Say you look at your PNL from a given strategy and calculation the skewness of the histogram. You find that the strategy is losing because of a few large outliers, but otherwise, the strategy has mostly winners. When is it NOT correct to put a stop loss to limit losses? If the market is liquid and I can avoid the loss with a reasonable loss threshold, am I cheating myself and introducing some bias from having observed the results of the backtest?
1)?......skewness......histogram......outliers......garch.......those are "big words". 2) Having observed the results of the "backtest", you're cheating yourself and wasting your time & money if you willingly allow a "small number of large losses" to ruin what otherwise may be a good system. 3) Strategies that have mostly winners are prone to large losses because the trader can be overcome with perfectionism and has difficulty limiting losses. 4) Lesson learned?......Cut losses!.....whether or not you use stop orders to do so.
I'm a bit too academic for my own good, so sometimes some guy telling me "cut losses, stupid" is the right medicine.
Haha, I almost went to Penn State. It's not that bad a school, kind of peaceful out there in the middle of the hills. I don't know if you read Ernie Chan's quant trading blog, but he avoids stop loss operations in mean-reversion -- or at least, I think I read he did. Don't remember. I was trying to balance his assertion against an equity curve I'm seeing where everything looks smooth, but there are two disaster points that are a bit disheartening.