I'm not sure what you're trying to say, but I'm replying anyway. If there's a sudden gust of wind, say, there's a possibility you might draw a different tile t

I'm not sure what you're trying to say, but I'm replying anyway. If there's a sudden gust of wind, say, there's a possibility you might draw a different tile than you would have otherwise. But, this wouldn't make individual tiles any more or less likely. As long as we can agree that wind has a purely random effect, it does not need to be part of the probability model. Putting tiles into the bag and redrawing them may, however, result in particular patterns. For example, it would be reasonable to believe that LIFO has some influence: last in, first out. This would depend on the individuals involved in shuffling and drawing the tiles, but some patterns common to human players could be identified with some study. I would suggest that word combinations could be more likely to appear than the given random model predicts (and non-word combinations less likely), as the groups of tiles may hang over from previous games.