A few weeks ago I shared the results of a Crude Oil trading strategy with the Breakout Strategy Masterclass. Here’s a chart of the yearly out of sample results we discussed in the coaching call:
What I found interesting about this strategy is the years 2009 and 2013 showed very low performance but were then followed by more profitable results the following years.
This is a pattern that I’ve noticed quite a few times in my strategies and was something we mentioned in the previous podcast episode with Bruce Vanstone where we discussed ‘throwing the babies out with the bath water’ and trying to determine if a strategy could potentially recover from a drawdown or not.
I made the comment to Bruce that I’ve observed a lot of strategies I’ve created seem to have a year of low performance but then recover, so assuming a strategy could be broken after 1 poor year may not be safe or accurate.
But is this observation a pattern or not?
Humans are susceptible to seeing patterns where there aren’t any, so I thought it best to test out the theory with a quick exercise to see if what I’ve noticed occurs as often as I thought.
Luckily, I have a database of over 900+ strategies to test the theory, with strategies across 21 different Futures markets and 6 different timeframes. (using the process shared in the Breakout Strategies Masterclass).
My aim in this quick study is to identify the strategies that have at least 1 year of negative returns and determine whether the following year was a positive or negative year.
Before I share the results, here is some additional info on the strategies:
- The total number of strategies in the database is currently 981,
- All of the strategy results are Out-Of-Sample using Walk Forward Analysis, no In-Sample data is used,
- All strategies take both Long and Short trades, there are no Long only/Short only strategies,
- All results
are excluding slippage and commissions – yes, this will probably impact the results of this study a little but I’m looking for general observations so I don’t think it’s entirely necessary at this stage (perhaps for a future study),(updated 29/3 to include transaction costs etc) include $25 RT slippage and commissions,
- All strategy results are between 2009-2016 only, 2017/2018 are excluded (I keep these years out of my process/database as an additional check before I trade them).
So, I analysed all 981 strategies in my database, and extracted the strategies that had at least 1 negative year between 2009 and 2016, which totalled
742 (76%) 885 (89%) strategies. Surprisingly, this means 11% of the strategies do not have a single losing year – I would have expected that to be much lower.
The Average Loss for this group of losing years was
-$2,317 -$2799, with the largest losing year being -$22,790 -$24,799 and the smallest being -$5 -$3.15
The total number of losing years amongst the group of 742 strategies is
1090 1680, indicating that quite a few strategies had multiple losing years.
Now that I’ve identified the strategies that have had at least 1 negative year, what happens the following year? Here are the results:
|Total losing years:|
|Following year is positive:|
|Following year is negative:|
That’s a big difference, so what conclusions can we draw from these results?
This is a very rough analysis (that I knocked up in an hour) so it obviously needs much more work, but the conclusion I’ve come to is that just because a strategy has had a negative year it doesn’t automatically mean it should be abandoned. According to these results there is a good chance it could recover the following year.
However, this research raises more questions, for example:
- What about the magnitude of the losing year, do larger losing years have a lower chance of a profit the following year?
- Would we get better performance results if we started trading a strategy after a negative year or a positive year?
- What effect is Walk Forward Optimization having on these results?
Are there any survivorship bias factors in the strategy selection process that are inflating these results?How much impact is survivorship bias actually having on these results?
- And much more…
Actually, this quick study has raised more questions than it answers (ha!), but it’s interesting to know the stats based on the strategies I’ve created so far. You’ll have to do your own research to see if the stats hold true for your own strategies.
And I think year-to-year stability doesn’t really matter that much anyway, because real stability can be achieved at the portfolio level – it doesn’t matter so much if one (or a few) strategies have a dull year, when you have other uncorrelated strategies to pick up the slack.
If you’d like to know how we build loads of breakout strategies following the same process that breakout trading specialist Tomas Nesnidal used to launch his own Hedge Fund, you can find out more details here.
UPDATE 29/3 – Updated to include transaction costs. The more I think about it, the more I’ve come to realise the potentially serious flaw in this study – without being able to quantify the impact of survivorship bias we can’t really say these results are accurate at all. I suspect I’m underestimating the effect survivorship bias is having here so more research is definitely required.
28 March 2018