What's the right statistical test for this?

I have 100 urns. Each urn has 1000 balls. The balls in the urns are labeled with a value, where the value is a quantitative measurement. An experiment has identified a selection of 100 balls, one from each urn, that are of particular interest to the experimenter. Are the values of the 100 balls I selected significantly greater than the values of the urns from which the balls came?

I'm not sure how to do this but I suspect some type of paired statistics are useful, or some type of combined p-value approach (Stouffer's test?) might be useful.

Thanks!