What is the acceptable range for the INFAMOUS "Maximal drawdown" percentage?

It all depends on how aggressively you want to trade, the nature of your trading system/strategy and how strong your stomach is for big drawdowns.

For my trading style I like to see figures around 10% or less and not more than say 15%. My experience is that when actually trading live, the draw down often exceeds the figure that you get from testing.

Regards

Tim Wilson

My elementary arithmetic gave me this table awhile ago which I keep visible, just to keep me always trading much less than perhaps think I should - compounding will always win in the end... IF have any funds left that is :/

.

.

Formula used: 100*%drawdown/(100-%drawdown)

FX trading is a stochastic "investment" model, and because it is stochastic** it means the bottom line will always fluctuate. However, it tends to fluctuate within statistical limits once you know the stochastic parameters of your EA. If your backtesting has successfully handled a 1000 or 2000 trades then you know your stochastic parameters with a high degree of confidence. If your backtesting has handled only a 100 or so trades then you do not know the stochastic parameters very well.

A year ago I created a Trade Simulator in Excel to gain knowledge on this very question from a purely statistical perspective. Just how much draw down is acceptable. I tend to agree with TSWilson because a 10% draw down in most years can occasionally have a 25% to 30% drawdown in the outlying years statistically speaking.

** (Please note when I say "stochastic" I'm NOT refering to the Stochastic Indicator --- I am using the word in accordance with its standard definition)

What I do is run a monte carlo permutation:

you take your returns (expressed in deposit currency, e.g. USD), this is a column vector, and then you randomly select

returns from your this vector into a new vector with equal length (replacement is allowed, so it is possible to select the same trade twice).

Then you do this over and over (I suggest 10.000 times or more, doesn't take long at all with a for-loop) and you write a simple function to

calculate relative drawdown (which can be 1 trade or more trades: minimum, smallest 2 trades, smallest 3, ...). You then draw a histogram

of all the drawdowns (this will be skewed to the left) and calculate e.g. the mean of that distribution.

This gives you a good indication of what you / your system can be expect to get as drawdown over time because the sample mean is an unbiassed

and efficient estimator of the population mean. Offcourse the initial vector with trading returns should have the correct sample size.

If your actual drawdown exceeds this number, something has changed and you should stop trading, and go back to the drawing board.

I really appreciate your replies. I've been provided with the answers I need, covering the empirical, stochastic and even psychological aspects. I guess will have to perform the suggested simulations. Your suggestions tell me that I should accept a value between 10% and 15%, as trying a smaller one will significantly affect the implemented trading strategy.

Thank you all.

A.M.