Two Dark Swans Of Stops

[Predictor 11]

Why Bounding Losses Doesn't Work Well

 

The basic strategy for choosing a stop loss size is to categorize trades into 2 groups: winning trades and losing trades. The optimal stop is in theory set where we can separate the groups best. The problem with this basic strategy is that it forgets about type 3 trades: losing trades that are essentially winners -- that is losing trades that have a greater MAE then the closed loss. Good systems, like good traders, are usually able to exit losing trades gracefully and that can add up to a lot over time.

 

The Tight Stop Worked

 

As I was running my trades through the analyzer, the trend was clear: larger stops were more optimal. I remember reading an article some years back that strategies that were at near extremes tended to do better (in a strategy game), and decided to see how my trades would have worked with a very tight stop. Surprisingly, the system kept a good deal of the profits and even some risk-adjusted metrics improved. The larger stops worked better overall but very tight stops, based on my overall assessment, were working as well as the moderate sized stops. Yet, the win ratio dropped significantly.

 

Explanations

 

The explanation was that I was skilled at exiting my losing trades at favorable prices. However, if I set a stop tight enough then my losses became very small in relation to my wins. In other words, I was always exiting losing trades at favorable prices. The cost for this was that my win ratio dropped from 80% to below 50%. Having monitored several systems, I have developed a feeling that systems with win ratios below around 55% can be trouble.

 

New Understanding

 

The sample that I tested was a volatile month. I was thinking about that and something from a blog that I read when I realized that tight stops are essentially a bet on expanding volatility while large stops are a bet on constrained volatility. The bets are similar to options but are path vs end point sensitive.

 

Dark Swan Of The Large Stop

 

The requirement for using large stops is to ensure they only rarely get hit. It is obvious that if the market becomes too volatile and the stops are hit then large losses can occur. Imagine a logical stop placed by imagining a bollinger band, as the markets volatility increases the size for the stop needed grows beyond reason.

 

Dark Swan of The Tight Stop

 

Okay, so it may seem that the tight stop reduces the risk. Yet, the tight stop is essentially betting on expanding volatility. There are 2 reasons that this is the case. The first reason is that if one sets a tight stop then one obviously anticipates the market might continue to move against them: its just an assumption built in. The second reason is perhaps less obvious but systems using tight stops rely on out-sized winners.

 

Any decrease in volatility can wreck havoc on any such system, and the small losses can grow into an avalanche.

 

Possible Takeaways

 

One possible takeaway is that during volatile markets that traders should pull their stops in and accept lower win ratios or sit out if they prefer larger stops if the volatility grows too great. Likewise, in less volatile markets traders should prefer the larger stops because the volatility isn't likely to produce the out-sized winners (or sit out). Both of these implications are counter intuitive.

-----

http://themarketpredictor.com

Edited August 9, 2012 by Predictor

[onesmith 10]

Imagine a logical stop placed by imagining a bollinger band

 

Standard Deviation is considerably different than Volatility. StdDev increases as Variance increases. Variance is the difference between current avg and previous values of close. Picky but hopefully helpful.

 

Thank you for all the topics you've started recently.

[BlueHorseshoe 10]

Imagine a logical stop placed by imagining a bollinger band, as the markets volatility increases the size for the stop needed grows beyond reason.

 

Thanks for another interesting thread.

 

In addition to OneSmith's comments regarding std/volatility, I would add that std is not a valid indication of probable price distribution around an entry that has a positive expectancy (note: this is not the same as the strategy's expectancy.). One would expect a skewed bell-curve.

 

I tried to start a discussion on this topic when r:r ratios were brought up in the daytrading the e-mini thread, but it was deemed off-topic, so good to see people willing to discuss it here.

 

BlueHorseshoe