- This topic is empty.
-
I have been asked about this a few times now so I thought it prudent to at least write my thoughts and outline an example.
Trading the equity curve means plotting a moving average of the equity growth chart. When real equity drops below the average, you stop trading the system in real time yet continue to plot equity in simulation mode. When equity again moves above the average then you start trading real time.
Goal:
– to ensure you’ve stopped trading the system should it start to fail due to lack of robustness or the effects of over optimisation
– lower drawdownsCaveats
– this method will not stop drawdowns
– should only be used on short term systems that DO NOT exhibit trade clustering
– is not appropriate for trend system or longer term systems
– most people calculate the MA with a post dictive error.The following chart shows a SPXL ETF strategy that exhibits 200 trades since 2008. Results on other indices or stocks are not overly strong, so we would suggest that (a) this is a single market system
(b) it doesn’t exhibit enough sample to offer confidence in robustness.(although we can use the SPY as a proxy back to 1993 with 660 trades)
However, can we trade it?
I believe we can so long as we have some trigger to stop and reassess. The next chart shows the same system with a 20-trade SMA (Green) and the subsequent equity curve created by only trading when the equity curve is trending up.
Summary
– The end equity is about $300 lower.
– Drawdown is now $3724 vs $8747 (no compounding used)
– 28 trades are skipped
– we have a clear OFF switch should this system start to fail..
And here is an example of why the logical can be wrong in reality and why the ONLY reason we do this is to avoid a catastrophic deterioration in the system. Here is a basic pattern setup that is trading Aristocrat Leisure (ALL). Whilst the strategy works ‘reasonably’ well across other stocks it really doesn’t stack up well enough to trade on ANY stock. If we pick the better stocks and use this method we may well be able to survive if it starts to fail.
This chart shows the raw vs. filtered data using our same 20 trade SMA. Note however that the filtered version actually underperforms. The simple reason being was due to the system switching off due to a bad trade but then missing a great trade. It switched back on again, only to take another bad trade, switched off and then missed two great trades.
Obviously the Achilles Heal is that chop between good and bad trades. The payoff for avoiding a catastrophic system failure is actually underperforming the real system itself – which then begs the question whether or not you’d trade the filtered version anyway?
This is why I prefer a robust strategy from the outset.
- You must be logged in to reply to this topic.