In all the books i could read, some say 20 trades (Mark Douglas), others say 50 or 100 (Van K. Tharp). Actually, nobody knows. The one who knows that, knows the future. The one who knows the future would be the master of the universe. That's the question at 1 Billion dollars.
I have being testing out a couple of strategies and i get that one should let their system run for a while so that the results are not just noise or chance but the merits of your strategy (like tossing a coin 10 time and getting 9 tails , which doss not mean that the coin is not 50/50) but how can you know that its a failing system, how long do you have to stick to it . to know objectively that it dose not preform as you wish (hope this makes sense ).
To test a strategy, at first you have to do some back testing to find out how it worked in the past. If the result is consistent, then you may trade on your demo. If it fails 5 out of 10 times but risk-reward is 1:2, it is a successful strategy. If it fails 6 out of 10 times, you may want to skip using that strategy.
After I've properly backtested the strategy (very easy to do incorrectly), I watch it for 6-12 months in real time to see if it performs well in relation to the backtest. Now, backtests are almost always optimistic, but the strategy in 6-12 months after development should make money (I compare backtest monthly profit to real time profit). And, of course, this assumes that the strategy you've created works the same in backtest as in real time (many strategies do not).
I will say this real time "incubation" has save me thousands over the years. It is really hard to do (who wants to wait for that long to trade?), but it will save you money.