Working on my trading strategy. The strategy is a trend-following one, like I have been attempting to trade for a long time. My assumption is that cornerstones include identifying trending conditions, spotting pullbacks and betting on when the pullback is/has come to an end.
Am trying to articulate these cornerstones in a rigid and objective manner so that I can run the subsequent rules through years of historical price data to determine trades, entries, exits, duration, excursions, frequencies, win rates and profitability.
Sharing some of work in progress and thoughts on this blog post….
Automated back-testing vs manual back-testing
The testing methodology I am attempting to use now is not the same as paper/demo trading in a simulated environment such as forextester. In contrast it moves closer to the type of testing undertaken by systematic traders. (Systematic trading, at its core, is an automated manner of trading, where setups are identified, executed, managed and exited without human intervention.) Systematic traders are generally also skilled programmers who use a raft of applications to test their strategies.
For my situation, the sample size of completed test trades will be significantly larger (thousands as opposed to hundreds) with my “large scale Excel-based testing” compared to “simulated environment testing”. It will also be time efficient and easy to test a range of what-if refinements and thus determine ways for making the strategy better (whilst of course being aware of the risk of over-fitting.
Am also very conscious to avoid strategy-hopping. Trader1961, one of trading mentors, has warned me that this is a common trap, where aspiring traders give up on a trading strategy when they hit issues and move to a new idea, as opposed to grinding it out and trying all the angles to make it work. I guess there is a fine line between giving up too early and knowing when it’s time to give up on a strategy.
Identifying trending conditions
Onto the first cornerstone…. identifying trending conditions. The key at this point is not finding entries/setups, but simply for determining whether the instrument is trending or ranging. Here’s a 1H chart of EURUSD for part of February 2016:
Here is the same chart again, this time including exponential moving averages (EMA) for 10, 20 and 100 periods:
Which chart is easier to evaluate and assess?
The top chart focuses on market structure – on identifying swing points – and utilises conepts outlined by Michael Voigt in the German text Das grosse Buch der Markttechnik. Incidentally, Carlos Diaz, who runs a trading for oil futures live trading room and also Foresightfx.com, is from the Americas and most likely doesn’t speak German, approaches the charts in a very similar manner to Voigt, just using completely different terminology. I like the manner in which both of these guys analyze the market. The past few weeks I have joined Carlos in his trading room and witnessed first hand how he executes trades using this framework on the 10M and 15M charts. Additionally I am finding this approach lends itself well to back-testing.
The bottom chart is very reflective of the way I have been analyzing the charts the last couple of years. There are pros and cons to this approach. For three years I sat next to a retail prop trader who pretty much only used these exact EMAs (together with two common indicators) for milking the DAX on a daily basis – it definitely worked for him. He taught me a lot about how to use them but I seem to have hit a wall with those. TraderDi, who has been coaching me for the past year, also is a fan of these moving averages (whether simple or exponential).
Alternatives I have looked at in recent days include the zig zag indicator. Discussed the pros and cons of this with one of the systematic traders here at the hedge fund. It’s great to be able to sit down with these guys and get some input. I am going to keep this one nearby – for now, I am thinking the testing would be more difficult with zig zag compared to market structure, not least because the zig zag values continue to re-calculate themselves as they go along.
Quantifying the pullback
The second cornerstone relates to identifying valid pullbacks. This is something where I have benefited significantly from observing Carlos trade oil futures. Trying to quantify the pullback by expressing the price action seems more logical, and far easier to formulate, than working with EMAs. The quantification is done by looking at the candlesticks (see zoomed-in chart, again on EURUSD). Here’s what I am working with at the moment:
- Lower Lows without Lower Highs
- Count of Lower closes (x out of last y)
- Lower Low Count (x out of last y)
- Decline at least x pips/ticks [where the measurement is volatility-adjusted]
- Decline at most x pips/ticks [where the measurement is volatility-adjusted]
In terms of volatility-adjustment I am planning to use ATR for the 1H timeframe for say the last 50 periods. Using the standard setting of 14 periods (used on daily timeframe) is too choppy – I need a much smoother line.
Thankfully, one of my trading buddies wrote a script for me that allows me to take information from Excel (such as trade entry and exit points) and automatically draw these as trades on my MT4 trading platform. This is a huge time saver and allows me to assess how the test trades look on the chart.
Let’s see how much I can progress with this today….