I am trying to figure out some formula that will help me get to a # that is neither over or under capitalized for each specific system I subscribe to. Obviously many factors need to be taken in consideration, anyone willing to share some idea’s?
It’s all about risk and everyone has a different risk tolerance. You need to answer two questions.
1. How big a drawdown can I handle without bailing on the system?
2. How big a drawdown should I expect in the normal course of affairs?
The first, only you can answer. You need to trade a good system small enough that you can stick with it through the inevitable drawdowns. It’s a bad feeling to bail on a good system only to see it go on to new highs.
The second is harder with the limited data available at C2. Ideally, the vendor would do a bootstrap sim on the historical trades (shuffle them into a different order, calculate the max drawdown, repeat thousands of times) and give you the probabilities of $X drawdown.
Let’s take two examples.
A young gambler who trades futures instead of going to Vegas might think 70/30 are pretty good odds. He might be willing to accept a 30% probability of busting the account (100% drawdown) hoping to hit the jackpot.
At the other extreme, a retired teacher might be looking to do a little better than t-bills but she would be very upset if she had a 10% drawdown. So she might trade small enough that there is only a 0.1% probability of a 10% drawdown.
Most of us are somewhere in between.
Of course, what I wrote above is a somewhat utopian view. It’s unlikely that many vendors at C2 will be running bootstrap sims. Another way to look at it is the old rule of thumb to not risk more than 2% of the account on a trade. If you’re risk averse, the number should be smaller. You can get an idea of the risk by looking at the biggest loser in the trade list. So a trader with average risk tolerance might need 50x the biggest loser to trade a system.
How does Monte Carlo differ from Bootstrap?
They’re similar in concept but different in execution. I wasn’t a math major and when I started doing what I described above in the 90s, I was calling it Monte Carlo. A math prof pointed out that I was really doing bootstrapping which I had never heard of. Wikipedia has some explanations of the two. Anyway, I’m not trying to predict an equity curve like the C2 MC does, just the probability of a single variable, the max drawdown.
Dennis, you wrote:
Ideally, the vendor would do a bootstrap sim on the historical trades (shuffle them into a different order, calculate the max drawdown, repeat thousands of times) and give you the probabilities of $X drawdown.
Do you know of an Excel marco or stand alone program that will do this?
In your experience, how well do the results concur with forward testing?
Alex Matulich’s ProSizer for Excel ($37) does that and more although it presents the info a little differently. If you want to write your own simpler one, you need a column of random numbers and write a macro that sorts the sheet based on that column.
http://unicorn.us.com/trading/prosizer.html
Dennis, thanks for the link.
I followed it to another link which yeilded a very interesting free online book “Trading Strategies” by Larry Sanders http://www.tradelabstrategies.com/
It explains in simple terms the theory of bootstrap (he uses monte carlo), and effects of varying paramenters. In essence a strategies return is a relationship of risk to a strategy’s probability and win/loss ratio. Probability being the most influential of a strategy’s characteristics and risk being a trade’s position size. The book also points out that there is an optimal relationship between equity return drawdown and risk. Drawdown is in fact desirable. After reading this book, your next strategy drawdown will be an elevating experience that “you are optimized”!
