The adage is that only 1 in 10 startups succeed (which is likely not true — the real numbers are pretty much all over the place depending on what you measure and in which industry your startup is). A rule of thumb says that ideas have a 5% chance of success.
If those are your odds — your efforts should be focussed on increasing your chances. Most people do this by trying to shift the odds. The problem with that approach is that it is insanely hard to do, and your net gain is likely minimal (doubling your rate of success only gets you from 5% to 10%).
Rather than focusing on increasing the odds of a single idea, you are much better off trying more approaches where you maximize your rate of learning. The beauty of rate-based learning is that by the time you have attempted 20 ideas, your overall systems-based chance of success has increased to a whopping 64%! Moreover, this does not even include any gains you make by learning from approach to approach.
The trick to being successful with this approach is to dramatically minimize the time to try things. That means you are much better off getting into a habit of trying many things instead of trying to master one thing.
The question you want to ask yourself is how to speed up your experimentation by 10x. If something typically takes you a year to try, figure out how you can do it in a month; if it is a month do it in a week and if it is a week, do it in a day. Of course — you will come to a point where you cannot speed things up anymore, but until then:
Maximize your rate of learning by dramatically minimize the time to try things.
P.S. Huge thanks to my friend Tom Chi, who taught me this way of looking at the world.