Losing Weight

Was just wondering about losing weight. I saw a clip on youtube yesterday before going to bed, there was Bill Burr being interviewed by Joe Rogan. Rogan is a bit a sketchy character in my opinion. Anyway Burr was making comments about people not bothering to go to the gym and lose 10 to 15 pounds. So I started thinking well is it actually easier to lose 10 to 15 pounds by just going to the gym, or by just changing your diet? So then I started wondering how to quantify these effects. For an example would be good to drink green tea every morning to lose weight, and if so, how effective is it? This involves things like caffeine increasing metabolism a bit or rate of it, and it resulting to increased production of heat within the body, and so it's called thermogenic. Basically if you drink some green tea, your core will spend a tiny bit of energy on it's own for no reason. But how big is that effect? Like if you drink this cup of green tea for 6 months, can you expect to lose a pound of weight? So from there I was just thinking about what would be the most effective way to statistically test the varying efficacies of different methods of losing weight, and also if there would be some correlations. So if you have a small group test volunteers then you could randomly like assign them different tasks like 50% of the group will have green tea every morning. Then some other random part of the group, like 50%, would also start jogging. Etc. But then I started thinking that well isn't actually pointless to do it randomly that way? Like you could just come up with these sieve for how you distribute multiple different variables within a group, like a for an example, let's say you wanted to distribute properties A,B,C into a sample of n people. You could just pick a representative sample, assign them to numbers, then juts randomize the number each test subjects gets. So you don't get negative effects of randomness