've often made oblique references to certain types of quotas that a deck wants to have, implying that there is some kind of ideal composition of a deck. There is nothing definitive about this; certain decks can win successfully without adhering to these guidelines, like the Dampen Thought deck in Champions of Kamigawa, but they are good guidelines to bear in mind whilst drafting and during deck construction. I want to reinforce that these are helpful suggestions, and far from solid rules.
The easiest quota to examine is how many land to play. I often hear some formats being referred to as sixteen- or eighteen-land formats. This is a good guideline, but the truth is that it is always deck-dependent, altering based on how mana-intensive a deck is. Mirrodin Block brought us mana-producing Myrs, Talismans, and cheap cycling Spellbombs that allowed us to cheat on land and in some extreme cases run as few as fourteen. Ravnica Block gave us Signets and bounce lands that again allowed us to cut land, often meaning we need only run sixteen, even in three-colour decks. Time Spiral Block, with the common exception of Prismatic Lens, only has acceleration in green, where almost all of its mana fixing also lies. This generates the effect that most decks are two colours and tend to run the standard seventeen or eighteen land.
What factors influence your decision to run a specific number of land? Sometimes it will be a poor mana base. In a bad three-colour deck light on fixers, the splash colour will usually demand at least three sources. Once these are taken into account, there are only fourteen or fifteen land slots left to be allocated. These can be divided equally between the two other colours, but it is often the case that one colour has more cards than the other or that one is greedier with pump creatures and double casting cost spells, demanding that you play eighteen land and allocate an eight/seven split.
Other times, a two-colour deck will have a combination of double-cost spells in both colours and several high-end spells. This is an easy example of an eighteen-land deck. In contrast, a deck with one colour heavier than the other, with an early, aggressive curve and all single coloured mana cards is an easy seventeen land deck. It is the murky shades of grey between these that cause difficulty. Some decks will be so full of power cards and card advantage that the only way they will lose is if they miss land drops, and so want eighteen and, in rare cases, nineteen. Generally, if both colours are heavy or if the twenty-third card that would make the deck is poor, then play eighteen land. If a deck has next to no need to make its sixth land drop, run seventeen.
Other influencing factors are non-basic land. If you want to run a colourless land, like Urza's Factory, you can only really afford to do it as your eighteenth land if your two colours are both very non-committal. A land that cleanly fixes both of your colours, like a River of Tears, will allow you to cheat on land based on colour commitment, as opposed to needing lots of land. Semi-fixers, like Calciform Pools, can only really cheat on land if the deck doesn't have a pressing need for coloured mana, and will otherwise be treated as a pseudo basic land or the main colour, with a bonus.
It is not so much the number of land that the format impacts, but rather the number of mana sources available to a deck. A solid main colour will want seven to nine sources, depending on how heavy it is, and a splash will tend to need three or four. You will often arrive at the correct land count as a result of totalling the ideal number of sources for each colour and then shaving these until you reach the correct number of sources and, hence, the correct amount of land.
Creatures are the next most important quotient. You need to win somehow after all. In general, a deck wants somewhere between twelve and fifteen creatures, just so that it can make room for the other card types. Decks can also win with seven creatures (if most of them are rares, or if the deck has a ton of removal/card advantage as it can then find the time to draw into them) or twenty (white-green decks are a prime example, with an aggressive curve, often backed up by just combat tricks), but I am speaking in generalities.
As a rule of thumb, it is your creatures that will be doing the most work, so you need them to bear the brunt of the effort so as not to force your removal or tricks to labour overtime. For example, in the early game the difference between removal and a creature is fairly insignificant, as most early creatures will trade with their opposition as well as a Flowstone Embrace or equivalent might. It is a shame to waste your removal trying to stay alive this early when you can just use the more easily acquired, and less highly picked, dorks to the same end.
In this regard, it is also important, as part of your creature quota, to pay attention to your mana curve. You will tend to want, give or take, three two-drops, five three-drops, four four-drops, two five-drops, and two six-drops. Two-drops can often fill in for three-drops, and so on up the scale. All these numbers vary depending on how defensive/aggressive your deck is. If you are very aggressive, then you will want around five two-drops with a curve ending with about two five-drops. If you have a lot of very defensive early creatures, like Thallid Shell-Dweller, you can possible skimp on three-drops in favour of more five-drops as you will be more likely to survive to the late game, and so on. Whilst drafting you will have an idea of how your deck is shaping up. Be sure to keep an eye out on its direction so you know whether or not to focus on early defensive picks if you are in need and shy of them, or whether to ignore them completely in favour of things that attack.
Tricks make up the rest of a deck with the exception of two types of cards: card advantage and acceleration. Despite card advantage being Limited's premium, there are on average only two or three slots available for cards pure card draw like Foresee or Think Twice. This is a budget that I have little problem stretching, but it should be noted that your card draw needs to get you somewhere. You do eventually have to draw removal for their threats and creatures as your own.
Acceleration and its partner in crime, fixers, have a similar-sized quotient to card advantage. The amount varies on how many colours you are playing and will therefore have to gain access to, and how tempo-based the format is-in Mirrodin, everyone was accelerating, meaning you had to as well. The main problem about most acceleration is that it sucks to draw them in the late game (Edge of Autumn is a welcome exception). This means that you want to play enough so that you have a reliable chance to draw one early enough for it to count, but not enough so that you get mana flooded in the late game; this means that acceleration has the bizarre quirk of almost being considered part of the land quota. You almost never want to play more than nineteen mana sources within your deck.
Tricks tend to be divided into three categories: removal, pump, and finishers. Some colours and archetypes use pump as poor removal that compensates by having a better tempo edge as they tend to cost less, whereas as others, until recently (in most cases), did not have access to pump so played more removal. In general, it is better to play a removal spell over a creature spell. This obviously cannot be taken to extremes as there are varying strengths of both types and there is a limit to how few creatures you can play. There is also a limit to how much pump you can play as they lose effect the more you have, as your opponent is even more likely to play around them in the second and third games. I would advocate around three pump spells to be a good number and almost any amount of solid removal. Finishers, as the name suggests, tend to only be needed in singles, so you do not really want more than one in a deck, although exceptions can be made for really powerful ones like Tromp the Domains.
One of the most crucial things to be aware of whilst drafting is how your deck will win. It may sound obvious, but you need win conditions. You can have a deck full of cool cards, but no reliable way to win. Many white-black decks in triple Time Spiral looked like this, where the player would sacrifice a family member to have a creature with power three or greater. Win conditions are finishers, mass removal effects, X burn spells, bombs, fatties, mill effects, and creatures with removal. It's hard to say how many you need to have in a deck, but you definitely need around five good ways to win a game. One thing to note about evasion creatures is how fragile they may be. Spirit en-Dal and Gossamer Phantasm are easily removed and cannot really be relied upon, while Castle Raptors, Corpulent Corpse, and Errant Ephemeron will almost always get the job done. You will often see pros filling their decks with massively overcosted fatties, like Phantasmagorian, simply because they need to have some definitive win conditions that will get the job done.
All of these concepts need to be carefully juggled to draft and build a good deck. Getting the right balance often comes down to experience within a format by being able to recognise what holes your deck has and what ways to best shore them up. This needs to be done whilst drafting, to recognise the flaws and pick cards to compensate for them; and during deck construction if there was something you missed earlier, and are now forced to play normally sub-par cards make up for it. Now for a few concrete examples.
This is a great example of a focussed, aggressive deck. It is classic for the red-black archetype, filled with just men and removal. It is lacking a little in quality two-drops, but the nearly endless removal means that, as an exception to the rule, it doesn't begrudge using them on the opponent's early men. Its curve ends early, because it wants to be racing and will spend the last few turns killing men rather than asking questions with its own fatties. The Urza's Factory is worth a special mention. This is one of the times where running it with sixteen other land is great because the deck's mana is so good; it helps to provide a late game without taking up a slot.
Not all my examples will be of awesomely powerful decks, but this one illustrates the importance of a mana curve. This deck is chock-a-block with bombs and removal but is distinctly lacking in both early drops and defence-I would've killed for a Mogg War Marshal! As a result, almost every game I played with it, especially those where I was going second, had me on the back foot, forced to make bad trades and failing to get maximum use out of cards as I had to waste good removal on poor targets just to stay alive until the raw power kicked in. This is also a great example of a deck which has no mana issues and yet has to play eighteen land, because it will lose if it stumbles on mana due to the lack of defence and it has enough power to compensate for most flooding that might occur. It wants to make a land for the first seven turns, so much so that I almost cut the Slipstream Serpent for a nineteenth land.
White-green is often an archetype which struggles to map out well. This is because evasion, removal, and pump are hard to find, meaning you will often end up with a deck that looks like this. There are several things that you can do about it, though. This kind of deck needs to be very tempo-oriented, meaning it wants lots of early drops and acceleration. Uktabi Drake is also an ace up your sleeve as you play it whenever you have a spare mana and it swings the tempo massively in your favour whilst also providing some much-needed evasion. This deck needed some help, as can be seen by the poor Lumithread Fields trying to help a bad mana curve-there are too many four-drops and not enough early threats, which places a lot of premium on the Edge of Autumns.
Here is a deck where it all went wrong. I took over the draft from a friend and had no idea which colours were open, as I had missed most of the first pack and, hence, the signals. The curve is horrible; there are almost no two-drops, far too many 6+ cards, and no acceleration to get there. On top of all that, there's a third colour thrown in for good measure and the deck lacks both card advantage and removal. What the deck does have is bombs. Lots of them. I seem to spend my life explaining to non-players and rookies alike that Limited is a skilled game, despite the fact that you randomly open packs. People place too much emphasis on opening bombs. This deck did that, but it is an awful deck. Bombs do not make a deck-consistency, removal, and a curve do. Sure, this deck won games thanks to it having bombs, but it is definitely a 1-2 deck at best.
Before I finish, remember that, as with everything else in Magic, these quotas are subjective.