The answer to your question varies a bit, depending on your conditions. I'll give you some general thoughts and then point you at a couple of our articles that show some of the uses for some of the settings.
First, what trace interval should you use?
I normally run at 2.5s trace interval here. This is enough to get a basic profile of performance without collecting too much data. This will catch most outages that will disconnect you from a game (most games can weather a 1-2 second outage). There are certainly times to do more or less frequent sampling (more = 2.5 seconds isn't enough and you suspect there is some data point your'e missing, less = you don't have enough bandwidth and don't want to waste any more).
Next, what samples should you include? Normally, I run with "Samples to include" set to 100. But I change it often based on the data I'm looking at.to "zoom in" or out on a specific piece of data.
We discuss a few analysis scenarios in our tutorial that will give you an idea of how to use some of the settings. Several of them that highlight some of the analysis capabilities are:
Recognizing Bandwidth Saturation:http://www.pingplotter.com/tutorial/ScenarioSaturatedPipe.html
A scenario about bad hardware:http://www.pingplotter.com/tutorial/ScenarioCustomerNetworkProblem.html
When posting a picture, it should be a picture that shows both a good period and a bad period (if that exists). If you don't ever have a "good period", then it should show as much time as you can and still really focus on the problem. This changes drastically depending on the network conditions you're troubleshooting.
If you're getting spikes, you probably want to create an image that captures several spikes (3+ of them), along with periods where performance is OK. If this is a time-of-day based problem, then you'll want ot show enough data to demonstrate that the problem happens at that period on a couple of days, and not other times of day.
Hopefully, you find this helpful. PingPlotter is certainly an analysis tool that enables you to see the data in a way that's not easily possible with any other tool. It doesn't do tha analysis itself, though.