The colloquial statistical mechanics explanation of entropy as if it is caused by probability is dissatisfying to me, in part because it allows highly organized (i.e. with a real potential for work) arrangements to appear as 'random fluctuations', though with very low probability. But as far as I know (not a physicist!) we don't even see tiny, less improbable but still significant fluctuations toward 'new' potential for work, much less the big, super-improbable ones. Is there a constraint on the fluctuations of 'random' systems like gas molecules in a box that would not appear if we simply add the probabilities at equilibrium? Another way of asking the same question: are there experimentally supported equations for how the probability distribution for a volume of gas or other entropically constrained system changes as the system begins to fluctuate away from maximum (i.e. equilibrium) entropy and toward some significant potential for work? My dissatisfaction with the statistical 'explanation' is in part because the arrangements of molecules in a box of gas are self-interacting, so that any shift in a counter-entropic direction, and toward ‘free’ work, should (at least to my layman's thinking) change the probability distributions in nonlinear ways that might reduce 'very highly improbable' to zero probability. Mathematical answers are welcome (‘are there equations?’), but I am a visual and intuitive not a mathematical thinker so translations into non-math or intuitive concepts would be greatly appreciated. Thanks!