TLDR answer for the ZWO 533:
Set your gain to 100. Set your offset to the minimum amount at which you no longer get any pixels with an ADU value of zero in your bias frames.
Non TLDR:
Speaking very loosely, gain is sort of analogous to ISO on a traditional camera. Higher gain = brighter image. You pick the gain based on looking at the gain response curves (provided with the ZWO documentation, available on any retailer's website). For the ZWO 533, the "high gain conversion mode" (a mode which gives you better dynamic range and low read noise) kicks in at gain = 100. So unless you have special circumstances, using a gain value of 100 is appropriate on the ZWO 533.
For offset: Increasing the offset essentially just shifts your histogram to the right by a fixed amount. Let me explain why this is desirable: Imagine that you take a very dark exposure (like a bias frame, perhaps). Your camera reads out the values in each pixel, but this measurement is affected by some amount of noise. Sometimes the noise makes the pixel brighter, sometimes darker. But if you're taking a dark image -- an image at which pixels are already at zero -- then a negative amount of noise just gets clipped to black. This messes up the noise profile in your image. Instead of a nice gaussian distribution, you get a clipped gaussian. (Err... is it a Poisson distribution? I don't remember. The exact distribution is unimportant). Stacking images with a clipped noise profile results in weird noise patterns. These look bad to the eye and confuse machine learning noise reduction techniques like NoiseX and DeepSNR, which are trained to expect a non-clipped noise pattern. So the "offset" setting shifts your histogram slightly to the right so that none of the darkest parts of your image are clipped to black.
