Anecdotal evidence, like pretty much every example stated in this thread, can be used to argue any idea relating to game shape. The only robust study I've seen of game shape was performed by our resident data guru Josef Ka.
(160760.2) Fairly conclusive evidence that there is some randomness to game shape.
Maybe after Wozz is done figuring out training speeds, he'll tackle the black hole that is game shape. Or even better, he and Josef Ka could team up....