Too few numbers to choose from guarantees that someone will guess at the right answer. As has been suggested before, using even the numbers 1 to 9 would make a "right" answer far more unlikely, and therefore hits more significant. If you increase this to 1 to 1000, a correct answer by chance is almost impossible with your small amount of test subjects. I would suggest using a truly blind group of numbers, say for example, a two digit number not smaller than 11 or larger than 99, chosen from a group of 50 contiguous numbers from the set, with the subset of 50 starting at a random number. 36 to 86 for example. Without telling the test subjects more than this, they would have to be receiving your telepathy to have a correct number.