I want to clarify a few things in response to my stop-loss post in which I stated adding a static hard dollar stop-loss is not always beneficial to your system/strategy (example provided).
I’ve received dozens of emails and comments and wanted to make sure I made a few points crystal clear.
The article was an attempt to show that using a static $$ stop-loss, specifically determined at or before the time of entry, does not test well. That is, using some logic like once I am down $x dollars close this trade is not a viable strategy or even beneficial to add to “most” existing strategies. Sure it feels good or safe or whatever, but does not work in many tests and real-trading. The market doesn’t care what feels good – you know that by now.
Secondly, I specifically mentioned I did not test any dynamic, technical stop-loss ideas for the strategy example in the post. For example, a dynamic stop of k*ATR may prove beneficial; however, nothing was tested like this. This is outside the point I was attempting to make. However, all the algorithms we trade have been tested exhaustively; I assure you of that.
Thirdly, I am not against ALL stop-losses. ALL of my algorithms have what I mentioned as an “oh s**t” stop to protect against a black swan, tail-risk event. Some algorithms even have trailing stops or other forms of stops; however, we only add these kinds of stops when the data says we should! We’re concious of risk. Our background is in the High Frequency Trading space; risk is very real and FAST.
Finally, if some exit criteria is met and an algorithm closes a trade for a loss then I do not consider that a stop-loss. For example, if your exit criteria is to sell when prices crosses below a moving average (excuse this rudimentary example) and this event occurs below your purchase price then this is simply a losing trade, but not a stop-loss.
Point being, we do not hold trades under water until they “come back” as a few of you have suggested in emails. We reassess our probabilities, recalculate our forecasts, make new predictions and send all that information to our decision making algorithm – which every one of our trading algorithms has. If this decision making algorithm decides the likliehood for price appreciation is more probable than price depreciation then we’ll continue to stay in the trade (loosley speaking of course) and visa versa.
In summary, not all stops are bad. However, static hard dollar stops usually are. If you’re worried about a black swan event add an oh s**t stop, but keep it outside normal market variations as it is meant for an abnormal market gyration and should hopefully never be hit! If you can’t keep it that far, then you’re trading too big. Furthermore, testing dynamic stops is always worthwhile and we always do. Finally, we’re believers in implementing bayesian or conditional probabilities in our algorithms – nothing static or hard dollar based; frankly, the market doesn’t care what dollar amount you want to use as your stop. Your decision should be based on the data – not on the fact that you think $500 is a reasonable limit. That $500 should change through time based on market conditions – that is more in line with our beliefs.
Thanks and keep a lookout for part 3 of this “Becoming a Better System Trader” series entitled “Why Sequential Trade Statistics (such as Drawdown and consecutive wins/losses) Can Be Decieving”