That being said, I find that they are quite the leveler; high tempo decks with less lands suffer less than a control deck, but need mulligans more often. It's a trade-off, and evens out the playing field more than people realize.
Even with mulligans, I'd argue that mana screw is one of the biggest failings of M:tG. Make your best estimate of what percentage of games you think are decided by mana screw or mulligan-reduced hands, then ask: "Is that too high?"
In my experience, it was about 1/3. Then again, I wasn't a great player.
In 2003, I came up with a Magic-like game that solved a lot of these problems. It wasn't a kill-your-opponent game, but a German-style TCG with a Victory Point (VP) system, making it more amenable to multiplayer games. There were no "life points", but you had Action and Focus points that you could use to draw extra cards or resources, to give creatures an advantage in combat (making it more nondeterministic), etc. The number of strategic options per turn was vastly greater than in Magic, so it was harder for a player to get "stuck" and unable to do something useful.
Unfortunately, at the time I knew nothing about programming, nor had any money (I was 20). It was pretty obvious that the only way the game had a shot would be to take it online. Now I could hack it up, but I've forgotten most of the mechanics.
The disadvantage of this game, and why I think it wouldn't have ever taken off the way M:tG did is that, despite having "better" design from a Euro-style perspective, it wasn't a game that would finish within 30 minutes. A turn involved multiple players (most phases of turns were concurrent) and had a lot more strategic options than Magic's "untap, draw, cast, attack" formula that describes 90% of turns, so a single turn would last 3-10 minutes.
It's nice. It reminds me of the "square root" hack for Gin Rummy, from the guy who came up with the card game Ambition.
Square-root hack is as follows: regular rounds are scored on the square root (rounding down) of the deadwood difference, capped at 5 points. Undercuts get a bonus of 3, plus the square root of the differential. Gin scores 4 plus the square root of the opponent's deadwood. So the ranges are:
Regular: 1-5
Undercut: 3-6
Gin: 5-13
The victory condition is to have win a round when you have at least 20 points, and you are in the lead.
Second-turn Gin is extremely rare, but first- or second-turn knock is not rare. With the regular system, a second-turn knock (pure luck) can be a 60+ round, and that's more than half the game. With this system, it's only 5 points.
If only someone could do something similar for mana drought/flood, Magic would be a pretty good game.