If not, then I apologize. I thought I regognized the specific wording and pattern of argument as Craig's as opposed to, say, Aquinas'.<quoted text>
No Poly it wasn't copied and pasted from William Craig.
Physicists have gotten away from calling things 'laws' over the course of the 20th century. The second law was formulated in the middle 19th century and uses the old terminology. Today, it would be called a theory.Now in reply, what good is a law if it's only statistical as you say?
This seems to be just a little too convenient. It again appears to be an intellectual escape hatch of poor design. If by your own understanding, the Second Law of Thermodynamics is only statistical, why is it considered a Law and not a hypothesis?
To answer your more basic question, you have to realize that macroscopic objects are made from a *huge* number of atoms. Avogadro's number is much larger than most people can imagine easily. Because of this, statistical analysis is not only useful, it is also necessary for understanding.
For example, no particular atom has a 'pressure' or a 'temperature'. Both of those properties of macroscopic objects are *averages* of microscopic properties (momentum and kinetic energy, respectively). But the averages are taken over so many atoms that temperature and pressure do not deviate from their averages to any measurable degree, even for things of the size of bacteria (which are still huge from an atomic perspective). At the macroscopic level, the probabilities at the atomic level are averaged out to give regularities and causality. In a similar way, properties like the conductivity of a metal are averages of what happens at the atomic level and so are predictable to high accuracy (causal).
But, if you get to collections of only a few tens or hundreds of atoms, the averaging does not smooth out the individual probabilities as well, so violations of statistical laws, like the second law of thermodynamics, can be produced. Once again, a single atom does not have an entropy. Entropy is an averaged property of many, many atoms.
It turns out that at the atomic and subatomic levels, it is almost never possible to predict exactly what any given observation will produce. Instead, we get probabilities for the various possible results. We also get predictions of more detailed statistical properties like the standard deviation, etc. So, when a sufficient number of observations are made, we can compare our predicted probabilities to the actual results and thereby test our theories.
The natural question is whether the probabilities we see are actually the result of a deeper causality that we don't yet see. Such ideas are called 'hidden variable theories'. it turns out that all such theories, merely by being causal, have certain measurable properties (correlations between observations) that are in contradiction with quantum mechanics (which is an acausal theory). The experiments have been done and agree with QM, and not with causality.
At a deep level, the universe is acausal.