How to calc how far a 50% event is off

[adsxvii]

Hi wasn't sure if this would get an answer here, but it can't hurt to try.
I use lots of sports stats to calculate and predict results. Well all goes well but when I look for new things to predict, I am not really sure that I use the data my stats return in the best way. For instance lets say I run a particular theory over which I get 5000 results of a 50% event, and my data shows me a return of 54%. Making my theory correct to predict an event by 4%. I am sure at 100 samples my theory would be inconclusive. Due to the lack of data.
So my question is, in an event that is 50/50 how many samples would I need to make my theory somewhat conclusive. Like for example at 15000 samples at 54% where I need a 52% result to be conclusive. Is there a formula that would tell me if this would be enough samples to where there would be less than a 2% error. Thanks in advance Andy

[robert111]

If any event can have an outcome of either pass or fail, and this is truly random, then by definition you cannot predict it at all. This is the theory of variation. You might devise a theory and start to make predictions, and find you have been right 55% of the time, but that is because the answers don't pop out pass fail pass fail pass fail pass fail. It is purely a coincidence that your guess matched the outcome in that time period.

[adsxvii]

I think my post was misleading. Ill try again.
Lets say I flip a fair coin 100 times, the variation (not sure if variation is the correct word) has to be higher from 50% thats probable, than if I flipped the same coin 10,000 times. My question is how do I calculate how much it may be off depending on my sample size. I'm pretty sure that if I flipped a coin 100,000 times and came up with a 54% heads ratio the event would not be considered as a 50/50 prop. Any help is appreciated. Thanks Andy.