I have thought up a wonderful thing, that I like to call: "conditional probabilities." I know some might say that Bayes & Laplace beat me to it but hear me out. Conditional Probability = (probability of B | given that A has already occurred) So, if you suck in the crucible, you have: (Low chance, like it is now) = (Legendary gear | loser) If you did the best: (1 in 10) = (legendary | winner) If you did well: (1 in 20) = (Legendary | Good) huh? eh?? hmmm? GIMMMEEE A YEEEAAAHHAAA???? NOTE: Behavioural Psychology (academic, often with animals) studies have shown that "random rewards" are the most habit forming. In one such study, a rat could hit a button to get cocaine intravenously injected. Under various conditions (i.e. once per minute; or after pressing a different button first... etc.) the rat demonstrated fairly strong behavioural dedication to pushing the button, but not continuously. Under a random reward system however, the rat was willing to push the button hundreds of times (without pause) without getting an injection just in the hopes of getting one based on button presses. I'm just saying there may be a reason why we are all struggling under an equal and low probability system for rewards (most of the time)...