Where we left you in the Prior Post of this Series
In our prior Part 1 post of this series we left our readers with an ominous warning that needs to be heeded in our retail world. For those traders seeking to apply the techniques of the diversified systematic trend following community to the retail environment of CFDs and Forex markets…..then there is an added complication to our little niche that makes this game a bit more problematic. Namely the higher relative costs associated with trading the retail products that attempt to mimic the cross asset class exposure achieved by the global diversified players in the industry.
In particular, as we are seeking to exploit the ‘outlier’ trends that can have very extensive holding periods, the ‘cost drag’ associated with SWAP (interest holding costs) on leveraged instruments are prohibitive requiring that our edge (before costs) needs to be sufficiently great to more than compensate for the cost burden that impacts your PL.
Now unfortunately when we apply a ‘classic’ simple trend following model across our entire test universe of 63 markets (which possess fairly wide cross asset exposure), while we see that on a ‘before cost’ treatment. a clear favourable bias exists towards ‘trading the trend’, the bias is just not sufficient to generate positive expectancy on any possible permutation of parameters used for this very simple trend following system.
The overall design simplicity of the chosen test system means that if we applied some other form of simple trend following system to catching the ‘big’ outlier trends…such as a moving average crossover technique, Bollinger breakout or Simple Moving average system…then we are still likely to run into the same obstacle where the cost loading’s are too great to allow for us to enjoy a ‘house edge’.
So we are left in a pickle. Does this mean that we abandon trend following and become seduced by more ‘predictive methods’ such as mean reverting methods that are shorter term in nature and possess those great but shorter term linear equity curves? Well not so fast “Poindexter”….that’s quitter talk and we know that those other methods are like playing darts with a blindfold.
Before we contemplate throwing in the towel we just need to unpack this riddle a bit further by examining how well our simple trend following system fares within different asset classes themselves. Perhaps there are some asset classes that are simply better than others in which we can apply our arcane Trend Following magic to?
This post…being Part 2 of this series will now do just that looking at Spot Forex first. Our later posts in this series will then examine different asset classes.
So let’s see what we can uncover.
Our assessment of this asset class was undertaken using 29 Spot Forex markets available through Dukascopy which comprised 7 Majors, 5 Minors and 17 Crosses.
Let’s have a look at the performance results of a range of different possible Parameter values using our simple trend following system across this entire asset class.
Results – Spot Forex – Individual Parameters
Ouch……what a mess. We are not going to get much joy here. Have a look at the ‘Bias on GP Column’ for a ‘before trading cost’ assessment versus the ‘Bias on NP Column’ for an ‘after trading cost’ performance assessment.
Stop Loss ATR (SL ATR)
On a before cost basis…the bias is not that evident. There is no defined clustered ranking appearing in the series of values chosen. The best value was an initial stop of 4xATR with the next base value being 10xATR. Not much inference can be made here apart from the fact that there is no definitive bias in the series present when looking at SL in isolation. But now look at the ‘Bias on NP’ column. Notice here how we now have a preferential ranking from the best being 10xATR, then next best being 8xATR and then next best being 6xATR. This preferential ranking is simply a result of the dominance of trading costs in the overall result where the better results are a result of a lower average cost burden per trade. Now if there was a bias that provided a degree of positive expectancy…then great…but in this case while there is a bias in the after cost results, the bias is still unfavourable in overall profit outcome.
Trailing Stop Loss ATR (TSL ATR)
The bias on the ‘before cost series’ is a bit more evident when looking at the trailing stop condition. There appears to be a preference towards a tighter trailing stop condition. This before cost bias…if it bears any definitive substance…therefore holds a bit of useful information worth pondering in relation to Spot Forex markets.
Having a preference towards a tighter trail gives an indication that perhaps the trending condition in Forex markets is more ‘stunted’ where we see a bias arising with the Law of Large numbers that prefers shorter holding periods than longer holding periods for this asset class. Not surprising really considering that trading spot Forex is a result of trading a pair of currencies where the spread relationship between the pairs oscillates between strength and weakness of the underlying currency pair relationship as opposed to adopting a preferred long or short bias over time that we may find in commodities and Equity Indices for example.
Now this pondering and navel gazing really doesn’t help us however as the overall bias ‘before costs’ is still simply not sufficient to give us an overall house edge when trading trends in this entire asset class.
Remember as much as we would like to infer some causal factor to this bias….in using quantitative methods we simply take note of the bias itself. We are not after a narrative. The causal implications actually are not important in playing this game…but what the data says is important to how we go about potentially exploiting any preferential bias.
Donchian Period Look-back (Donch)
Now have a look at the bias on a ‘before cost basis’ for the Donchian look-back. That is interesting. A clear bias exists towards a very long look-back. Even stronger when looking at an ‘after cost basis’. This is starting to become a very common story with efficient markets where a definitive edge is hard to find. I would like to conclude with a narrative about the noise and mean reverting tendency of modern efficient markets such as the highly liquid Forex market….but I have to stop assigning meaning to this stuff. Leave that to Tolstoy.
Break-even ATR Multiplier (BE Multi)
Mmmmm…still a split biased series. Not much can assist us with this parameter at the moment.
Results – Spot Forex – Combined Parameters
In terms of possible permutations using combined variables (of Initial Stop, Trailing Stop, Donchian Look-back and BE multiplier) there are a total 3000 different possible combinations which we will then rank in terms of overall profit performance when applied across the entire sub-universe of Spot Forex.
So what we see here is that of the total possible 3000 possible permutations of combined Parameters. then the top 50 performance results by overall profit across the Spot Forex asset class possessed a mean Donchian look-back of 790 and median of 900 daily bars. That is a pretty long look-back and supports what we find when we look at the bias that exists in the Donchian Look-back as an individual parameter.
The top 50 result also possessed an initial stop of approximately 6-8xATR, a TSL of 2xATR and a BE Multiplier of 4-5xATR. This sits within the range of the bias we observed in our analysis when assessing the bias for these Parameters in isolation.
Let’s now fast forward to the meat of the story in this Forex asset class……as you are probably getting bored by now.
In fact….now that you can see the process we are adopting here we can get to the conclusions quickly without having to go through this ‘painful slow process’ of narrative description. No one ever said this was a fun exercise.
Major and Minor Forex Currencies with Positive SWAP Only
Ok….so we have fast forwarded now to where we can now actually exploit some of the significant bias that exists in the data that assists our trend following cause.
This is where the ‘house edge’ exists in the Forex universe where the bias is in our favour and exceeds the impact of the costs of trading.
Unfortunately however the preferential bias that we now observe is one relating to cost structure as opposed to the nature of Forex markets themselves.
In fact according to this analysis there is no sufficient bias on a pre-cost basis in the Forex universe that favours trend following models. What this means is that there are no universal trend following models that we can apply to this particular asset class that over the last 15-20 years demonstrates that allow for overall positive expectancy.
There may be individual markets such as EURUSD, GBPUSD, GBPJPY that suit trend following due to their particular unique market characteristics…but there is no universal model that can be applied to this entire sector to reap any bias in the entire Forex universal series of this test.
So then we are left with a well known trading practice called ‘the carry trade’ which is a feature of the bias that exists in the cost structure itself as opposed to the nature of the market condition to give us our slight house edge. Provided that there is an overall positive SWAP (interest paid on holding) for relevant Major currency pairs and Minor currency pairs, then this contribution to overall profitability is sufficient to generate overall positive expectancy for these simple styles of trend following system.
But be warned the beneficial bias associated with the carry trade that sufficiently augments the power of our trend following models only lies in the Major Currency pairs and the Minor Currency pairs. Avoid the Cross currencies as the favourable bias of positive SWAP is still not enough amongst the Crosses to gives us that house edge to apply our technique.
Here are the results and I will let you make your own conclusions….but as a hint…. focus your attention on the ‘Bias on NP’ column. You will see where the favourable clusters lie.
Also refer to the average profit per trade column and you will see that we have overall positive expectancy across the series of Majors and Minors with positive SWAP. Then also cross-check the individual Parameter results against the Combined Parameter series of Top 50 results.
Once you absorb this analysis you will then be able to conclude that there is actually a clear House edge to exploiting the bias associated with the ‘carry trade’ principle using trend following methods providing a long average hold.
This bias is significant enough to allow us to focus on the shorter Donchian look-back periods in our decision for when to jump on board the trends. Neat eh!
So here is the central takeout from trend followers looking to take advantage of a general bias in this asset class that favours trend following which has been relevant for the past 15 to 20 years of market condition..
If you want to get best bang for buck in your diversified trend following models in the Retail space using Forex pairs available from retail brokers, then consider at least limiting your universe to those Major and Minor currency pairs that offer a high positive SWAP making it beneficial to use simple trend following systems that offer long average holds.
It is the ‘carry principle’ of the preferential interest holding costs that provides a general bias in your favour within this particular asset class.
The overall edge that lies in the market data itself in this particularly efficient and very liquid but noisy asset class makes it less favourable to trend following methods and perhaps makes it more favourable towards other styles such as mean reversion.
Unless you have strong positive SWAP on your side….then there are better asset classes to focus your energies on to achieve diversification benefits when the bias is on your side.
Now this is a feature for us Retail traders seeking to trade Spot Forex. We cannot conclude that this is a general feature of trading all currencies such as Currency Futures. We haven’t undertaken an analysis on Currency Futures and so our conclusions must remain restricted to Spot Forex with Retail Brokers only.
Now don’t get upset….as there are more parts coming to this series and there is a happy ending to all of this for us Retail traders which do in fact mirror some of the statements made by many of the successful trend following FM’s who trade the trend in a diversified and systematic way…..but until we get to these posts we won’t pre-empt.
Stay tuned for a happy ending to this mind numbing crunching of extended data sets. I even find it hard to write this Blog as my head occasionally crashes onto my writing desk while trying to keep it all light and airy 🙂
Trade well and prosper
The ATS mob