I know it's just geeky fun, but consider: 1. Do you think there's a problem with you knowing it's always an A/B choice? That is: do you think the results would be different if you were ever given two same Pepsi? 2. What would happen if you added N cans of M brands? That is, let's say you get 7 extra cans and 2 will be RC Cola and the other ones will be some store brands or whatever. 3. Would it be a good test of caffeine hypothesis to tamper with a number of cans by adding a predetermined caffeine solution? If so, should you only tamper with one brand? 4. Can any of scenarios 1-3 be combined in a way that'd, to say intuitively, add information while reducing noise? Since apparently two is enough to even mention confidence: I bet that if we were playing Russian Roulette, you'd absolutely insist on letting you re-spin when going second despite the difference being noticeably less than 25%.with confidence of 75% (barely better than a 50% guess)