Bits for Thought #2: Stops for the Long Run?

All of us have read all sorts of comments about stops and their applicability for the short/long run. Too much noise, no clarity at all. But there is a reason for this. Remember what we said before? This problem has an infinite number of solutions, so there is room for all opinions, and also for a great deal of statistically acceptable efficient stops. There are many combinations of slightly different stops time series with similar 'good enough' statistical properties, just choose the one that suits you. We will make our contribution to the short term/long term dilemma knowing that, almost for sure, our view will be just another drop in the already messy and infinite sea of opinions.

Stops for the Short Run

This is the easy one. Consistent stops should be developed considering from the very beginning that they apply to a very specific (and short, of course!!) time frame, after which they die replaced by new stops (it's kind of a walk forward procedure). Without fulfilling this premise, stops are just an almost worthless number void of any statistical properties that might render our investment process inefficient. When we think about the short run the reasoning is simple, on breaking the stop level we want us out of the position to avoid losses that, with a high degree of certainty, might take place in the next few bars. 

We opted to calculate ExtremeStops in a weekly basis because the idea was to help the short term trader to minimize losses in the daily basis without any consideration about the long run. It follows that: 1) those stops won't be very far from the last close (asymmetric; on average 2 to 3 percent points above/below the last available close), 2) they apply approximately to the next 5 trading days and 3) short term re-entry strategies should be considered carefully. This is the case of the high frequency trader and the reason for ExtremeStops to exist (we later found some unexpected side effects that we will tackle in a future BFT release).

Stops for the Long Run

Now, lets think big, lets set the problem for the long run (12 months). Here comes the mess we want to disentangle. For the sake of simplicity we will assume the case of someone that is long the market. If you are investing long term there is no point in betting downwards for a sustained period of time. Your bet will, on average, be that the market continues its march up.

Mess 1: Investor Z makes decisions with a 12 month time frame in mind. He thinks that if he invests for the long term he should accept more losses before he exits his position. In mess 1, the size of the maximum accepted loss is implicitly proportional to the time horizon of the investment (the longer the investment term the higher the loss he should be willing to accept before exiting a trade). WRONG!

How much would this individual be willing to lose before he exits his position? Since the longer the term, the higher the loss he will accept before exiting, what's the right figure? A 5% loss?, 10%? The longer the time of our planned investment the wider the stop is sort of an insane philosophy. There's no need to scale the size of the stop loss with time because that way we are increasing the chances of materializing a big loss just when the pain is about to end.

Mess 2: Investor Z wants stop losses that 'save him' from 10% to 20% crashes (those that last for several weeks; remember july'11) but never be forced to exit unnecessarily from a nice long and correct long term position. Z doesn't want any of those obnoxious stupid useless stops, unless all these 'very reasonable' conditions are satisfied. WRONG AGAIN!

OK so Z wants perfect stops. Those wide enough during the uptrend so that intermediate small corrections don't exit him, but very tight the moment a serious (and lasting) crash is about to happen. Doesn't this sound stupid? In the first place we need to have in place a convincing tool that deals with long term re-entry points efficiently, just in case we are stopped out in an intermediate correction. In this case, please, use common sense and a moving average to signal when to exit and don't burn your brain searching for the perfect stop. A moving average (50 days, maybe 100) coupled with some sort of fundamental tool (for conviction and re-entry signals) will do a fine job. This combination will help to avoid some of the mistakes the average will surely make.

Conclusion

A few thoughts:

1. Decide first what type of investor YOU are, high frequency-short term or low frequency-long term
2. Remember that scaling your stop with time does not necessarily make it more efficient. The  relationship between the size of the stop loss and time is a little more subtle and convoluted. The stop loss should ALWAYS be linked to the probability of more and unacceptable losses materializing in a specified future period of time.  
3. If you are a long term investor then don't ask for miracles when using short term stops (and don't complain about short term stop losses!!, of course they are going to exit you more frequently than you would desire). Use instead a moving average to generate exit signals and some kind of long term fundamental tool to provide some meat to the decision making process. It is not the percent you decide to lose what makes the stop long term or short term. It is the nature of the underlying tool that generates the stop what makes it. If you are a very long term investor, search in the web for the Mebane T. Faber paper about the 10 month moving averages study.
4. If you are a short term investor use a device suited for that purpose. The idea is to preserve capital and be able to shoot again if you made a mistake in your last trade. Be ready to compensate the stops system errors with a conviction/re-entry tool. ExtremeStops/DTPA is an alternative, though not 100% flawless. Try yourself, you may come up with better tools.