When studies are conducted to determine whether Drug X cures Disease Y, success is declared if the link is proven "at the 0.05 significance level." In layman's terms this means there is less than a 1 in 20 chance that the result was caused by insufficient data. Think of it this way: if you see a bunch of nurses one day and they are all female, you might conclude that all nurses are female. But in reality, there are male nurses; you just didn't happen to see one that day. You were fooled by the smallness of your sample size.
An amateur faker makes up data to support a hypothesis. We have a cute name for this: "dry-labbing." An amateur also may think up bogus reasons for throwing out data that don't support the hypothesis.
A real professional faker with a big research budget does lots of studies and publishes the one that by chance showed a positive result at the 0.05 significance level. 20 is the average number of studies needed, assuming there is no correlation between administering the drug and curing the disease. It might take 10 or it might take 40 depending on how lucky you are. Even 40 studies are a small price to pay when billions of dollars are at stake.