And now to the Lessons Learned
We have been trading five strategies with the FX Autotrader Elite from 24 April 2014 to 21 October 2014 using Version 03.01.01. These are the Bonus Bundle SET files that were provided with the Autotrader upon purchase.
Overall the portfolio of strategies has done well returning 29.4% with a drawdown of only 5.1% over
this six months period – as shown below. A monthly return of 4.5%, well within our target range of 3 to 5%.
But we need to delve a little deeper into the performance to see what we can learn and how we can improve.
Lesson 1: Overall insights
- I have run this portfolio on two platforms – and I get different results. Why? Because from time to time when broker connections fail, trades are missed. Nothing we can do about that.
- I have also tried closing all trades on portfolio-wide TP and SL basis, but those results were worse than letting each strategy run on its own TP and SL levels.
- And I have confirmed that for these strategies, trading during the Asian and London sessions only is far better than trading all day, 24/5. Compare the results below! Profit of $1400 trading the Asian and London sessions only compared to $200 if we had traded all day, 24/5!
Lesson 2: The importance of TP and Partial Close settings
There are two elements to developing strategies:
Selecting the right indicators and settings to determine entry points is relatively easy – can be done visually on the chart and best done over several periods of trading history. We then have a choice of letting the Exit condition close out the trade and/or setting TP and partial close levels also. The problem with the former is that Exit conditions normally materialise after peak profit has been reached.
So to overcome giving back some of our profits we try to estimate appropriate settings for TP and partial closes. But partially closing profitable trades at low TP levels can lead to bad results since losing trades will close as full trades leading to a situation where we can be profitable in pips but lose out in dollars. Please read the User Guide for a detailed explanation of this.
So how did we do with the Bonus Bundle strategies in this regard? Take a look at the comparison charts below where I was trading 0.1 lots (ie trading at approximately $1 a pip). So a 1000 pip profit should roughly equal $1000.
But on this portfolio we made 4900 pips profit but only $1470 in real money, a ratio of 0.3 $ per pip. The results for each pair are shown below. The best ratio for $ to pips is GBPJPY at about 0.4 and the worst is GBPUSD at 0.07. We are profitable in dollars overall and for each strategy but things could have been better.
The worst performer in this respect was the GBPUSD (H1) 191 strategy. It has a partial close of 80% at 40 pips ($32) and SL averaged around 30 pips (-$30) when trading with 0.1 lots. Full TP at 100 was rarely reached and so the remaining 20% of the trade was closed at 0 in most cases. So with 52% wins and a 1:1 effective risk:reward, we essentially broke even.
Remedy? If we were to keep this GBPUSD strategy I would set TP to 60, no partial close,
trailing stop (TS) starting at 30 and trailing 20 pips in 10 pip steps for Initial trades and no
trade Re-Entries (refer Lesson 3 below). Trailing can be done using the DIY Trade Manager
Plus or with the built-in TS feature of the new Autotrader version, out soon. But for myself, I
am going to design a new strategy for this pair on the H1 timeframe mainly because a 52%
win rate is too low for my liking. The USDCAD strategies on the other hand had around 80%
winners so optimising Trade settings for these pairs seems more worthwhile.
So we need to optimise TP and partial close settings – not a simple task. We can do it manually by scrolling back through the charts to see if we can find better average settings; or we can run some trials in parallel as I describe in the Planning Guide part of the FX Autotrader Elite User Guide.
We could also of course use the MT4 Strategy Tester to try and optimise these settings. But there are several drawbacks to that approach. Unless you import and use good tick data the results will be meaningless. Even if you do, it does not mean that your broker will replicate such results because they do all sorts of funny things to the data and orders like slippage, freezing feeds etc and of course there will be connection issues when trading live. And lastly, the Autotrader with all its logic and rules performs extremely slowly on the Strategy Tester. It’s something a colleague of mine is working on, but for me its visual back testing and forward demo testing.
For example, considering a H1 chart, if we set SL at 40, or if 40 is the average SL from Exit condition (dynamic SL), then we should close 80% at 50 pips in profit. (80% of 50 = 40 and so equal to SL). Since 50 pips is a fair distance on a 1 hour strategy, we might set a BE at 40 and a TS of 20 or 30 also starting at 40 to cover the instances where the trade reverses before reaching 50. But we need to remember the advice of “giving your trades room to breathe”.
Also of course, we should examine recent history of the strategy on the charts to see if such levels are reasonable.
Lesson 3: Initial and Re-Entry trades
It’s worth examining the different performance of Initial and Re-Entry trades and consider using different trade settings for each. After all, a re-entry will normally occur only at some time during the latter portion of a trend and hence a smaller TP compared to the Initial TP setting may be justified.
Below is the data for the five strategies or SET files. Let’s discuss what we have here.
If we look at the EURJPY strategy, Magic 151 (Initial) and 1511 (Re-Entry) we see that 151 made 24 trades for a profit of $168 ($7 per trade) while 1511 made 16 trades for $145 ($9 per trade). Both Initial and Re-Entry trades have the same trade settings.
When we also examine the actual trade results (which you can also do with the FX Blue Publisher EA analyser) we can see that if we had traded without partial close for Initial trades our net profit would have been $240 better because 4 of the trades went the full distance – and we would have had 24 trades for $408 net profit, not just $168 ($17 per trade).
On the other hand, re-entry trades never reached the full TP settings and so perhaps we should just do a 100% partial close at 40 pips or keep the 80% partial close and have TS for the remainder.
For the GU strategy, 191 and 1911, only 191 was profitable so we could either set Re-Entry lot size to 0 for 1911 or make other adjustments to see if we can make it profitable.
As an aside, a careful examination of the charts shows that several of the bad re-entry trades were triggered on MA cross-over conditions in accordance with the rules (MA1 > MA2 for a Buy and MA1 < MA2 for a Sell) but at times when the MA1 Line had actually started to turn over and was pointing in the opposite direction to the trade entry. In other words, the trend had already started to reverse. This issue has been addressed in the new Version of the FX Autotrader Elite.
Lesson 4: Hard Stop Loss vs Dynamic SL
A good rule of thumb for trading is to have a maximum risk per trade of 2% or less. When trading 0.1 lots ($1 a pip) on a $5K account that means a SL of 100 pips.
However in these SET files I have used dynamic SL (ie relied on Exit conditions) rather than having set any fixed or hard SLs. The average SL that has resulted for all the trades using the dynamic SL approach was 32 so well within the 2% risk limit. But out of the 215 orders/trades there were four large losses of over 100 pips, two resulting in more than 3% loss each.
The problem with hard SL is that we can get spiked out even though the trend stays in place. A dynamic SL does not suffer from that same fate because the Exit condition takes a few bars to materialise and so if it just a one bar spike, the Exit condition will not be triggered. But if it does get triggered in a few bars time it could be at a far bigger loss. So it becomes a compromise.
But, as I point out in the User Guide, a hard SL is a good safety feature in case your platform loses internet connection with the broker. When that happens, a dynamic SL setting will not be triggered because the connection has been broken, whereas a hard SL will be triggered because that SL order was put in place with the broker when the trade was first opened.
Lesson 5: Too many Break Even Trades
The win rate of 60% for this portfolio is on the low side but that is caused partly by the fact that Break Evens (BEs) are classed as a Loss. An examination of the trade history shows that we had many BE trades.
A BE is better than a loss but having so many BEs begs the question as to whether a TS and perhaps a BE function with a small TP step would have improved our results. It would certainly improve the % winning trade statistics and it could well have resulted in more profit.
I say “could have” because a TS can also close out trades early on a temporary retracement that would otherwise have gone on to make a large TP.
New Improved FX Autotrader Elite – Version 03.01.10
Analysis is a key aspect of trading and we need to respond to the results. Some of that response can be done with the existing Version of the Autotrader. Other responses require modifications to the current Version. And I have been working on that.
Version 03.01.10 will be available in November and all existing customers will receive this upgrade and a revised User Guide for free. I also plan to do some more training videos, and ones with better sound quality (!!) than the current ones available at www.diyforexskills.com/videos
Key features of the new Version are:
- All actions are currently triggered only when the conditions have been met at the close of the candle or bar. In the new Version we can select whether we want this or whether we want the action to be triggered whenever the conditions are met, and independently for Entry and Exit.
- The BE function now has a Profit Lock setting. So instead of the BE closing at 0 pips, you can set it to close the BE at say 5 pips.
- A stepped Trailing Stop (TS) function.
- Three multi timeframe (MTF) filters for entries based on MA crossovers to enable you to take trades only in the direction of the trend on higher and/or lower timeframes. One of the MTF filters can also be used for exits and can hence be run on a shorter timeframe than the one your EA is trading on, and thus act as a TS or early exit signal.
- The zlMACD module can now have separate settings for Entry and Exit and each can be set with two levels. Eg you could enter when zlMACD histogram is above zero and Exit when the histogram is > than some level or < zero. (This is similar to how the RSI module works).
- The MA of RSI and the double MA crossover of RSI settings have now been separated so that you can use both of them in a strategy rather than only one or the other.
- The rule for MA crossovers has been amended so that two conditions now need to be fulfilled. MA1 must > MA2 for a Buy (as we have now) and MA1 must be sloping upwards.
The opposite for a Sell. This extra condition is especially important for avoiding bad re-entries.
Existing SET files will need to be modified to work on the new Version. A revised Bonus Bundle of
SET files for this new Version will be supplied with the upgrade.