no, its because everyone's complaining that "You have to wait for 3000 mutations per pixel to get a clean image...". Unfortunatly, 3000 mutations per pixel is a LOT of time. My interior scene is about 500,000 polygons. I got to 1500 mutations per pixel in 14 hours. With Photoshop's De-Noise filter, it almost looks clean. (almost) is a keyword.
Now, Path tracing, Bidirectional [path tracing], and Metropolis Light Transport are all rendering algorthims that use Monte Carlo aproximation (random samples, averaged. Enough samples and the average will converge to the right answer).
Path Tracing traces rays from the eye to the light (this is completly efficient and correct, since it is the EXACT mirror of what happens in real life). Unfortunatly, Path Tracing is bad with caustics, since the light has to be refracted throguh a refractive material or reflected off another into a light to work. That is why caustics with path tracing are very grainy.
Bidirectional Path Tracing traces a ray from the eye to the light, as well as one from the light to the eye. Thus, it can simulate caustics much better than stand-alone path tracing. It can be slower, though, since it does two algorithms instead of one at the same time.
Metropolis Light Transport is a very complicated algorithm that takes an existing path, made by bi-dir or single path tracing, and changes its direction randomly. Then, if this change is OK, (eg, it doesn't pass through a wall), then that new path is traced, and the result added to the average. This way, if we take a complicated situation, such as a light separated from a room by an infinite wall with a tiny hole, once the Metropolis algorithm finds this hole, it will expand and explore the hole, and the stuff behind the wall. This kind of situation is almost impossible for a Path tracer or a bidir-path tracer alone, since each new ray is completly random.
The MLT algorithm won't work very well if that infinite wall had 2 holes instead of one. It would explore one, and it wouldn't try to find the other hole, since it already found one way and needs to explore it. Thus, the lighting result would be incorrect, since the second hole would be [almost] ignored.
All of these algorithms are "Unbiased". The only incorectness in the estimation of the renering equation will show up as noise. Leave it on enough and the aestimate will converge to the right one.
There's your crash course in Rocket Science, Mirek!
), and I'm fighting it every day 
