Get one with as high a distance to spot ratio as you can; i.e 10:1 is better than 1:1. This determines the size of the measuring area compared to the distance from the surface being measured. If you get one with a 1:1 ratio (usually the cheapest), you would have to hold it against the pipe to get an accurate reading. This is because it averages the temperature over the field of view.
Accuracy is about the same for all devices and, as you are measuring temperature differences, it's relative temperatures, not absolute that you are interested in.
Because these devices measure the heat emitted from the surface, the temperature reading will depend on the type of surface. They are calibrated for a black surface, so if you try to measure a shiny copper pipe it will give a lower temperature reading. The way to get round this is to wrap a turn of black insulating tape round all pipes and measure off that.