In the game of StarCraft, a metagame describes a certain way of thinking about a particular matchup. A new player might treat any matchup (for simplicity's sake, we will work with ZvT as an example henceforth) in terms of what is happening in that individual game. He might have one build order regardless of opposing race, intending to transition only when he scouts an initial strategy from his opponent. The thought process exists somewhere along the lines of "If he makes this unit, I'll beat it with that unit." The best players in the world, though, do not do this. The game itself is nice enough to give you metadata (wouldn't the NSA be proud?) "Meta-" is a greek prefix that refers to exceeding beyond a given constraint. A metagame, then, is a game played in a mental realm outside of StarCraft. The moment a Zerg player finds out his opponent is Terran, he checks off certain things in his head from a list of strategies his opponent could do. A Terran most certainly won't be cannon-rushing you anytime soon. This metadata can give you a significant edge, though this is one of the most basic forms of metagaming.
Using advantages gained by playing on certain maps is a form of metagaming, as is becoming aware of matchup trends with the intention of exploiting them. Terran really liked their proxy two rax or SCV-pull all-ins early in Wings of Liberty. I would therefore make extremely safe choices (as opposed to greedy choices), knowing that the longer I could survive his onslaught the better likelihood for a Swarm victory.Tune into this episode of Crash Course for more information on recurring metagames
Since that time, Terran have taken greedier and greedier openers, oftentimes not beginning stimpack and infantry upgrades until 8 minutes or later. These builds delay the core units of a Terran bio army (marines [also, barracks and reactors to reinforce said marines], medivacs, et cetera) in exchange for harassment potential, an earlier third base, and map control.
It isn't that proxy two rax is any worse now than it was then. Marines don't do less damage, and bunkers don't cost more to make. No in-game changes really killed early cheese; expanding the size of the map pool, however, had a huge amount to do with it. After that initial BitByBit era of constant SCV pulls, 4gates and 10pools, economic play became the norm. Protoss switched into fast expand builds, Zerg started hatch firsting (and even sometimes two hatching before pool!) while Terran grabbed a second base before all-inning. The map size is important due to how long it takes to figure out where you opponent is. If your timing is supposed to arrive at 3 minutes, but you scout your opponent last on a four player map (~3:30), your all-in has failed before it ever had first contact with the enemy.
The most interesting part though is that the units, the buildings, the building prerequisites, everything is the same now as it was then (with a few minor changes for Heart of the Swarm). All that has changed are player trends, and reactions to player trends. Trends tend to be shaped by map pool (but are not enslaved to them!) This current map pool's longer rush distance makes the most desperate of all-ins (6 pools, worker rushes, et cetera) coin-flippy at best while still encouraging some level of contest for map control. King Sejong Station in particular comes to mind, with a wide open natural ramp and easy access from there into the main; too much static defense will easily be punished by running past the spines (or if resources are used to make extraneous buildings too early, a fast third can be taken by the Terran). Because the trends are predictable -- changing only with the map pool, or in some rare cases, with major innovations to the matchup following a patch or breakout tournament -- they can be anticipated and punished.
In StarCraft, metagaming is intricately related to a game design concept David Sirlin calls "yomi layers." Yomi, which refer to reading the mind of the opponent, are a model for situational decision-making. Any decent competitive game will allow you to counter your opponent if you know what he is going to do. What happens, though, when your enemy knows that you know what he will do? He needs a way to counter you. He's said to be on another level than you, or another "yomi layer." You knew his initial action (yomi layer 0), but he knew that you knew (yomi layer 1). What happens when you know that he knows that you know what he will do (yomi layer 3)? You'll need a way to counter his counter. And what happens when he knows that you know...?
For simplicity's sake, let's look at this concretely in the example of a fighting game before circling back around to StarCraft. Say you have this move you really like; let's call it "Fireball." You are perfectly content spitting fireballs at your opponent all the live-long day. Every time. This is yomi layer 0. Finally, your opponent, let's call her Sasha, sees what you're doing and uses her own really awesome move which hardcounters Fireball. Let's call her move "Flash." This is yomi layer 1.
What are you supposed to do now? Fireballs, no matter how much you like them, can't hold a candle to Flash. You need another option so you move in close for a Drop Kick -- exactly the type of move Flash is weak against. Now that you have this counter to Sasha's Flash, you can mix in fireballs again, because Sasha will be afraid of Drop Kick (yomi layer 2). Sasha has to find a counter to Drop Kick or you will be able to return to spamming Fireballs again. Luckily, Sasha has just the one -- Cartwheel (yomi layer 3). This completely terrifies you and stops you from countering Sasha's Flash.
Yomi layer 4 is actually yomi layer 0 because a full loop of yomi has been completed. There is no need to find the perfect counter to Sasha's Cartwheel. In the example above, Sasha can neutralize your Fireball (Y0) with Flash (Y1), but Sasha's Flash is negated by Drop Kick (Y2). Your Drop Kick is nullified by her Cartwheel (Y3), which is in turn nullified by your Fireball (Y0). As mentioned above, early in StarCraft's history, Terran liked to perform some pretty ridonculous cheese: bunker rushes and SCV pulls only the top of the list. Let's group all of these cheese and early strategies into one massive yomi (yomi 0). Zerg would turtle up, 14gas/14pool being most common before finally expanding (yomi 1), allowing some incredible safety and versatility against Terran's more coin-flippy antics. Terran would then learn to play more of a macro game, but the pendulum may have swung too far in that direction. By delaying marine production for ground-based map control, Terran reveal a weakness in lack of anti-air, and because of the gas expenditure of this greedy style, cuts into vespene gas for engineering bay and upgrades. No stim pack, no missile turrets, no reactors and no upgrades are a Mutalisk playground -- as long as said Mutalisks arrive before these things are viable (yomi 2).
There is no Yomi 3 yet developed for this build. That's not to say it doesn't exist in the game at present, simply that it hasn't been discovered yet. Metagames are constantly shifting, forever evolving. What works today because your opponent does X may not work tomorrow when your opponent does Y. And what happened yesterday when your opponent did W might not be as laughable tomorrow when you start doing Z in response to Y.
Welcome to Yomi. The possibilities are endless.
For more on this subject, please tune into the above-listed episode of Crash Course, where I analyze one such yomi build. Pool first instead of hatch first in ZvT -- how could that work?
Afterall, everyone knows Terran wall-offs prevent early pool builds from scouting or damaging an opponent, and the maps are way too big for Terran to cheese on... aren't they?