Reading Notes from Rob Carver's Blog

I intend to regularly post notes and comments to Rob Carver's new blog posts. The purpose of this exercise is to get an easy-to-read table of contents of the studies Rob ran, as well as to force myself to regularly catch up with his work. 

Does it make sense to change your trading behaviour in different periods of volatility?

Conclusion: yes - but as volatility gets higher, faster trading rules do relatively badly, but actually the bigger story is that all momentum rules suffer.

Using maximum drawdowns to set capital sizing - not as bad as I first thought

• Rob usually measures risk using annualized daily standard deviation of returns, and thinks that ATR is a good approximation as well
• Rob has strong opinion on how to calculate MDD. 
• Suppose you start with £10,000; but then you lose 1% of your capital for the next ten days. So what's your maximum drawdown? Is it 10% or 9.6%?
• His preferred way of calculating is simply adding up the ten of 1%. Note that this could be over 100% in extreme scenarios
• Rob's main argument against using MDD is that using MDD could underestimate the risk and hence over-allocate risk
• Given backtest period is likely short, MDD derived from backtest is smaller than actual, which means risk allocated is larger than needed - risky!
• In addition, there are many factors in the backtest that makes it look better than actual (all kinds of biases, for example). Then again, more risk is allocated than needed
• Distribution of MDD is skewed, so that observed MDD might be at a much higher percentile in the distribution
• Rob advocates the use of Kelly Criterion to set the risk (vol target = SR), and realistically use half of that. Still, similar skepticism exists so that you should probably be more conservative than what the number tells you
• One more reason: if you use SR based metrics, then as new data come in your weight won't change too much but only gradually; and if you use MDD based metrics, then your weight won't change at all, and with a single day it changes drastically - bad

Dynamic trend following

• Rob discussed "dynamic stop losses" which means having tighter stop losses at the beginning, and increase the stop loss after some good profit to have a better chance of hitting a home run. Rob believes otherwise that one should run the system in a "stateless" fashion, meaning that the trading history shouldn't play a role in future decisions
• 
  
•