I'm surprised how little theorycraft backed up with stats there seems to be for MTG? Do players normally copy mana curves from someone else, guess/playtest or keep their stats secret? Am I missing something?
I've been experimenting in excel with looking at the probabilities of playing 3 lands in the first 3 turns to see what the optimum amount of lands in deck is for that, taking into account being an aggro deck and not wanting to draw to many and not having enough creatures to play. I assumed I was mulliganning on 0,1,5,6 or 7 lands in opening hand and put down having only 2 or 6+ lands on Turn 3 or mulligan below 5 as mana flood/screw.
24 and 25 Lands gives you a 77% chance of no mana flood/screw on T3, which is the highest possible under my assumptions.
21 Lands gives you a 72% chance and is the sweetspot for avoiding dimishing returns (adding more land adds a smaller success percentage than it used to - so is it worth it?).
23 Lands gives you a 76% chance and seems to be the best of both worlds.
I did the same with the chances of playing a one drop on turn 1, two drop on turn 2 and three drop on turn 3 (again being for an agreesive aggro deck). It didn't fit into the deck with 23 lands but shaving off a percentage from each to use the next optimal and goign for multiples of 4 & 3 gave me:
16 One Mana Spells (89% Chance).
12 Two Mana Spells (79% Chance).
9 Three Mana Spells (67% Chance) - and you can play multiple cheaper spells to deal as a backup.
My best FNM was actually when I copied the curve of a tournament deck withexactlythis distrbution (maybe he ran the same numbers?), so I will try it again this time with more powerful cards.
I'm sure I can refine my assumptions... am I going down a dead end here or is persuing this kind of data helpful?