A trading system very often is based on long standing observations of market behaviour. Market participants changes but crowd behaviour never. That is the reason various market patterns can repeat itself with high reliability. If a trader observed/noticed a certain market pattern, he may then place his trade based on the pre-cursors/sign of that pattern. Then as the market moves, there will be a good chance it will be moving in the trader’s favour.
There will never be a system which is 100% accurate. A system with a >70% accuracy is considered a very good system. The trader will be profitable as long as he follow the system. In this example, it is assumed the system has a reasonable risk to reward ratio. (Risk to reward ratio is the amount you are risking for a gain. 1:1 means you are risking a dollar to get a dollar).
On the other hand, a 30% accuracy may not be a bad system either, say if the risk to reward ratio is 1:5. So out of every trade, you expect to get $500 for every $100 wagered. So in a distribution of ten trades, you will lose 7 trades and win 3. But you still ended up positive (3 X 500 – 7 x 100 = 800 profit).
Do not be fooled if one claim to achieve a 90% accuracy when you need to risk 500 to get 100.
Yes, trading can be compared to gambling. But a good trader plays the house, the unprepared trader plays the player.