Bits for Thought #1: Good Stop (loss), Bad Stop (loss)

Bits for Thought:
....are the philosophical premises that precede any quant project or idea, no matter how simple. Yep!, your are right, they are more important than quant itself, because they help to set up problems properly before we mess them up with our stupidly complex quant. Quant, by the way, ends up being adding and subtracting and actually not much more than that, a thorny issue that we, quants, of course will never admit.


So here comes our first BFT on stops (the so much overlooked problem): Good stop, Bad stop.


The Good Stop Loss is the one that if triggered is for a good forward looking reason, in which case will be located somewhere between the High and Low of the next trading bar (yep! otherwise it won't be triggered!!). If it's not triggered, the good stop is the one that is located right above the next bar high (for a short position), or right below the next bar low (for a long position). In this case they minimize the ex ante risk of our position.


The Bad Stop Loss is triggered for no good forward looking reason, meaning it exits us from a winning trade too early. In this case it happens to be located somewhere between the H and L of the next bar when it shouldn't. If it's not triggered, the bad stop is the one the leaves us open to an infinite (or lets say high) amount of risk when looking forward to the next bar. A bad stop for longs is way below the next bar L, and the one for shorts, well above the next bar H. The reason it is bad is because it leaves us unnecessarily exposed to risks, even when these risks end up not materializing. This is of extreme importance because when a rare event (war, terrorist attack, statistical glitch on US GDP data like in july'11....) hits the market, it might find you unprepared.


By now, "Good stop, Bad stop" should have been enough to open your mind to the true depth of this problem. If you conclude that this is a multidimensional problem with a huge number of alternative solutions, then you are right. The only thing we did was to chose the one that fitted best our requirements (in statistical terms) without overoptimizing. The result... ExtremeStops. Remember, there is always something you have to give up, a trade off.