When poker players refer to "meta game" they generally refer to taking a course of action in a hand, different from what they would have done if the hand had been played in a vacuum. A good example of a meta game play is to make a loose call on the river, a call you wouldn't normally make, in order to create the impression on the other player that you will make "bad calls" on the river. This, then, might stop him from trying to bluff you in future hands.
Or, you might bet a weak hand on the river, for much the same reason but in reverse; if he calls, he might note that you bet weak hands on the river and in the future be much more inclined to call you down light. As you can see, in order to call it a "meta game play" it must be something that you do once (or maybe a few times) and then change how you play. If you always make loose calls on the river, your opponent will adjust correctly and only bet his stronger hands, and you will lose money - both now and later. The idea is to make a slightly "bad" play and then take advantage of how you opponent adjusts to it.
The word "meta", in this sense, means "beyond" or "trancending". It's not, then, about how you play your hand, it's about how you play your session. You might start out loose and aggressive in order to create the image of a maniac, only to switch gears and turn into a rock. By the time your opponents catch on, it may be too late. It's a fascinating aspect of poker, but it's also fairly unimportant, in the larger scheme of things. If you're going to successfully employ a meta game play, three conditions need to be met:
Being able to get into your opponent's head and decide that "now, he should view me as a maniac" is not easy. Online, he might be multitabling and not be paying attention to any specific table very much at all. I presume it's easier live, but still, you need to trust that he notices what you've done. Furthermore, you need to trust that he will adjust the way you want him to, which is an easy trap to fall in. For instance:
You make a bluff on the river, and he calls you. Now, you think to yourself, I can start value betting the river a lot more because he will call me down light. So now you bet the river with two pair, and suddenly he raises big. If he thinks you're bluffing, calling would have made sense, but he raised, so he must have a really big hand, you think, and fold. Of course, this is only good if you were right about how he adjusted. Maybe he took the bluff personally and decided to fight fire with fire, and you've now turned a reliable opponent into someone who will make big bluffs on the river - but it might take you awhile to figure that out and re-adjust yourself!
Your opponents are not you. Don't make the mistake of thinking that they think and act the same way you do.
And what if you have a really savvy opponent? One that has played with you before and realizes that what you did is not really up to your usual standards. Such an opponent would perhaps not fall for the bait, but see your play for what it really is. And suddenly, he's ahead of you, anticipating that the next time you're in the same situation you will actually make the opposite play and he will read you perfectly. These opponents are very rare, but they will cost you dearly.
And the last point is not to be ignored. You need to have reason to believe that you can make back the money that you lost on your meta game investment later in the session (or in a later session, theoretically). This won't work if your opponent is likely to get up and leave sometime in the next few hands, or if the adjustment that you want him to make isn't bad enough that you won't be able to make the investment back even if the session lasts for a long time.
If it sounds like I'm recommending you stay away from meta game plays, then it sounds the way I intend for it to sound. Avoid them, unless you're very sure of what you're doing. However, there's a special kind of "meta game"-like plays that I do suggest you employ; the so-called "balancing plays."
A balancing play is when you occasionally do something different than you normally would in a given situation. For instance, you might limp/re-raise with pocket aces rather than just open-raise. Or you might re-raise with pocket sevens, despite that you'd normally just cold-call. The difference between a meta game play and a balancing play is that with balancing plays, you actually expect to make money immediately, not just on future hands. A meta game play, by my definition at least, is a calculated cost that you will make up for in future hands. A balancing play is a play that has immediate positive expected value.
Let's look at the situation where you 3-bet preflop with pocket deuces on the button, after a regular player has opened in middle position. If you always 3-bet every pocket pair in position on this player, he will - or at least may - adjust so that he will come over the top more often, or bluff you out on ace-high flops. But if you 3-bet with deuces only some of the time, you're making it hard for your opponent to adjust. Ideally, you'd 3-bet with weaker hands with such a frequency that you're actually making it theoretically impossible for him to adjust, but deciding what that frequency is, is difficult away from the table and near impossible while actually playing. "Occasionally," "some of the time," and "rarely" is how these things are usually phrased.
But the key feature of a balancing play, that sets it apart from a meta game play, is not that it's an "unusual play," it's how you want your opponent to react to it. With a meta game play, you're intentionally trying to get your opponent to play a certain way. When you employ a balancing play, you're trying to keep your opponent guessing. Adding an element of randomness to how you play, in other words, is not something you do for awhile, and then stop when you think your opponent has noticed what's going on. You balance your plays continuously.
Even here, however, you should be cautious. Don't go overboard with 3-betting small suited connectors or just limping with aces. You're adding deception to your poker game by being tricky, but that deception is useless unless you actually use it to cloak your actual game. If limp/re-raising becomes your most common way of playing aces, your opponents will learn from it, and rule out aces whenever you open-raise - something you surely do NOT want them to be able to do.
In short, develop a fundamentally solid game. From that platform, spice it up with some balancing plays. And generally speaking, stay away from attempts at meta game plays, unless you have a very good idea of what you are - and your opponent is! - doing.