Be the first to read about algo trading analysis, research and news

Don’t worry, we want spam you! We only send about 2 emails per month.

Subscribe to: Newsletter
Algorithmic Trading Blog

Algorithmic Trading Blog

Automated Trading System Development
Quantitative Trading Research
Portfolio Backtesting

In my experience , drawdown can be one of the most misleading metrics when analysing system performance. Every investor wants to know what they can potentially lose or what a system has lost in the past, this way the investor has some expectation going forwards. This is a measure of risk and survivability of a system especially when you consider margins (futures markets are traded heavily on margin) are involved.


Reasons why I feel drawdown is not so important

I find drawdown really is a misleading metric when considering past performance:

The drawdown percentage depends entirely on the starting capital. Most system back test reports are run with the assumption of starting $100k capital trading 1 contract with no compounding of returns (this is for futures trading systems). However, some developers will then inform investors that the system can be run with as little as $10k and still have the same performance.

  • Let’s assume the backtest report stated a 10% drawdown on $100k right of the start of trading = $10k.
  • For a trader starting this with $15k the dollar value drawdown will be the same because the assumption was trading 1 futures contract. In this situation, the drawdown percentage is now 66.6% and now we must assume margins will not be met in the future and the system is at ruin.

 

Normalising drawdown to market price

Another factor that nobody talks about is normalising the drawdown to market price. Just like in real life prices increase overtime (phenomenon known as inflation) as do most futures markets.

Moreover, markets move in percentages and care very little about dollar value. See the following example where we are not discussing drawdown loss but the number of ticks on a daytrade lost converted to a percentage at that current market price:

 

  • A strategy with max close to close drawdown from 2007 - 2016 was $2,000. This strategy is daytrading only and closes at the end of each day.
  • The above system max percentage daytrading loss was 2% in 2008 (this is the high of day minus low of the day when we talk about the daytrading range). In dollar value this 2% move = $600 per contract.
  • Now in 2017 with the market price more than triple what it was in 2008 should we have the same 2% max loss then the dollar loss will be triple what it was in 2008 e.g. $600 * 3 = $1800.
  • As I mentioned markets care only about percentages and futures move in tick represented by a dollar per tick value. Clearly the higher the current market price there will be more ticks for every 1 % move meaning more dollars per every 1% move in the market.
  • Hence, with the example we show that drawdown is completely irrelevant when looking at past backtests. This is only in respect to futures trading with a small account e.g. 1 contract.

 

This whole post implies futures trading with a small account, as we know futures trade in contracts so drawdown for an individual trading 1 contract is entirely misleading when looking at past data.

Investors have very little understanding regarding drawdown and often see a system with a max 10% drawdown as promising. We really must know the starting capital and then normalise this to current market prices. We need to convert each daytrade made into ticks lost/gained and then turn this into a percentage at that current market price the trade was made. We then convert the percentage gained/lost to today’s market price to get an accurate assessment of dollar drawdown.

This is another reason why bogus systems with fixed dollar stops or targets cannot and will not ever work as market price changes and as mentioned many time markets move in percentages and hence dollar value increases as price increases.

Primarily this is a problem in Futures trading where an investor can only trade 1 contract, for larger accounts we can adjust the number of contracts being traded based on capital, risk tolerance and the percentage drawdown risk of the trade.

 

Comments
Anonymous commented on 04-May-2018 09:57 AM
I have been a systems trader for a long time, I am shocked that I have not paid this more attention. Makes sense why futures testing should be done as percentage of account equity for position sizing.


Post a Comment



Captcha Image


Ready to Learn More?

We welcome your messages/calls and we reach out to you as soon as possible to provide more information including in-depth reports for our results and performance. Our CEO will contact any prospective Client personally.

Our office info

  •   info@quantsavvy.com
  •   US/Canada Toll Free: 1-888-414-0759
  •   Rest of World: +44(0) 208-935-5852
  •   The Techno Centre,
    Coventry University Technology Park
    United Kingdom, CV1 2TT

Connect with us

Contact us - we respond very quickly, or call anytime from 9am to 11pm GMT London

Send us a Message