Looking for statistics help

Hi all - I'm a new member of the forum - I'm looking for statistics help. A friend of mine is convinced he can dowse water and wants to test his ability. I agreed to help him and I want to do a good job. I'm thinking of having 6 identical new buckets, six numbered spots in a row or circle (previously dowsed as having no interference), one bucket will have water. I'll roll a die and record the number, move the water-bucket to the designated location, move all the buckets slightly so he can't tell which was moved, retreat from sight, let him dowse and record which spot has the water. Then he retreats from sight... repeat the trial.

My question is - how many trials should I run to get sufficient results? Something that would suggest to you-all, who are familiar with these sorts of trials, that there's a real effect and not just a couple random correct guesses?

Thanks for the help! :)

via International Skeptics Forum http://ift.tt/1JQWgJA