The Challenges of Creating Intraday Trading Systems
Today’s article is a short reply to a reader’s question:
“Is it possible to utilize the Weekly/Daily Systematic Pullback model described here on a smaller timeframes, for example using it on a Daily/1H?”
In principle, the financial markets have a fractal nature and as such, the patterns that manifest themselves on a Daily chart will also be visible on a 4H chart, a 1H chart, all the way down to a 1Minute chart. However, empirical testing brings fourth differences in system performance that would suggest more of an imperfect fractal nature.
In this article we will illustrate some of the issues of applying Daily models to intraday prices.
The Model in Question: Abraham’s Pullback Setup
In our recent article on trade management we utilized Andrew Abraham’s Weekly Trend/Daily Pullback setup. Here is a reminder of what the setup looks like:
And transferred onto a Daily/1H combo looks like this:
(Black lines are where the MACD conditions neutralize)
By visually inspecting the model, it does seem like there’s some validity to it, and the initial testing does confirm that there is some value in the model. However, the basic model isn’t tradable “as is”.
Initially we applied the basic model with different trend filters. The D1 Ichimoku filter proved to be better than the Daily MACD. So we then tested various exits with the Ichimoku filter.
Sure, the trailing stop methods came out positive over this backtest (Dec 2017 to 9 Jan 2019, i.e. about 1 year). The number of trades is acceptable, but the average profit is by no means “meaty” enough to be traded, alongside the SQN which is still not high enough to be traded with confidence.
Once again, the impact of trade management is clear. The SQN for the Ichimoku filter with both SuperTrend trailing stops is better on the 1-hr time frame than on the daily (SQN 0.56 & 0.51), but caution is needed. The calculation includes the square root of the number of trades in a year, which is much higher than on the daily time frame. The 1-hr equity curves are not as
compelling visually, and also the 2-bar Trailing Stop didn’t perform as well as expected.
These findings make us nervous that the results may not be robust. So what are the next steps?
Additional Common Sense Filters
Trading systems are made up of broadly 4 components:
- asset selection
and all 4 components must come together like a jigsaw puzzle. They must work in unison. Once you have studied an entry strategy based on a repetitive behavioural trait displayed by the markets over & over again, additional filters are usually necessary in order to create a viable trading system (whether mecchanical or rules-based).
These filters basically help to enhance the efficiency of the setup, exclude false signals or enhance the risk:reward of the strategy. Daily systems should be viable with 3-4 filters. Any more than that and you’re probably optimizing/curve-fitting. Once we drop below the daily timeframe however, the number of filters usually increases to 6-8. There is more “fine-tuning” to be done on the lower timeframes, which adds complexity and exponentially increases the risk of overfitting.
It is still a work in progress for us here at FXRenew, but we can definitely share some basic venues of exploration:
- Volatility Filters: these filters attempt to exclude periods of rangebound markets, when the risk of false signals is higher and the P&L would in any case be lower.
- Time/Date filters: these can be significant in FX markets because most times the Asian Session is a consolidation session, London is the trend-setter and the NY session is a reversal session with higher volatility. Depending on the model, it may be useful to exclude signals that appear in certain timezones. Also, trading on the first and last days of the month may not be viable due to rogue flows going through the pipes; a similar consideration might be said of Fridays or when there is important data due.
- Stop Loss Tweaks: on sub-daily frames, it might make more sense to use a combination of profit targets (based on ATR or some other adaptive metric) as well as trailing stops.
Good filters can enhance a system’s parameters by 30% even, but they cannot transform a losing system into a winning system. The underlying logic still has to be valid. Filters are not a panacea.
To Be Continued
Tony & I are still in the process of completing our work on Trade Management using Daily Trading Systems. After our work is completed, we will then dig into models like the Abraham Pullback, and see whether we can find suitable filters that can make a viable system out of them.
It is ver much a work of exploration, to be done with an open mind and always keeping “robustness and common sense” as principal guides.
To be continued…
About the Author
Justin Paolini is a Forex trader & Coach. He is a member of the team at www.fxrenew.com, a provider of Forex signals from ex-bank and hedge fund traders (get a free trial), or get FREE access to the Advanced Forex Course for Smart Traders. If you like his writing you can subscribe to the newsletter for free.
The post The Challenges of Creating Intraday Trading Systems appeared first on FX Renew.