I've made a deckbuilding game, Shadowrun: Crossfire, which will become available for purchase sometime in 2013 (sorry, no official date yet). In doing so, I was reading up on deckbuilding games in general, and I found this lovely blog article.
In that blog, the author begins to discuss the basics of deckbuilding games. Naturally this includes commenting on starting the game with a hand of 5 or 6 cards. They note that there is little variety from these numbers, and briefly wonder why. I'll tell you why.
Sometimes numbers in games are just right. The reason almost all deckbuilding games give you 5 or 6 cards is that more or less makes for a less fun game. Sure, maybe all the designers of those games are just lazy, and went with the "industry standard" number of cards, but if the rest of their game is any good it's more likely they investigated the hand size, tried various numbers, and concluded for themselves that 5-ish was the right number.
Fewer cards would mean fewer choices of what to do each turn.
More cards would mean analysis paralysis in how to play your hand, increased variance in hand power levels, and hands that are too powerful for the desired pace of the game. You'd need bigger starting decks to compensate, and at that point you might as well just divide by two and be back at 5 or 6.
Later on in the same blog, the author goes on to question why cards are purchased. It's great to question such things, but pretty weak to just ask why and not think about it even a single step further. Why purchase cards instead of letting the players choose any one card at the end of their turn? Well, I tried that early in my design of Shadowrun: Crossfire, and I can tell you why it didn't work out.
First, there were some good things about "one free card per turn." For one, you are more guaranteed that players' decks are always making progress; their decks are improving turn over turn, and they never have a "dead" turn (they always at least gain a new card). Also, if players gain a free card each turn, the main abilities of those cards can be focused on something other than purchasing the next card. This will make for a very different feeling game, and open the doors to more options for you as a designer.
The drawbacks to "one free card per turn" are quite large. First, a parallel to one of the positives: you've got to think of something for the cards to do besides help acquire more cards. You're forcing yourself into new design space, and cutting out a major category of card functions. We managed to do this in Shadowrun: Crossfire by putting money somewhere else. Your cards don't make money, but you use your cards to defeat things, and those things are full of money.
But why money at all? Without money, all of your cards have the same cost. This is the second drawback to "one free card per turn." Turns out this situation is awful. Or at least, awfully hard to design around. If all cards have the same cost, they must have the same power level. (Or else, if they have different power levels, players will all buy the best one first, and all players will end up with the same deck - or at least all the players who know what they are doing will.) Let's say you solve that - you could have a centralized, randomized market of some number of cards (again, the math says 6 is pretty good), like Ascension, Legendary, and Shadowrun: Crossfire all use. This would mean players can only choose the best card of the market, so that would fix it, right? Nope! This dramatically increases the luck factor in your game. The players on whose turn the best card appears in the market will have a tremendous advantage. The player who goes first will see the most "new" cards to choose from and then be very likely to be the most advantaged. They'll get a powerful card, that is randomly replaced by (chances are) a weaker card. Now the next player takes the next best card. The two or three weakest cards in the market will never get taken, and your market, over the course of the game, naturally reduces down to just 1 card slot. In my attempt at this, I tried to fix that by clearing the market every time. An early prototype had players flipping 5 cards at the end of their turn, taking one, and then the other players drafted the rest. With 4 players there was always 1 left over, so you had some choice but... It was so-so-so-soooo random. You couldn't have any kind of strategy in what you were doing, and you still always took the most powerful card. You had no real choice and thus the lack of player autonomy was severe.
Without a cost system, the weaker cards always look far worse than the stronger cards. (Don't get me started on why you can't make all the cards "equally powerful" - it's a natural fact of reality. If you can't understand that, try making the assumption that it IS true, and try to figure out WHY it is true.)
The cost system helps fix this for you. You can make weaker cards cheaper, and now players are evaluating them not just on their abilities, but their costs. (Yes, there will still be a "most powerful" card at each cost, but you open up a lot of design space for yourself this way. The other axis, not discussed here, is situational power, which, of course, you are also trying to achieve with narrow cards, but in a deckbuilding game narrow cards naturally aren't useful in every game, so they only partly solve the problem.)
So to sum up: many times numbers are established in games and used in many games of that genre because they are the naturally "best" numbers for their application. This can also be true for entire mechanics. However! It doesn't mean you shouldn't try to do something different. Just be aware of how you are spending your time. You can only afford so many wild goose chases during design before you need to settle down and worry about finishing the product. Chase the right goose.
Oh, and be sure to buy Shadowrun: Crossfire when it comes out in the next, um, sometime this year. We got a lot of new and interesting mechanics into it because most of our goose chases ended with a golden goose.