"Simple question. Simple answer.
The table. Or tails. Or he substituted double head coins for the ones you examined. Not so simple.
Using probability to show that the 2LoT is violated by evolution ignores energy used to perform work. The formation of molecules larger than 2 atoms, steam engines, Earth's water cycle, the growth of plants all use a local increase in entropy to perform work. As far as linking entropy exclusively with disorder this is a mistake since it is possible to increase entropy and decrease disorder. The author of this article is conflating the statistical mechanics definition of disorder with the common usage definition.
The three laws of thermodynamics were taught in both the engineering and physics classes. If you first approach thermodynamics from the classical engineeing path of rediscovering how to bore out cannons effeciently, I can see why you'd have the (false) complaint of conflation that you do. There's no sense in learning of heats enthalpic, latent, and transfered of the more general fundaments of the mathematics and physical models involved.
Just to note, that like you, I was always bothered by the non-respect for local phenonmenom that the laws of thermo have -- how they impose perfect gases, perfect diffusions, perfect mixing so as to make use of the classical thermodynamic distributions -- Maxwell-Botlzmann, etc..
So I try to work from the old simple physics mindset. Make simple analogies -- such as the fifty pennies. Use those analogies to provide a mental model to consider aspects of the problem.
Here the fifty unsequenced pennies are being used to to show how the laws of *information* thermodynamics develop and what entropy means, what order means. I could *presto* change it into a heat problem by saying that if heads the pennie is in energy state E1, and if tails, E2. There is a complete parallelism to classical thermo in that regard.