones is one of the top-rank journals in the field.
You can put as many negatives as you want together to end up either positive or negative
but you can't put two positives together to make a negative.