_Looking at how each buff contributes to a team’s chance of winning the game, however, tells us the buffs are statistically pretty close to each other. _
Really? Because it seems to me that people take the Cloud drake only if they are winning. This means they take the Cloud Drake because they are winning while for the other buffs the teams are winning because they take the drake.
I personally take that air dragon at every possible opportunity. People don't contest it as often as the other dragons, and it gives me and my usual rotation/split push strategy a nice boost to safety.
They might not consider it very strong, but they better not complain when their top and bot inhibitor turrets are gone and they can't catch me to make me stop pushing.
Here's how each buff ACTUALLY works:
Infernal: Hey, you have more damage while means your teamfights are snowballing harder and three Infernals almost guarantees you win fights.
Earth: You Siege Objectives better INCLUDING BARON. Since the game is objective based this snowballs harder, and a Ziggs with three Earths essentially means 'Towers no longer exist.' Cho'gath or Nunu can pretty much solo Baron if they didn't feed out the ass and have 3 Earth Dragons.
Ocean: If you aren't in combat you get missing health/mana regen, this is actually stronger for Roamers as laners will need to enter combat to clear waves.
Wind: You take four/eight/twelve seconds less to get to lane... Also three Wind Dragons means 'You aren't stopping that Tryndamere/Singed/Other High-Mobility Split-pusher.'
I will say the Ocean drake seems noticeably strong with the sustain, especially with high sustain champions that build a 
. I think it is healthy for the game that teams fight over dragons, but I also think "giving up" a dragon shouldn't significantly effect game outcomes unless it is the Elder dragon.
It seems through nature of drakes being random, that the 3rd dragon of the same type really seals a game. A third infernal, ocean, or earth in a row really gives a team a very strong edge even if gold, kills, and turrets are even. But a team that has 3 dragons and all of them are different types does not.
I don't know how Riot could address this though. If the dragons were sequenced people would only fight for dragons that are advantageous to their comps and know when to fight for that dragon. If dragons were all the same in the game, then it could lead to less contestation over them if a particular comp draws no benefit from it.
Maybe the best way is that dragons are not completely random (1/4) chance of each dragon, but that the odds of getting the same dragon over and over is lowered after that dragon already appeared. I don't know if they already have that metric but if an infernal dragon appears first with a 1/4 roll, then the chance of an infernal appearing on the next roll should not reset to 1/4, but have a lower weight.
