Sometime ago, I saw an advertisement of a render engine describing it as "fully unbiased". It seems that a lot of renderers are trying to make a statement these days about their quality by being unbiased (which always helps deliver accuracy without fighting with settings). But... "fully unbiased"? And is there accordingly "partially unbiased" as well? These terms simply do not exist, it's the same like saying fully and partially pregnant. Talented programmers were always fond of developing a ray tracer, but they occasionally suffer from lack of scientific background, or so it seems.

On the other hand, the unbiased approach is only a means of rendering (accurately) an image. The quality comes from other elements though; materials, textures, geometric complexity and of course user skills and artistic talent. It looks like that the unbiased term is rather abused in the context of an aggressive marketing strategy.

Even if a technique is theoretically unbiased, this does not mean that it will resolve to 100% correct results. I am going actually to insist here that there is no unbiased technique that will deliver you 100% accuracy in practice. What? you didn't know? That's right, there is no truly unbiased technique in practice. And this is due to one plain reason; finite calculation accuracy. No matter what you will use, your processor does not have enough bits to avoid adding a small bias there, negligible many times but important some other times. You may now understand why some unbiased renderers cannot resolve sun-pool caustics; yep, even if you leave them running for a year they will never converge to the right result!

So, perhaps it's time to ask the right question, which sounds like an oxymoron at first; could we say something about unbiased renderers according to their practical bias? Yeah, we could! But unfortunately, this requires some very complicated analysis on the technique and its custom implementation. Nevertheless, in most cases, the general methodology used plays the most vital role in characterizing the accuracy of the technique. And for all those praising a path tracer as "fully unbiased", the fact is that path tracing is the method with the worst practical accuracy and robustness when resolving global illumination.