Actually a surprisingly complex question considering that "coinshooting" is such a fundamental aspect of metal detecting. As for all the variables? Got it, been there with almost all the detectors, blah blah. Besides the ergonomic aspects, (weight, balance, feel, size of important display characters, etc), and varied and good coil selection, I think one of the main things required to make a good coinshooter is resolution in all the normal coins of a particular nationality. In other words, a coinshooter detector should more often than not afford the operator a high degree of confidence that a target is a copper penny instead of a clad dime, a typical wheat instead of a copper memorial, a nickle instead of a rolled beaver tail or old ring tab, any silver coin denomination over it's clad counterpart, etc. Depth can be a factor in making those same "calls" in the field, but anyone that has much experience knows that relative target depth is not necessarily ultimately defining. Historically, there is usually one of those segment areas that is more guess than by golly. Next would be relative ease of operation vs max capability, IE, how much do you have to know in order to wring out "the best" of the ability of a certain detector?. Next would be relative ease of stable operation and its relativity to max performance. IE, if running stable at location "x", are you loosing any or much depth and or separation ability? Last but not least, hopefully it should be able to, more often than not, "SEEM" to find more when used behind other detectors. Yeah, I know that is tough to quantify but (experienced) user confidence/warm and fuzziness is pretty darn important to achieve huge sales figures. A detector that can not necessarily excel in all (still waiting for one of those) but at least be competently effective in all would surely be a formidable coin shooter in my mind.