I missed a bunch of discussion.
Here's my thoughts.
Disclaimer: of course exceptions don't just exist, but abound in sentinels, we all know them, so i'm not covering them.
1. Card Play: card play is strongest when your play area is empty, and for most heroes, diminishes as you get more cards in play. Extra card play early is incredibly valuable, letting you set up faster, get in front of limited destruction, and generally get past the initial set up of most villains faster and in better shape. In my experience, the better you handle the initial villain set up and the faster you recover from destruction/reach the more likely you are to win, that is why in my value system, card play is the most powerful of the 3 actions.
2. Powers: powers always have value, because heroes have innate powers, but they tend to increase in value as heroes get more cards in play. They tend to be less valuable early, because innate powers tend to be less powerful than one shots, and rarely accelerate setup. Powers are stronger the more setup heroes have, and also when card play options diminish due to hand size.
3. Card Draw: card draw is most valuable when your hand is empty, and loses value as you get better cards in your hand. Card draw increases your options, but once you have the best option in hand adding more cards will not help your next card play, but only increase probability that you'll have the best card later. Card draw value reaches 0 once you have access to all your options.
Card draw is also more complex, because it does nothing immediately, it has probability of improving later actions, but no direct impact. Team Leader is the best example of this. The card draw is fantastic, it makes the team almost completely immune to hand denial, and gives everyone more options constantly. Once you have 10 or more cards in hand though, it isn't likely to improve your next card play so much that you would take more cards over using a power or playing a card. At that point (H) card draws aren't even worth a power use, let alone a card play. That's one reason Team Leader gets replaced by Research Grant eventually, because that discard becomes more valuable than every other hero getting a card (unless they can discard for damage on their own).
The value for each is dynamic, if your area is empty card play is best, if your hand is empty, card draw, if you are setup powers.
My work on valuing the health impact of each (from right after Shattered Timelines) ended with card play at roughly 3 points, powers at roughly 2 points, and card draw at roughly 1 point.
That means Team Leader is valued at (H) points, making it one of the stronger powers in the game. (Of course it isn't perfect, card draw is actually stronger in low (H) games, because increasing options for hero play from 12 to 13 is more valuable than increasing from 20 to 21, or to put it another way, the more hands and card plays your team has, the more likely one of you has a card to deal with the current situation, so digging for the right card drops in value)
But again, that valuation is from my own testing, which means it is biased by the data coming solely from games and experiments I was running. Im my experience early game strength and recovery from destruction are the most important factors in a win or loss, averaged across all villains.
Also, note that while I finished the process and came upbwith the 3-2-1 point valuation, the real answer is "it depends", because each situation in Sentinels is so unique. The chief value of the process has been the ability to look at innate powers in playtesting with a rough guide on how good they are.
A fun thing from that process: Impromptu Invention gives you a card draw (1), a card play (3), a search that hits the majority of valuable cards in your deck (this effect came out to around 5 points of value) and playing that card (3), for a total value of 12 points, which beats every other card in the game that isn't Wrathful Retribution (which came out around 17, and was a hard one to value), which was good, because I would have considered the process invalid if it hadn't.