Card to player ratios

That's a toughy without knowing more about the game.

Some questions: Is there a central deck(s) that all players draw from, or does each player get their own deck? Is there hand management and/or cards-on-table management (such as rummy games and most CCGs use)? Will there be discards? If so, will the discards get re-shuffled and used again later in the game (self-refreshing draw deck)?

Based on those answers, you'll want to make sure you have enough total cards so that players neither feel like their always too lean on cards or mostly swamped in cards. They should feel like they have enough cards-in-hand to plan ahead, and able to draw enough to possibly fill-in any gaps and possibly get some nice surprises/gifts while drawing ... but still be able to do "just the right amount" of card-handling to keep the game's tension where you want it to be.

How does scoring work? How powerful are the rares compared to the commons?

These answers should help steer you toward the mix you want to have.

For example, in a game I'm currently working on, players end up grabbing "resource cards" through various means, which come in several different types. Some types score more points per card -- but there are fewer of those cards. Some types don't score well early on, but build-up to provide better scoring as you gain more of them ... and they are more common in the deck.

I'm trying to weight the scoring available through the various types at both the mid-point of each type (What score would a player get if they gathered half of the available cards in this type?) and also at the far extreme (What score would a player get as they near collecting all of this type?)

I'm using the comparisons between the mid-point and extreme as a way to balance things out. In a very simple example, I might have 6 cards available in one type that each score 5 points and 10 cards available in another type that each score 3 points. In both cases, the mid-point score would be 15 and the extreme score would be 30.

Hopefully those questions and examples will help spark an approach for you to take.

-Bryk

Brykovian has the right of it, but I'll throw in my two cents anyway.

Ask yourself how likely you want a card to be drawn from a complete deck. This tells you how many cards (of that type) should be in the deck, relative to the total number of cards in the deck.

Suppose you have a deck composed of brick, sheep, grain, stone, and wood. You want players to (on average) have 4 brick, 3 sheep, 2 grain, 1 stone, and 0 wood in an average hand of 10 cards. Now you know your ratios. You'll have 39% brick, 29% sheep, 19% grain, 9% stone, and 4% wood. 39+29+19+9+4 = 100

You're designing the game for four players. You expect them to start with 10 cards each and draw 2 cards per turn (on average). You don't want to make them shuffle the deck too often, so you decide to have a 104-card deck. (52 would be too small, and 104 is the next multiple of 52.)

Now that you know how many cards in the deck, it's easy to know how many of each card there are. Since 100 is so close to 104, I'm not going to even bother with multiplying the number of cards into each percentage. Instead, I'm going to shunt the extra four cards onto the first four commodities: 40 brick, 30 sheep, 20 grain, 10 stone, 4 wood.

The point is that my design goals drove my decisions. I thought about what I wanted the average player to have access to in an average hand, then designed the deck to support that design goal. Think about what you want an average player to hold, then design with that end in mind.

A bit off-topic, but this comment caught my eye:

Do early playtesting by yourself. I call this the "teddy bear tea party" approach: all the other players are teddy bears, and you must pretend to be each one of them in turn and decide what each one does as you play. This will help you find your card balances and iron out any big problems in your mechanics, deck composition, and rules.

You will need live playtesters eventually. They will show you things about your game that you will never discover on your own. I use teddy bear tea parties until a design seem stable and playable (that is, until I can find no more big changes to make, and the latest rules seem to make an okay game), and then start dragging in live playtesters. This approach is faster than waiting to do all your playtests with live people, and is easier on your playtesters since they won't have to suffer through too many new-and-untried designs.