One doesn’t: it requires very careful, methodical, statistically significant testing. There [U]isn’t[/U] an [I]overall[/I] “best approach”.
My own opinion is that in general, for most systems, most of the time, trailing stops lose money. I have several different reasons for believing this …
(i) When I tested my own five little systems, only one of them gained from a trailing stop: the others all lost;
(ii) In other forums, when I’ve occasionally made this comment, I’ve always had some “Good heavens - you’re right!” comments from people who have methodically backtested it for the first time and realised that they were more profitable without the trailing stop;
(iii) All the authors whose textbooks I really respect and whose opinions have proven right and beneficial to me in other areas seem fairly opposed to trailing stops.
They always look and sound attractive and appealing, but it’s very easy to lose count of the times a trade starts off doing well and then retraces a bit (just enough to take out the trailing stop, which has of course moved during the initial phase of the trade), before continuing on its merry way in the “right direction”. So the times that trailing stops cost money tend to be “opportunity cost” money, i.e. you make less than you might have done, i.e. they’re profitable trades anyway (often) which is why one doesn’t always “notice” as much as one would if one were looking at actual losses.
Yes - I believe that trailing a SL manually just above/below the most recently formed swings high/low is, overall, [U]very[/U] significantly likely to be a better approach.
For myself, I’m not willing to use automated trailing stops in the absence of [B]really[/B] clear-cut and statistically valid proof that they’re better than other methods.