Robert Carver: TTU: Diversification of Trading Strategies - PursuitOfEdge/podcasts GitHub Wiki

  • people generally have a strong pref for systems that make money every month but every now and then lose a lot, which is the opposite of trend following
  • realize there are known risks and unknown risks, and we can't forget/ignore that unknown risk can happen
  • some people add skew/kurtosis to make their equity curve more blocky and realistic, co-skewness correlation
  • the more fancy your risk model the more calibration you need to do, and you're mostly just capturing expected risk, not enough unexpected risk
  • so he prefers to just keep it simple, just looks at risk as standard deviation and corrrelation
  • assumes the worst e.g. what if all correlations go to 1 - how much am i able to lose?
  • emphasizes going for a simple risk model that is easy for you to understand so it feels better
  • "if you get this risk management part right, the entries/exits don't matter as much"
  • says how stops are written about is incorrect e.g. people usually say "don't risk more than 2%" so if you have more money, you have a bigger stop, but the market doesn't know how much money you have and doesn't care, so the smaller trader gets stopped out earlier and the bigger trader gets stopped out earlier and this is wrong, and therefore smaller traders trade more frequently
  • but the truth is the holding period / volatility is the market has nothing to do with how much money you have
  • so when setting a top it should be based on how volatile the market is and how long you want to hold it
  • then it gets on to the position sizing which is actually how much money you want to risk. divergent systems can risk bigger than convergent systems based on their skew of returns
  • he does continuous trading which means he doesn't really have a stop, but he has a target position which is implied by forecasts and if the price changes the target position then he trades from long/flat/close
  • when trading discretionary, he recommends just setting a stop and you can't get out of the trade until the stop is hit, which forces someone to manage the trade correctly
  • most pro managers have a vola target, say car analogy 50 mph that they try to hit over a number of years, and how they drive on a specific trade is just determined by the forecast, with some sort of limit
  • his limit is 100 mph so he would never have an expected known risk of more than 2x his long-run average target
  • but could go down -10% or -15% on the whole portfolio regularly
  • if starting a research team he would look at size, intelligence, diversity in opinions
  • says a good system design is like building a building, then you just have to maintain it. says these firms are too big with too many researchers
  • trend following is a self-fulfilling prophecy
  • favors sharpe ratio, skew
  • negative skew: relative value, skew, long/short
  • is worried about negative skew strategies because they have not have their big down day in the backtest
  • is always worried about strategies that win often, make a lot of money, but haven't had a big loss - then it must always be exposed to one
  • his own trading:
    • 40 futures markets
    • diversification is the only free lunch
    • a whole bunch of assets and geographies
    • if he had more capital he'd have more markets
    • theoretically there is always a benefit in adding another market that isn't 100% correlated
    • his system is actually only 40% trend following
    • also has a carry/contango yield system
    • has a short bias on the VIX and european V2X - he believes being short the implied vol vs. realizable premium is a source of return, with an allocation of about 5-10%, these have about 4x the skew of the equity markets they track and if 2008 happened again it would be pretty dangerous, but he's also running carry and trend so if anything really bad happened he'd be out of the system so it's not much of a concern
    • when he was younger he would be more flexible with his system, but now that he is older he is more rigorous and prefers simplicity, so if he's gonna put something in this system he needs strong evidence that it's legit
    • test with some random data for real drawdowns
    • if you had a risk target of -20%, aiming for a sharpe of 0.5, it'd be about -10% drawdown, there may be 10 year periods where average drawdown is -50%, worst drawdowns are about -40% over a 10 year period about 2x risk target, if you're running a system for a few years and it goes down -40% you shouldn't panic because that's about what you should expect (WHAT?)
    • you really want to be a position where you can lose ALL the money you're trading, like -80% drawdown, what?
    • "The Predictors" - by Don Farmer
    • if he could pass down one thing to his children, it's optimism. he's a very optimistic person who has never regretted something in the past