Now that you've been given the answer, let me expand it a bit.
The first thing is: What do you mean by "a real computer"? Not only is a Pi (any model) a "real computer" (in the sense of "a general purpose digital electronic computer"), but it's a lot faster--and therefore, I suppose, more "real"--than many a minicomputer or mainframe I've worked on.
The second thing is: When you say that you want to play games on the computer, you need to specify what sort of computer they are designed to run on. Most games are designed and compiled to run on either a "PC" running some form of Microsoft Windows or some variety of pre-defined console. Games--like most programs--specify a hardware architecture they are compiled to run on. PC games run on the Intel (or workalike) "x86" architecture. Different consoles use chips with a variety of hardware architectures, with each model of console being nearly unique, at least across families. The Pi uses ARM CPUs, aka the ARM architecture. The result of this is that if you want to run a program compiled for one architecture on a different architecture, you have to use a simulator program to fake at least the underlying hardware and usually the operating environment as well. The problem here is that simulators are slow. The way around that is to run the simulator on hardware that is much, much faster than the targeted hardware. thus, you can run games (or other programs) that are meant to run on an MS-DOS system on a Pi because the Pi is far faster than any machine that was intended to run MS-DOS.
The third thing is: There are ways to combine multiple, discrete computers (of any type, not just Pis) to get more processing power. This is the essence of a technique called "clustering". You can read the gory details by looking up "Beowulf Cluster". There are two fundamental issues with this approach. The first is that, in order to be effective, you have to trying to solve a problem that can have various parts run in parallel. Most games, and particularly older games, aren't written that way, so there is no gain to be had. The second problem is that the efficiency of parallel computing is heavily dependent on the connection speeds between the various parts. Here, the Pi is not a terribly good choice becuase it doesn't have any really fast interfaces. Still...building a cluster of Pis can be worthwhile. One wouldn't use it to run games (though you *could* write a game to use one), but--rather--it can be used as a very inexpensive way to learn about parallel computing and how to do it.
Bear in mind that if you just want to learn a bit about parallel computing, you can use a single Pi. This is because all Pis other than the Pi0/Pi0W, A+, B+, and CM have 4 CPU cores and that will allow you to write a program that uses all 4 cores at the same time. (And if you use a single core model...the operating system will make it *look* like you have multiple processes running in parallel...just more slowly.)