Reducing Drawdowns

[chesl73]

I have a system I'm working on and in general I'm happy with the results, except for the drawdowns. The peak-valley drawdown is 25% (this is for the period 2003 - 2009), for 2003 - 2007 its 15%.
I have two questions I'm hoping some of you more experienced system designers could help me with..
1. Would you say that this level of drawdown is too high? (I know it depends on the individual's risk tolerance etc but in general?).
2. What I'm unsure of is what is the area of the system that I need to address in order to try and reduce the drawdown? ie, is it the exit? is it maybe my stoploss application? Changing my stop losses just seems to reduce the profitability and leave the drawdown unchanged. Is there anyway to reduce the drawdown or is it just an inherit part of my system and it'll be difficult to do with changing the overall system?

Any help would be appreciated.

 

[CascadeX]

Try going back and looking at the instances where there is high drawdown.
See why that is the case, and then write a formula to exclude holding onto those positions.

If you don't do too many trades with your system, the other alternative is to hedge those positions, but this is something beyond the scope of AB so it won't be included in your backtest results, not unless AB allows you to include derivative positions in the system.

 

[beerwm]

-reduce your PositionSize [double-edged sword]
-increase winrate of trades [at expense of R:R/easier said than done]

-take less correlated trades -
eg.

>maybe long + short simultaneously/different entry concepts[pullbacks/swing trades, etc]

>keep system risk exposure the same - but spread it across move positions
eg. 20 trades @ 1% PositionSize is less correlated than 10 trades @ 2% PositionSize

 

[Chorlton]

I have a system I'm working on and in general I'm happy with the results, except for the drawdowns. The peak-valley drawdown is 25% (this is for the period 2003 - 2009), for 2003 - 2007 its 15%.
I have two questions I'm hoping some of you more experienced system designers could help me with..
1. Would you say that this level of drawdown is too high? (I know it depends on the individual's risk tolerance etc but in general?).
2. What I'm unsure of is what is the area of the system that I need to address in order to try and reduce the drawdown? ie, is it the exit? is it maybe my stoploss application? Changing my stop losses just seems to reduce the profitability and leave the drawdown unchanged. Is there anyway to reduce the drawdown or is it just an inherit part of my system and it'll be difficult to do with changing the overall system?

Any help would be appreciated.

1. Unfortunately it really does come down to the individual but as this statement is not that helpful I will add that for me I'm looking for a much lower MaxDD.

It is worth remembering that your biggest MaxDD is yet to happen (but happen it will at some point). A lot of System Developers suggest doubling the MaxDD for real-life trading. If you double it and its value is too high then you should re-assess. Personally, I look for a value of around 5% or less when backtesting.

2. As Beerwm has suggested increasing the WinRate athough it should be noted that this will probably have a negative impact of overall profitability.

I will assume that you are looking at a longish term trend-following type system? If so, then it is my belief that the real focus of attention should be on the exits. The longer the trade duration, the less important the entry IMO.

When you refer to "Stop Losses" are you referring to Initial Stops, Trailing Profit Stops or are you simply referring to all of them?

 

[chesl73]

Thanks Chorlton. I also think that the maxDD is too high for me as well hence the question.
It's a longer term system yes.
When I'm talking about stops, I"m talking about any stops. I've been looking at anything from chandelier to parabolic to MA's. My exit at the moment is quite basic and probably requires a bit of work.
I'd also like to test having one stop initially with a bit more leeway and then after xR profit then tighten it up but I'll have to work a bit on my Amibroker skills to do this first!
And also maybe have a different, tighter stop in place depending on market conditions (bull/bear conditions). It's a work in progress. Thanks.

 

[chesl73]

Also Chorlton, when you say you look for 5% drawdown, you can't be getting that with a long term trend following system - can you?

 

[yonnie]

maybe not a low drawdown is possible if you are getting into slow moving stocks.
however if you are aiming to get into rockets which you will follow until they burn out, you can tolerate very little loss if they move against you.
and onto the next launch pad.............a bit like darvas

 

[Synergy]

A couple of things that may help:

Play around with the variables to try and isolate the trades that are performing badly. I found that moving variables to the extremes (high and low) often uncovered small groups of trades that performed badly. Whether the variable is price, volume or something else, it seemed that these extreme cases were often very volatile trades and not so good for DD.

Also, have a look into time based stops, especially if you have an entry with a positive expectency and have excess buy signals. Your stop should be there not just to minimise losses, but to cut a trade that is likely to underperform a new trade.

If you still have excess buy signals, you can afford to trim your system to better match the number of trades you can take. It should be fairly easy to decrease the number of buy signals while increasing your systems stats.

Drawdown may be lowered by improving other aspects of the system, rather than trying to specifically minimise losses.

 

[howardbandy]

Hi Chesl, and all --

Let me make a couple of observations and recommendations:

1. You will hear me say this over and over -- be certain you are evaluating out-of-sample test results when you estimate what the trading system results might be in the future. In-sample results are always much better then out-of-sample results and will always underestimate losses and drawdowns.

2. In general, stop loss exits hurt trading system performance. If a system has several trades that show large losses, say 15% or more, it is tempting to put stop loss exits closer to the entry than those large losses, say 8% or so. Making no other changes in the trading system, the closer the stop loss exit is to the entry point, the worse the overall performance of the system will be. Spending more effort in designing exits based on characteristics of the trade will be a better solution than stop loss exits.

3. Drawdowns are proportional to the square root of the time in the position. This is independent of the market or the trading system. The constant of proportionality varies market by market, stock by stock, and year by year. For example, based on the period 1/1/1990 through 1/1/2006, for the 2700 stocks that were listed on the NASDAQ as of late 2006, you could expect a 5% drawdown for a 9 day holding period. The range, based on which year, was 2% to 13% drawdown for a typical stock in a 9 day holding period. (Due to the very significant survivorship bias, the results that would have been observed if the list was those issues that were listed as of 1/1/1990 would have been much worse.) I recommend analyzing the issues you plan to trade with your system to see the specific relationship between holding period and drawdown.

4. If you find that the drawdown that results from a typical trade (measured by out-of-sample tests) exceeds the maximum drawdown you want for your trading account, divide the account risk you are willing to take by the per-trade risk your system shows. The result will be a number less than 1.00 -- say 0.35. Then take positions with only that portion (35%, in this example) of your funds, leaving the remainder (65%) in cash. In general, systems that trade stocks or ETFs cannot take full positions without violating the account risk limits -- there will always be some money left in cash earning short-term interest.

Thanks for listening,
Howard

 

[CascadeX]

Ah ... Howard Bandy.

Nice to see you're on here.

I enjoyed reading your Quantitative System book immensely. Get's me excited about developing a system.

I will be waiting for your next book.

Regards.

 

[Synergy]

Just another quick one on this...

Having the system turn itself off during unfavourable times was probably the thing that had the largest effect on reducing drawdowns for me. But i trade volatile stocks that get hit harder when the market moves lower, so may not apply.