Hi all, I have a system that's losing a lot of money recently. I considered adding a rule to the system, which is much like "circuit-break" or dynamic risk reduction when things go crazy. The change adds 2 parameters to the system. It improves Sharpe ratio marginally, but it improves the overall average return/max drawdown ratio by 40%. It improves the ratio by risk reduction when things go crazy. In backtest, breakdown into each year, for most years, it improved the annual return/annual max drawdown a bit and for the rest years, it actually degrades the ratio. But overall, it cuts the overall max drawdown at a little bit expense of the overall average return, thus achieving a 40% improvement in the overall average return/max drawdown ratio. Shall I add these two parameters and put the circuit break mechanism on?

What is crazy for you is insane for someone else while seems normal to others , business as usual. Too general.

Do what YOU think is right, ignore the recommendations of others. It's not their money at risk, it's yours.

Return/Max Drawdown is a skewed number. If you are trading for 1 year, say you have 10,000 return and 1,000 drawdown. Your ratio is 10. Continuing at that pace for 10 years, your return would be 100,000, and your drawdown would be 1,000. Now your ratio is 100, but the performance did not really improve. Sharpe/Sortino ratios annualize returns. Therefore, the number would remain static over the 10 years in the example above. Additionally, these ratios use volatility as opposed to maximum drawdown, and a larger drawdown does not neccessarily mean you have higher overall volatility. Perhaps the number you are looking for is: (annualized rate of return / max drawdown)

only use a new rule if it doesnt significantly effect the scalability of the strategy. I assume your testing equities... see how the strategy performs in terms of scale (i.e with millions, and even more) without the new rule versus with the new rule. Some rules are kinda fudges that seriously limit the viability of a strategy. Remember people want controlled risk and reduced exposure to vanilla equities, so dont add a historically biased rule to clean up the occasional mess...

With all respect, you need to lose some money. Only when you have done that will you know what the right rules for you are.

Absolute return to drawdown is meaningless, as a previous poster said for the right reasons. Examing the Calmar ratio of APR to max dd is much more useful, generally try to target 2, 1's fine if max dd is not excessive, and 4+ is a holy grail. Sharpe ratios are really only relevant with at least two years of data, and you need to watch which risk free rate your simulation program is examining to come up with that statistic. 2 is excellent, 1 is still good over long periods of time.

oh thanks for pointing that out. (annualized rate of return / max drawdown) is exactly what I have been referring to... sorry for the confusion...

Oh I am testing futures... so the notional scalability for me is an arbitrary number... I agree "so dont add a historically biased rule to clean up the occasional mess..." but where is the fine-line there I am not sure...