ETMA_The Evolution

Zorro version 1.46

A new Zorro version was just released and can be downloaded from this direct link: http://server.conitec.net/down/Zorro_setup.exe.

Zorro 1.46 got many improvements and new features, for instance a new free Z8 long-term trading system. The complete list of new features can be found at http://manual.zorro-project.com/new.htm.

Traders, Looks a lot like trading Forex These Days, a deadly game.

Out-of-Sample Testing and Optimisation in TradeStation .

In the previous chapter we addressed optimisation tests which are the essential tool for robustness testing. We spoke about the optimisation tests primarily in relation with the sophisticated software platform TradeStation. Another tool which I use very often in this platform and which I perceive as the platform´s major competitive advantage is the Cluster Walk Forward optimisation (CWFO). Before I explain the principle and the greatest advantages of CWFO in some of the next chapters, it is necessary that you fully understand the principle of testing and validation on Out-of-Sample data. It is because that the Out-of-Sample testing is in fact the fundamental basis of CWFO. First you have to understand the difference between In-Sample and Out-of-Sample historical data. When you backtest on historical data, it is always good to set apart a certain period for testing the results´ validity. So if we have a 10-year data set of historical data, we split the data in a certain ratio. Everyone can choose the ratio at their individual discretion. There are many traders who use, for example, the ratio of 70% for In-Sample data and 30% for Out-of-Sample data. Thus if we have historical data covering 10 years, we split it into 2 parts:

70% = 7 years of In-Sample historical data
30% = 3 years of Out-of-Sample historical data

In Fig. 1 you can see the graphical division to In-Sample and Out-of-Sample data.

We have set the ratio of our In-Sample and Out-of-Sample data and now we backtest our trading system on the In-Sample data (the first seven years of data). If the backtesting on the In-Sample data with a certain parameter setting (we explained the parameter values ​​in the previous chapter) gives stable and profitable results, we perform the same backtesting with the same parameter values on the Out-of-Sample data (the remaining three years). In case the trading system is stable and profitable on the Out-of-Sample data as well, we have validated its functionality on unknown data. In other words, the division of data into two samples is a very important step in evaluation of an automated trading system´s (ATS) potential profitability because it allows us to test the system settings on an unknown data sample (Out-of-Sample) that was not used in building of the system. This is called the ATS optimisation. One of the testing criteria can be, for example, the condition that the Out-of-Sample (OOS) results are equally profitable as the In-Sample (IS) results. Given that the In-Sample and Out-of-Sample data always have different percentages (e.g. 70% IS and 30% OOS, 60% IS and 40% OOS, or 80% IS and 20% OOS), we have to standardise the ATS´s efficiency by calculating the profitability per day. Otherwise, for example, the 70% IS results will obviously be always better than the 30% OOS results. The result then show a certain efficiency percentage of the Out-of-Sample results (0% and more). If an OOS data efficiency is 55%, we can say that the ATS has a 55% performance on the OOS data in comparison with the IS data. In TradeStation you can find this percentage ratio under the Robustness Index value. It the Robustness Index reaches 120%, it means that the ATS is more efficient on the Out-of-Sample data. The exact calculation procedure of Robustness Index is not the subject of this chapter. You can find the calculation procedure in Chapter 24.

Let´s add another tool to the In-Sample and Out-of-Sample testing – the ATS optimisation.

In the previous chapter we have explained the optimisation on the example of an ATS based on crossing of two moving averages. Yet there are never enough examples. This time, I am going combine the issue of parameter value optimisation with the In-Sample and Out-of-Sample data division.

We will call our trading system HIGHLOWCLOSE CONDITIONS. To make the example more illustrate, we will focus on the entry conditions only and we leave all the exit conditions aside. The entry condition for the Long side is: If the current today´s Low is higher than the previous day’s Close, then enter the market at the current High by the MARKET BUY order. Now consider a situation that we changed the previous-day-Close condition to a Close two days back, Close three days back, or Close 4 days back. Why should we be interested in trying such an optimisation? Obviously, our goal is always to find the most efficient ATS, i.e. the highest Robustness Index, because our premise is that these systems have the greatest profit potential for the future live trading. The point is that Out-of-Sample data simulate live trading, because it is an unknown data for our trading system. In order to be able to perform a trading system optimisation by In-Sample and Out-of-Sample data, we first have to set some testing criteria and methodology by which we will choose the best In-Sample parameter settings ​​to be validated on Out-of-Sample data. This methodology for evaluation and selection of ATS parameter settings is called the Fitness Function. It may be, for example, the highest Profit Factor, Net Profit, Net Profit/Max Drawdown, etc. It is virtually impossible for discretionary traders to optimise their systems on In-Sample data because it is an incredibly time-consuming process. Yet for a trader using automated trading systems such an optimisation is definitely not an insolvable problem. The trader simply sets the Close variable to X days back and then test various values of X (from 1 to any number of days). If the trader sets the variable X values from 1 to 10, TradeStation automatically performs all optimisation tests (i.e. 10 tests in this example). Optimisation tests showed that the best results of backtesting In-Sample data for the selected Fitness Function (in this example the Net Profit/Max Drawdown) has the setting “Close two days back” because it showed the highest profitability and the lowest drawdown of the 10 tests. Now we apply the HIGHLOWCLOSE CONDITIONS trading strategy with the Close two days back condition on the Out-of-Sample data. Out-of-Sample data substitutes the future data and gives us important information about whether the trading is likely to generate profits in live trading in the future. The metrics for this assessment is the already mentioned Robustness Index. In other words, if a trading system shows good results in terms of Fitness Function in time within In-Sample data backtesting and Out-of-Sample testing confirms the system´s stable profitability, the trader gets a very good signal that the trading system may be robust. You can add the term robust to your trading glossary because you are going to meet with it very often. Robust trading system is a system that is able to adapt to changing market conditions and that has a potential of stable long-term profits. But be careful, robustness does not mean that the system may not have losing periods! The point is that losing periods do not exceed the level predicted and defined on the basis of backtesting. The combination of In-Sample and Out-of-Sample tests gives us a high probability of a successful identification of robust trading systems. Yet it would be foolish to think that all we need is a common IS and OOS data sample. We should definitely employ also another very complex and sophisticated tool available in the TradeStation platform – Cluster Walk Forward Optimisation (CWFO). CWFO performs a huge amount of In-Sample and Out-of-Sample tests with different In-Sample/Out-of-Sample data ratios (e.g. 80% IS and 20% OOS, or 70% IS and 30% OOS) for various time periods. CWFA is such a complex instrument that we will address it in detail in other chapters .

In conclusion, let´s have a look at two equity curves divided into In-Sample and Out-of-Sample data. In Fig. 2 you can see a promising Out-of-Sample test with potentially high Robustness Index and in Fig 3 you can see an Out-of-Sample test which clearly indicates that the ATS is not profitable on unknown data and therefore has a low Robustness Index.

The In-sample test (blue curve) of the trading strategy shows promising results because the equity curve tends to reach new and new peaks. In this case, the Out-of-Sample test (green curve) indicates that the ATS has the potential to generate profits on future data in live trading because on unknown data in the form of Out-of-Sample test it showed stability and correlation (similarity of distribution of profits and losses) with In-Sample data. Of course, we need to carry out further tests via the Cluster Walk Forward Analysis.

The In-sample test (blue curve) of the ATS shows promising results because the equity curve tends to reach new and new peaks. However, the unsatisfactory Out-of-Sample test (red curve) has clearly indicated that the trading system has not the potential to generate profits on future data in live trading.
We explained the principle of In-Sample and Out-of-Sample testing combined with optimisation tests and the basic concepts relating to this topic, such as test criteria and Fitness Function. It is necessary that you get acquainted with these concepts as much as possible because we are going to use them within the Cluster Walk Forward optimization in some of the following chapters.

Trend Following Wizards – June
July 25th, 2016

June was a good month for the Trend Following Wizards, with all funds but one returning positive numbers, with a handful in double-digit territory, including a 27% return for the highest performance. The monthly return average over +5% takes the YTD figure back in the black.

MANAGED FUTURES BOUNCE BACK IN JUNE
July 25, 2016 by Shane Wisdom

This has been the case for my Trading also in both my Trend Following and scalping Portfolio, here is a glimpse of my Results, Notice the upward Equity Curve ( Scalping Portfolio ) in the recent 3 Months.

Two Stocks Walk Into A Bar, Part 1

Trend Following by Scott Welsh.

ATR Breakout Entries.
EXCEL Spreadsheet: JUNE 2016, appeared in the May 2016 issue of Technical Analysis of STOCKS & COMMODITIES.

http://traders.com/Documentation/FEE...adersTips.html

MAY 2016
“Zero In On The MACD"

http://traders.com/Documentation/FEE...adersTips.html

Robustness of the Trading Strategy – How to Measure It?
Published April 27, 2015

When you possess a profitable trading strategy, one of the first things you concern yourself with is how robust your strategy is. There are several ways of testing the robustness of the trading strategy. Before you start reading further, please note that it is not an intention of this article to describe in depth the nature of these different approaches, but to describe briefly only some of them. Most of these approaches have their own limitations and there is no good or bad way. The best trading systems should achieve good results in majority of these tests but it may not always be the case.

Firstly, it should be defined what the robustness of a trading strategy means. The robustness can be described as an ability to withstand or overcome adverse conditions. So when referring to the trading strategy one can say that the robust trading strategy should withstand or overcome adverse MARKET conditions. By withstand or overcome we mean to maintain similar performance characteristics as had been achieved in historical testing without losing a greater amount of capital than expected.

“Practically speaking, a robust trading strategy is one that produces consistently good results across a broad set of parameter (input) values applied to many different markets tested for many years.” Perry J. Kaufman

This means that the robust trading strategy should be profitable in different market conditions (trending x choppy market, bull x bear market) across the wide range of strategy parameter values many years back into the history and also in a basket of diverse markets.

So how can we measure the robustness of a trading strategy?

1) Historical Performance Testing (Backtesting)

If we want our tests to be statistically significant we should always work with a big enough market data sample. By big enough we mean the more the better. When we use market data reaching many years back into the history and do backtesting on them we, in fact, perform the basic robustness test. In other words we are testing if the trading system is able to handle different trading conditions that had appeared during that period. There can be periods of volatile bull or bear market, trending or choppy market and our system must handle all of them. What can happen when we do backtesting only on a few recent years of a market data (or even less) is that throughout this period market conditions remain the same but when we start to trade market conditions change and our strategy stops working.

The amount of market data is related to the number of total trades. Again the higher the number is the better. We cannot declare our test to be statistically significant when we have only 10 or 20 of the total trades.

At the same time we have to care about the quality of the data. These are really important aspects that could skew the test results.

2) Strategy Parameters Stability Test

This type of the robustness test can be performed easily with built-in optimization tools in TradeStation/Multicharts or other software. The purpose of this test is to examine the effect of using different parameter (input) sets on the trading system performance. If the test has revealed both positive and similar performance results across a wide range of parameter sets it is a sign of a trading system robustness. We examine mainly the following characteristics:

A) The distribution of performance across the optimization space – here we want to see a profitable average simulation, a small standard deviation and a small range (max-min) of trading system net profits. We can perform a stress test: Is the average simulation minus the 1st. standard deviation positive? Is the average simulation minus the 2nd. std. deviation positive? And is the average simulation minus the 3rd. std. deviation positive? Using the excel program our analysis could look like this:

http://simplealgotrading.com/wp-cont...est3_thumb.png

* for illustrative purposes

B) The shape of the optimization space – to examine the shape of the optimization space we can use 3D graph in excel or other program by plotting two parameters (inputs) on the horizontal axes and profits on the vertical axis. When we have more parameters than two in our strategy we can perform so called sensitivity analysis where we research an impact of every single parameter on a trading system performance when other parameters are left constant (then we could alternatively proceed with two parameters with the biggest impact on a trading system performance and compile a 3D graph). Generally, we are looking for a flat shape in distribution of profits and when performing sensitivity analysis we do not want to see any parameter to have a big impact on the trading system performance.

http://simplealgotrading.com/wp-cont...s-3D_thumb.png

* for illustrative purposes

3) Out-of-Sample Data Test

This is one of the most common tests used by the majority of traders. It is based on dividing the available market data into In-Sample and Out-of-Sample data segments. The In-Sample data are used during the development process of the trading strategy when the Out-of-Sample (OOS) data are left aside. When the development process is finished we can use the OOS data for testing of how our strategy behaves on an unseen data. The reason for using the OOS data is simple: when developing a strategy based on historical market data we can easily slip to tailoring the strategy to this data. This is often calledoverfitting or curve fitting. This phenomenon causes our strategy to look good on paper but it fails in real terms. Therefore, it is always good to leave some OOS data aside.

Setting the right proportion between the In-Sample and the OOS data is at the discretion of everyone. Most often the OOS data segment is set at 15-30% of dataset. If we have used all of the available historical market data during the development of the strategy then we have to wait to get new data that come up from the market. This can be a lengthy process because we have to wait to get statistically significant number of total trades. However it may be a good option to use both the historical OOS data and the new data coming up from the market for this test. In such case you have to leave the strategy in “incubation” for a period of time and only when it proves to behave as expected you can launch it live.

http://simplealgotrading.com/wp-cont...SS-3_thumb.png

*Out-of-Sample and In-Sample segments of Gold Super Simple strategy

4) Walk-Forward Testing

Our testing is always limited by the amount of the OOS data available. Therefore new methods of robustness testing were created. The most widely used is probably the Walk-Forward Testing (i.e. Walk-Forward Analysis – WFA). Among other methods we can name the Cluster Analysis from TradeStation, which is actually a multiple WFA or Cross Validation method.

The WFA is a multiple In-Sample and OOS testing which walks-forward in time. It is performed in several runs where each subsequent run moves forward in time by a proportion equal to the length of the OOS period. As shown below:

http://simplealgotrading.com/wp-cont.../WFA_thumb.png

In every single run we optimize the trading strategy parameters (inputs) values according to fitness function (e.g. max net profit) and then test the best setting on the OOS data. The purpose of this method is to determine whether the strategy is capable to work on the OOS data in the similar manner as on the in-sample data. If the strategy performs well on the OOS data its performance can be considered realistic and not just the result of curve-fitting. This method can be particularly beneficial if we plan to periodically change the strategy parameters values.

Tools like the Walk-Forward Optimizer in TradeStation can do even more. The so called Cluster Analysis performs multiple WFA in which the results are calculated for different values of the Out-Of-Sample % and the Walk Forward Runs.

5) Cross Validation

This method makes even better use of available data than the WFA. It uses every part of data as the OOS at some point. Not only the data following the In-Sample period.

http://simplealgotrading.com/wp-cont...tion_thumb.png

As a disadvantage of this method it is often given its sensitivity to non-stationarity of the markets (changing market conditions). It is caused by using data way back from the beginning of the dataset for the OOS testing and then incorporating these results into the overall test results. On the other hand the robust trading strategy should be capable of handling the non-stationarity of the markets. Good results in this test can be only beneficial to our strategy. We can presume that if the strategy achieves good results in the parameters stability test it will also end up well in the WFA and the Cross validation test.

6) Multimarket and Multi-timeframe Test

When we are going to perform a multimarket test we must first determine whether the strategy was created to trade on multiple diverse markets or just on one specific market or market sector (e.g. index futures). Both types of strategies are valid. However the strategies performing well on the multiple markets are usually more robust.

If we are testing the strategy created for trading on the multiple diverse markets we will form a portfolio of diverse markets and we will test the strategy on it. To build a highly diversified portfolio we can calculate a correlation matrix of closing prices of markets and then make use of those markets that will have negative or slightly positive correlation. At the same time we should choose the markets that do not have any crucial fundamental connections.

A smaller portfolio of diverse futures markets could for example consist of: e-mini S&P 500, euro, T-bonds, gold, crude oil, sugar, soybeans and lean hogs. For more accurate testing it is appropriate to use a bigger portfolio.

The robust trading strategy does not have to succeed in getting good trading results on a single market but it can achieve great overall results when applied to a portfolio of markets.

If we are testing strategy created for trading on one specific market or market sector we may not necessarily require good results on other markets. But if we want to be more confident that these results are not just a matter of luck we can test the strategy robustness on similar markets. For example, when we have a strategy that was created to trade only e-mini S&P 500 we can test if the strategy is also profitable on other US e-mini index futures (Dow Jones, Nasdaq 100, Russell 2000, e-mini MidCap S&P 400) and on their larger counterpart contracts which have more historical data available. If the strategy achieves good results it is a sign of a certain degree of robustness. However these markets are usually highly correlated and that is the reason why we cannot consider these results as convincing as the results of testing on the multiple diverse markets.

Conclusion

It is not easy to find a robust trading strategy that would fulfill all requirements. The robust strategies can be also associated with high requirements on the trading capital, equipment and background. Therefore, there usually has to be some compromise. For example, the strategies trading on the multiple diverse markets usually employs daily time frame or higher and they hold positions for more then one day. This brings higher requirements on the trading capital and therefore it is very hard to trade this type of strategies with a smaller account and if so then usually with catastrophic results. For the smaller accounts it is more appropriate to trade intra daily (opening and closing positions in one day), but the intraday strategies may be more prone to a market noise and it can be more difficult to develop a robust intraday trading strategy. SO even the size of the trading account can play a key role in achieving a higher level of robustness.

The whole topic of measuring the robustness of the trading strategy is far more complicated. This article has outlined only some of the basic directions in which we can proceed. The robustness will still remain to be one of the main topics of the automated trading.

Author: Richard Palas

Sources:

1) Pardo, Robert. The Evaluation and Optimization of Trading Strategies. Wiley; 2nd edition (February 8, 2008). ISBN 978-0470128015.

2) Clenow, Andreas. Following the Trend: Diversified Managed Futures Trading. Wiley; 1st edition (January 14, 2013). ISBN 978-1118410851.

3) Aronson, David; Masters, Timothy. Statistically Sound Machine Learning for Algorithmic Trading of Financial Instruments: Developing Predictive-Model-Based Trading Systems Using TSSB. CreateSpace Independent Publishing Platform (June 1, 2013). ISBN 978-1489507716.

Walk-Forward Testing of the Gold Super Simple Strategy
Published May 1, 2015 By Author: Richard Palas

http://simplealgotrading.com/wp-cont...mage_thumb.pngThe previous article has introduced several methods of testing the robustness of the trading strategy. One of them is Walk-Forward Testing. Lets try out this type of test on our Gold Super Simple strategy!

Firstly, I would like to say that I have never walk-forward tested the Gold Super Simple strategy before. When developing the strategy I left some data aside and then performed just one Out-of-Sample test (results can be seen here). Strategy was developed on 06/03/2013. Data after 12/01/2012 and before 11/30/2008 were never used during the development process and were left aside for out-of-sample testing.

To test further the robustness of this strategy I will use TradeStation and perform walk-forward testing using the TS Walk-Forward Optimizer. This tool allows us to perform Cluster Analysis as mentioned in the previous article.

In order to make the whole topic clearer, I will divide the whole procedure into several steps:

1) Data preparation and costs assumptions

At the beginning it is necessary to obtain quality market data for the testing and estimate trading costs. In this test I will use TS data of COMEX gold market @GC 60min from 12/04/2006 to 08/13/2014. I will count with 2.67 USD commission per side per contract and assume 15 USD slippage per side per contract. That are trading costs 35.34 USD in total (round turn per contract).

2) Optimization

Before we proceed to the walk-forward testing it is necessary to perform walk-forward optimization in TradeStation. I will perform an Exhaustive optimization because the Gold Super Simple strategy has just 3 input parameters and I will exclude last 30% of data (i.e. from 04/26/2012) during the optimization. (If a strategy has more than just few optimizable inputs it is advisable to use Genetic optimization method).

The results: the In-Sample and Out-of-Sample equity curve of the best In-Sample candidate of the Gold Super Simple strategy according to Profit Fitness Function looks like this:

http://simplealgotrading.com/wp-cont...ofit_thumb.png

We can see that the Gold Super Simple strategy is profitable on the out-of-sample data. The strategy achieved an average OOS score (i.e. Robustness Index Average) 50.78% and WFO Suitability marked by TradeStation as High. However, the Robustness Index Average around 50% is rather small. We would like to see numbers reaching 100%. The reason is that the last 10 months were something like a stress period for this strategy. As it works best in a volatile market and catches all the big moves but the gold market went sideways in the last 10 months. But the strategy is still profitable on the OSS data so I will proceed to the walk-forward testing in Walk-Forward Optimizer.

3) Walk-Forward Testing in TS WFO

Firstly, we have to set up our Test Criteria Settings. How hard limits to set is on judgment of each user. I am intentionally using harder limits than default. For this test I will use the following:

http://simplealgotrading.com/wp-cont...age_thumb1.png

Criteria 1 and 4 are left default.

Criterion 2: Walk-forward Robustness is probably the most important test criterion. In this criterion we are setting up at what rate we require our system to perform in future compared to the results achieved in optimization. For example, when we set up 60% we require our system to perform in future at a rate of 60% and more of the results achieved during optimization. I usually set this limit between 60 and 75%.

Criterion 3: Consistency of Profits is also very important and I set it up also between 60 and 75%.

Criterion 5: Maximum Drawdown setting depends on your level of risk, but 30% should be an upper limit. You also have to set the amount of your Initial capital.

For Walk-Forward selection rule I usually choose between the 1st, 2nd, and 4th option.

In the next step we have to set up the Optimization settings. For a start, it is good to keep the default settings.

http://simplealgotrading.com/wp-cont...age_thumb2.png

Now we can run our Walk-Forward test. It usually takes several minutes.

..

And here are the results of the Gold Super Simple strategy:

http://simplealgotrading.com/wp-cont...age_thumb3.png

Walk-Forward Efficiency

http://simplealgotrading.com/wp-cont...age_thumb4.png

Walk-Forward Efficiency is the most important outcome for me because it measures the robustness of the strategy. The highest cluster average is 68.2% and the matrix average is 61.2%. The Gold Super Simple strategy has passed the Criterion 2: Walk-Forward Efficiency > 60% in 7 out of 9 measurements in the highest cluster (green and red fields). For me this result means that the strategy has passed a lighter test. But an ideal result would be to get the matrix average > 70% and the highest cluster average > 75%.

Consistency of Profits

http://simplealgotrading.com/wp-cont...age_thumb5.png

Criterion 3: the Consistency of Profits > 70% was met throughout the matrix with the highest cluster average 93.3% and the matrix average 89.1%. This is a pretty good result.

Annual Re-optimized P&L

http://simplealgotrading.com/wp-cont...age_thumb6.png

Annual Re-optimized Profit reached 47,475 USD in the highest cluster average and 45,737.49 USD in the matrix average. That is the average walk-forward return of 45.74% p.a. of an initial capital 100,000 USD.

Maximum Drawdown

http://simplealgotrading.com/wp-cont...age_thumb7.png

With the initial capital of 100,000 USD the maximum drawdown went to 21.7% in the lowest cluster average and to 22.5% in the matrix average. The criterion of 30% was met in 8 out of 9 cases in the lowest cluster. The biggest drawdown reached 32.6% of the initial capital.

OVERALL RESULT

http://simplealgotrading.com/wp-cont...age_thumb8.png

Overall the strategy has passed the test with 7 passes in the highest cluster average and has ended up even in the matrix average with 15 passes and 15 failures. This is a decent result. I would like to see slightly higher numbers in the Walk-Forward Efficiency and lower numbers in the Maximum Drawdown section. Maybe an increase of the initial capital would be appropriate but as I have written above that depends on everyones inclination to risk.

Now if you want to re-optimize the strategy periodically in future you can choose the re-optimization interval by selecting the highest cluster central cell (blue one in the previous image) and use the re-optimization schedule.

http://simplealgotrading.com/wp-cont...age_thumb9.png

In the Optimization Summary (In-Sample) tab it is also possible to choose recommended parameters for the next period of trading. But do not take this approach as the only good one. There are different ways of choosing parameter settings and you can also trade without any constant re-optimization.

In the Graphs tab we can examine the continuously re-optimized equities of all tested variants of the OOS% and WF runs. We can also plot the OOS intervals and/or underlying closing prices of the market. Below is shown the equity of the Gold Super Simple strategy (OOS 20%, WFRuns 10) with the OOS intervals followed by the same equity with underlying closing prices of the gold market.

http://simplealgotrading.com/wp-cont...ge_thumb10.png

http://simplealgotrading.com/wp-cont...ge_thumb11.png

Below you can see Walk-Forward Performance Summary of the Gold Super Simple strategy continuously re-optimized in 10 WF Runs with the OOS 20%.

http://simplealgotrading.com/wp-cont...ge_thumb12.png

4) Monte Carlo Analysis

This method is based on a repeated random sampling in order to determine which alternative performance scenarios could possibly arise from the historical data distribution. It is possible to choose between 4 simulation methods: Bootstrap with replacement, Bootstrap without replacement, Monte Carlo with normal distribution and Monte Carlo with random trade order (description of these methods can be found for example in the TradeStation Platform Help). These simulations enable us to see the probabilities of various performance results in categories such as the Total Return, the Average Return per Trade, the Maximum Drawdown or the Longest Drawdown Period. Below you can see the results of the Gold Super Simple strategy using the Bootstrap with replacement simulation method.

http://simplealgotrading.com/wp-cont...ge_thumb13.png

The Monte Carlo Analysis indicates a 95% probability of the Gold Super Simple reaching 301% total return.

http://simplealgotrading.com/wp-cont...ge_thumb14.png

The Monte Carlo Analysis indicates a 95% probability of the Gold Super Simple achieving 0.12% average return per trade.

http://simplealgotrading.com/wp-cont...ge_thumb15.png

With the initial capital of 100,000 USD the Monte Carlo Analysis indicates a 5% probability of the Gold Super Simple reaching the maximum drawdown of 32% (i.e. a 95% probability of the maximum drawdown being lower than 32%).

http://simplealgotrading.com/wp-cont...ge_thumb16.png

The Monte Carlo Analysis indicates a 5% probability of the Gold Super Simple reaching the longest drawdown period of 418 trades (i.e. a 95% probability of the longest drawdown period being less than 418 trades).

Conclusion

The Gold Super Simple strategy passed this walk-forward test with smaller partial deficiencies in all the criteria tested. The results of testing give us a more accurate estimate of the strategy robustness and future performance. The strategy would probably struggle if we would set up a higher limit for the Criterion 2 because the Walk-Forward Efficiency was measured around 61.2% in matrix average. It will be surely interesting to watch how this strategy will succeed in future.

hello did you find that strategy？could you help me？thank you so much.

AUGUST FALL FOR TREND FOLLOWING
September 1, 2016 by Jez Liberty

August was mostly one-sided, sliding down to a strong negative performance, and taking with it the Year-To-Date performance to a similar level. Interesting to note the shorter timeframes weighing on the index while the longer timeframes are still positive/neutral on a 12-month horizon (we do offer trading systems with long-term timeframes).
Note the drawdown level as well, getting a little bit closer to the Max Drawdown (since the start date of the index in Jan-2000). These occasions have historically proven a good time to start investing in a trend following strategy. It will be interesting to monitor the evolution of the index over the last few months.
Despite this recent performance, the index is still well positive (+35%) since launch in live monitoring, in 2013, which interestingly was at similar levels of drawdown. The performance since the start of the backtest in 2000 is at +1,279% compounded.

Systems
One of the main goals of the index is to mirror the performance of trend following in general. As such, the trading strategy is based on simple, public trading systems that use trend following principles. Each system is declined in three different timeframes (long, medium and short) to cover a wide spectrum of trend duration and increase overall diversification.

Something i posted on another forum.
http://www.forexfactory.com/showthre...06#post9140606

Monday, September 2, 2013
Turtle Trading System Automated. Method 2. (Forex Expert Advisor)

The legendary story of the Turtle Trades. Pretty intriguing isn't it? If you don't know about the Turtles and the story behind them, you can read here. My question always was, are the mechanical systems that they used still valid today? Are those systems still profitable? Are they applicable to Forex?

The Turtles traded several markets with 2 systems (System 1 and System 2). These systems didn't have any discretionary rule. Everything, from entries to exits, position sizing to money management, everything was clearly established to the very last detail. This document is the most comprehensive guide I have seen about the Turtle Trading Methods and I used all that information to automate System 2 in Metatrader. I automated System 2 first as it is simpler to program. My plan is to eventually do the same thing with System 1. Let's see if System 2 still works and if it is applicable to the Forex markets, in particular the EUR/USD pair.

The Rules
The system uses the daily timeframe and the rules are surprisingly simple. These are the rules for entering Long positions:

1. Buy when the current price is higher that any other high in the previous 55 days.

1. Put a Stop loss 2 Average True Ranges away from the entry (by using the ATR the system is flexible to current volatility. When the volatility is higher the Stop Loss will be set farther away and viceversa). The ATR used is evaluated over 20 days. So ATR(20).

1. Don't put any Target Profit on the orders. The goal is to let them run as far as possible in your favor.

1. Dynamically calculate position size so that if Stop Loss is hit you only lose 2% of the account. So, if you have a farther away Stop Loss, the positions size will be smaller, and viceversa.

1. Once inside the trade, if price moves in your favor by half the ATR, then add another long trade. And you do this until you have a maximum of 4 open trades in the same direction.

1. Every time a trade is entered, move up the Stop Loss of the existing trades, to the same price of the Stop Loss calculated for the new trade just entered.

1. Exit all the trades when the current price is the lowest price of the last 20 days (This may happen at a profit or a loss). Or exit when Stop Loss is hit.

For short positions the rules are the same but the other way around. You enter when the price is the lowest it has ever been in the last 55 bars and you exit when price is the highest price seen in the last 20 bars. This trading system has all the ingredients recommended by professional Forex traders: it cuts losses short, it lets winners run, it adds to winning positions.
https://1.bp.blogspot.com/-Wmdcef0ab...ble-EURUSD.PNG
Notice in the above picture how all 4 trades were profitable. However, notice how much of the profit was lost. With some optimization that I will explain in this post, the system can be much more profitable and let fewer portions of the profits slip away.

Pretty simple system but incredibly powerful. Also mentally hard to trade. Buying at a 55 day high is tough as that's the point where most traders think the move is done and start fearing a reversal. Same for shorts when you initiate a short position at the 55 day low. Entries are definitely hard to digest psychologically speaking. Exits are even harder. Waiting for the price to reach the lowest point of the last 20 days is painful while you inevitably watch part of your profits evaporate. But this is the best way to ride a trend as far as possible without arbitrary target profits that cut your winners short. This mechanism allows the strategy to ride monster trends once in a while that skyrocket profits.

I ran a mathematical expectancy analysis of the entries and verified that both Short and Long entry signals have a significant positive edge. This is good as a first step to trust the system. Then I coded the rest of the strategy and started to have fun with back tests.

Here's the result of the back test for the EURUSD currency pair from January 1, 2000 to November 14, 2012 (Fully trust worthy paid data from AlpariUK obtained through direct request to the broker). The spread used was 2 pips which is worse than what most brokers offer today.

(Click on image to enlarge)
https://1.bp.blogspot.com/-Q6IWEc485...ults-Forex.PNG
One note here about the draw down. The 31.69% relative drawdown is calculated by Metatrader taking into account the biggest drop peak to valley in the equity curve considering floating profits and losses (of non closed trades) in the calculation. That is why the draw down is reported as a higher number than what it actually is. When considering only closed trades instead of temporary unrealized profits and losses, the Maximum Relative draw-down is 21.34% in almost 13 years.

Some more statistics:
Average Annual Return: 13.19%
Ulcer index: 12.06
Sharpe ratio (Compared to S&P500): 0.70
Martin ratio (Compared to S&P500): 0.89

This performance easily beats the returns of audited professional Forex traders reported on the Barclay Currency Traders Index. A high Ulcer Index of 12.06 reveals what we already know: The system is hard to trade. But who said profitable trading was easy? Also, the Turtle trader must be patient as he will not be trading every single day. The system can stay inactive for weeks waiting for the expected Donchian Channel Breakouts in either direction.

Some further optimizations
The Turtles applied this method using the 55 & 20 days combination to all the markets they traded. The system was not optimized for specific instruments. It is more of a "universal", a "fit all" approach, which is great because it exploits a more universal market inefficiency that seems to work across multiple instruments. It is a more fundamental occurrence and therefore you would assume it is a market inefficiency that is harder to disappear. But that doesn't mean the parameter selection can not be improved for specific markets and that's what I wanted to do for the EURUSD pair.

First I ran an Optimization test to make sure that the Parameters space provides good results even though parameters were drastically changed.
https://1.bp.blogspot.com/-sGtOycX75...s-Method-2.PNG
A coarse optimization was run only for two parameters (days for entry signal and days for exit signal). The great news is that the Parameters space is very solid and profitable across the board (All green). Meaning the 55, 20 set is not the only profitable one. Any combination of days for the entry between 40 and 80, along with any selection of days to exit between 4 and 20 consistently yields positive results. Which is great because it means that even if you are not playing with the most optimal parameters, you still have a high chance of seeing equity growth over the long run.

Once I was confident the Parameters space was that wide and good, I ran a Rank Analysis for the parameters selection. The idea is to obtain the parameters that yields the most stable draw down values and the most stable yearly returns. It looks like the 70 day entry with an 8 day reversal exit is the most stable parameter set, which offers the most stable draw-down values yer after year along with the most stable returns. Here's the back test using the 70 & 8 combination.

(Click on image to enlarge)
https://1.bp.blogspot.com/-ZNISIiEhd...ults-Forex.PNG
Now the maximum drawdown according to MT4 (which as I said, considers floating unrealized profits and losses) is 25.20%. But in reality it is only 15.66% considering the equity curve based on closed trades.

Average Annual Return: 18.33%
Maximum Drow-Down: 15.67% in 13 years
Ulcer index: 8.98
Sharpe ratio (Compared to S&P500): 0.74
Martin ratio (Compared to S&P500): 1.77

This is, in my opinion, a far superior version. The essence of the system hasn't been changed. We simply came up with parameters that consistently deliver superior results and that are likely to be profitable going forward regardless of market mutations.

If you think this trading system can be a good addition to your trading arsenal, consider the purchase of the Expert Advisor for only $67. Much less than what is typically charged for unprofitable, over curve-fitted unprofessional EAs out there that can't survive a month of live trading. As part of the purchase you get an Installation and Configuration guide, plus my personal support setting it all up if needed.
Turtles Trading System 2
Expert Advisor and Pdf Guide $80

Buy Now

Beyond any doubt, the Turtles Trading System 2 has a positive edge. It is profitable and applicable to the Forex markets (you can test it on other currency pairs as well). Although profitable, the method is psychologically hard to trade. The hardest part is letting profits run indefinitely without a predefined profit target. That factor explains why some Turtles were not profitable in spite of the fact that they were all taught the same system. Letting your profits run, as originally sated is what ultimately separates the great Forex traders from the amateur ones.

Thanks for your interest. I hope you enjoyed this article and I also hope that if you make the Turtle Trading System EA part of your trading arsenal, you can benefit from it. For any inquiries and support of any kind feel free to contact me at [email protected].

Update
As a proof that Long term Profitable automated trading is possible, the Turtles Trading system continues to deliver good results after this document was written. +35.6% in 2014, +22.8% in 2015 Read the articles below for more details:

Results - 2014
Results - 2015

Tuesday, April 12, 2016
Impatient Sniper - A Profitable Gold trading strategy

A complete Gold trading system with positive expectancy. Metatrader Expert Advisor included.

In order to diversify my trading, over the past couple of years I have been researching strategies to effectively trade Gold. Gold is an extremely emotional market that can move several points in a day in one direction, and in the same way it can quickly give it all back. These wild rides, the noise and volatility make it sometimes harder to trade in a discretionary fashion. There is a huge emotional component that ends up hurting Gold traders more often than not. That, and the fact that it is a 24-hour market, makes it more suitable in my opinion for fully automated trading through a robot, a.k.a expert advisor.

At this point, I have developed three Gold trading strategies with a positive edge: a Trend Following strategy, a Volatility based breakout strategy that uses Bollinger Bands and a Counter-trend strategy. In today's article, we'll explore the Trend-Following strategy.


STRATEGY DESIGN

Using pure price action and no indicators, we can determine that Gold is in an uptrend if today's closing price is higher than every closing price over the last X days. We can also say that Gold is in a downtrend if today's closing price is lower than every closing price over the last X days.

I started testing this simple hypothesis with X values between 50 and 120 to determine if there was momentum in favor of the trend, by measuring Maximum Favorable Excursion (MFE) and Maximum Adverse Excursion (MAE) after all the entries. This allowed me to quantify a positive edge. For more information on calculating MFE, MAE, what they represent, etc, read How to measure a system's edge.

I finally settled on 100-day break-outs, as they provide a solid and clear edge. Meaning, if Gold closes today, higher than it has closed in all of the previous 100 days, then it usually has some upside momentum that carries it higher over the next few days. The beauty of this is that, this is not only true for 100-day break outs, but also for 80, 90, 110, 120 and all the values in between. That's to say, we have a reliable entry parameter that just "didn't happen to yield positive returns in the past". If all the surroundings (80, 85, 90, 95, 105, 110, 115, 120), or even some of them had yielded negative results then our 100-day entry rule would not be reliable and we would just be curve-fitting our system to past history, with fewer chances of survival in the future.

After this, I decided to set a Stop Loss at two times the daily range using the Average True Range indicator (ATR). This distance is usually enough to give your trades enough room to work out the noise while at the same time we protect our capital against unlimited risk. 2xATR is a common rule applied by many mechanical trading systems that use the daily time-frame and it was pioneered by the popular Turtles Trading system. A 2xATR move against our original assumption is usually a lost cause. Better to take a manageable loss at that point and move on.

So, now, we have a 100-day breakout entry combined with a standard 2 ATR Stop Loss. Time to think about exits. With the LT Trend Sniper system, we sometimes follow trends for 2 months or more. The Average holding period is 21 days. For this strategy I wanted to do something different. I wanted quicker exits, which is why I thought about a simple time-based exit and started testing automated exits 4 days, 6, 8, 10, 12, 14 days after the entries. All the time based exits showed positive results in the 12 year back-tests performed (2001-2012), which is a great sign. By choosing a simple 10-day time-based exit, we are not choosing a parameter value that just "happened to be" profitable in the past. We are indeed choosing a value in the middle of a sea of profitable possibilities, with good solid returns also in the surroundings to withstand future market mutations. By the way the 10-day exit DOES NOT yield the best returns in the back-test, but it does have a solid neighborhood which is what I like the most about this particular value. There is no guarantee that the parameter value with the greatest return in the past will continue to be the most profitable one in the future. The main concern is always the robustness and future reliability of the system.

In addition to the 10-day exit, the trades are entered with a 2:1 reward to risk ratio. In other words, if during the 10 days of life of the trade, a reward that is twice the risk is achieved, then we will be satisfied and will close the position for a nice profit. Therefore, upon entry, the trade is set with a Target Profit (TP) of 4 times the ATR. So, now we have Stop Loss at 2 x ATR and Target Profit at 4 x ATR. By the way, here, once again, we obtain positive returns regardless of whether we use a 1:1, 2:1 , 3:1 and other reward to risk ratios (3:2, 5:2, etc) , which is again a great sign.

With these simple rules all that was missing was our capital allocation. I wanted to keep a constant risk per trade so that results are not skewed by a few occurrences with greater weight. Because the Stop Loss distance is dynamic, our position size is also going to be dynamically calculated. Meaning, if the Stop Loss distance is small, we can afford a slightly larger position size. However, for larger Stop Loss distances we have to use smaller position sizes. This way we keep a constant risk throughout all the trades. For more information on this, read How to risk a percent of your account per trade.


SYSTEM'S RULES:

Buy GOLD on a 100-day closing high.
Short GOLD on a 100-day closing low.
Stop Loss at 2 x ATR
Target Profit at 4 x ATR
Close the position 10 days after the entry if neither SL nor TP have been hit.
Risk 5% of the account per position. Meaning if SL is hit, the account loses 5 percent. Of course, this risk level can be reduced for more conservative traders.
Never trade two positions at the same time. If a new signal is triggered while in a position, it simply means that the trend is progressing in our favor. We won't increase our capital exposure at that point. Instead, we just ride the original position.

Ok here are the results of this simple, yet effective strategy from January 1, 2001 to December 31, 2012. We are assuming a 0.40 point spread throughout the test. (Most brokers offer better spreads than this nowadays. Anywhere from 0.20 to 0.35 is common). The simulation starts with a $100,000 account balance:

(Click on image to enlarge)
https://1.bp.blogspot.com/-iGhNke3a5...Day%2BExit.gif
(Click on image to enlarge)
https://1.bp.blogspot.com/-uuxE6I5jF...ld-results.PNG
Not too shabby. The strategy achieves a +9.83% Average Annual Return, which easily beats the S&P500 Index and the Barclay's Currency Traders Index. The worst draw-down in the equity curve was 19.06% and it took place from October 2008 to February 2009. Of course, by using a smaller than 5% max risk per trade this draw-down can be reduced, but at the expense of also smaller Average Annual Returns. Still, with a 19.06% max draw-down, the strategy easily beats the 55% draw-down suffered by the S&P500 benchmark during the financial melt-down of 2008.


Reducing the magnitude of losses

Despite the positive initial results, we can still explore more approaches to control our draw-downs. For example: Trailing Stop Loss mechanisms. Now, usually, when you bring Stop Losses closer, you expose yourself to getting stopped out more often. As a consequence, you start taking smaller losses BUT more frequently and this may result in a null net effect in many cases or even worse outcomes. For this reason, we must be careful and not trail stop losses too aggressively. We will always keep a 2 x ATR Stop Loss distance, and we will move the Stop Loss by 0.5 ATR whenever the position moves in our favor by 0.5 ATR. For example, if our original Stop Loss distance was 32 Gold Points (ATR = 16), then we will move it by 8 points once the trade has gone in our favor by 8 points and we will keep doing that indefinitely. Once the trade has moved 2 ATRs in our favor, we have moved the Stop Loss four times and it is now at the break-even level, the same price at which we entered the position. As the trend progresses in our favor, we keep trailing the Stop Loss level in 0.5 ATR increments, or 8 Gold Points in this hypothetical case.

Let's look at the results now with the addition of the Trailing Stop Loss mechanism.

(Click on image to enlarge)
https://1.bp.blogspot.com/-Ac-B-k1SH...Day%2BExit.gif
(Click on image to enlarge) https://1.bp.blogspot.com/-Ga7l3bHOt...-stop-loss.PNG
A much nicer equity curve now and improved statistics. Average Annual Return is now +10.94% and the Max Draw-down ever suffered is 15.94%. (Meta-trader over-estimates it with 18.17% because it assumes higher equity peaks and larger losses as it considers open positions' balances). My preference is to only consider closed positions in the results.

This is now a more respectable trading system with very few degrees of freedom and a very low degree of curve-fitting to past data. We have a strong parameters space with many combinations of entry + exits and reward to risk ratios, all of them yielding positive long term results, which speaks of the robustness of this approach and the huge mutations the Gold market would have to endure in order for this simple trending principle to entirely stop working in the future.

How did the system perform in the Out of sample period (January 1, 2013 - January 1, 2016)?
Well, here you are:
(Click on image to enlarge)
https://4.bp.blogspot.com/-YrPaZXQIp...Day%2BExit.gif
(Click on image to enlarge) https://3.bp.blogspot.com/-GB8XGk0Hz...-trader-EA.PNG
In the three-year Out of Sample period, the account grows +25% and behaves within the expected parameters. This is important, because this period was not taken into account at all during the initial design of the strategy. In other words, this is Gold data that the system had never seen before and it is what would have resulted from trading the strategy live during those 3 years.

Combining both the Back-test and Out Of Sample periods we have 15 years of trading history from January 2001 to January 2016.

(Click on image to enlarge)
https://4.bp.blogspot.com/-IrKoELjD_...Day%2BExit.gif(Click on image to enlarge)
https://3.bp.blogspot.com/-LLtagWG0v...stem-rules.PNG
Beautiful and rather smooth equity curve. The original investment is quadrupled in 15 years, representing a +10.29% Average Annual Return. Max draw-down is 15.94% (closed positions only, not open floating profits or losses). Notice how the Average winner is much larger than the Average Loser and the system wins about 50% of the time, resulting in an excellent 2.01 Profit Factor.

As my free contribution to the trading community, in particular those who specialize in Gold trading, feel free to download the Impatient Sniper (Gold Only) for free.
LT Impatient Sniper (Gold Only)
Expert Advisor for Metatrader
Strategy design, Back-tests and Configuration Guide

Free Download


Exploring Other Instruments

Once the system was entirely conceived, I started to wonder how the strategy would perform in other instruments. Same break-out principle, same 10-day exit, same Stop Loss distance and 2:1 Reward to Risk ratio. It was a very nice surprise to see this system showing solid results in instruments other than Gold. The simple edge it exploits is more universal than I thought.

With the Euro for example, we obtain even better results than with Gold. We have good results whether we trail the Stop Loss or not, and whether we trail it indefinitely or just until the break even level and not beyond. Here is a 16 year back-test from January 1, 2000 to January 1, 2016:

(Click on image to enlarge)
https://2.bp.blogspot.com/-2WrD85aP9...BBreakEven.gif
(Click on image to enlarge)
https://3.bp.blogspot.com/-NONl40O6C...rt-advisor.PNG
Solid +13.94% Annual Return (AAR) with a Maximum Draw-down (MDD) of only 13.60%. A great 1.02 AAR to MDD ratio.

So, now if we want to diversify our risk a little more, we can combine both GOLD and EURO in the same portfolio. GOLD and EURO are not historically strongly correlated instruments. They don't inherently move in the same direction all the time. This is great as it provides us with some built-in risk diversification in a context of a trading strategy with a positive expectancy. By combining both instruments, the draw-downs in one of them can be slightly mitigated by positive results in the other instrument. Of course, we also have the risk of increased draw-downs if they happen to coincide in time.

Here's the GOLD + EUR portfolio. 2000-2016:
https://1.bp.blogspot.com/-Zu03CxJOb...strategies.PNG
https://2.bp.blogspot.com/-PO7mW-v1_...profitable.PNG

Here is the break-down of returns per year:
https://2.bp.blogspot.com/-q5WNRI213...er-results.PNG

A 23.37% Average Annual Return is nothing to sneeze at. It turns a $100,000 account into 2.48 millions in 16 years and this is without ever injecting additional fresh capital into the portfolio. It comes however, at the expense of a painful 23.35% draw-down and a period of 536 days during which the portfolio failed to reach new equity highs. It is therefore, a psychologically difficult system to trade. Imagine going through a 20% account draw-down, more than a year without achieving new equity highs and it just looks like you are wasting your time. Profitable trading hurts. For more on that, read my article The Torture.

Personally, I prefer less aggressiveness when trading more than one instrument simultaneously. Using a3% position size per trade instead of 5% better fits my pain tolerance. This results in still nice +14% Average Annual Returns, but with a much more manageable worst historical draw-down of less than 15%.

This is a nice example of a simple long term profitable system. I will consider it in my own trading and will follow up on the results at the beginning of every year.

Thanks for reading, and I hope that by studying this system you become a better trader whether you end up using it or not. It might help you explore and develop similar ideas on your own, and if that is the case, I would be glad to hear from you in the future.

Cheers,
LT

Notes:
- Back-tests run assuming a 0.40 Bid-Ask spread in Gold and 2 pip spread for the Euro (0.0002). Both numbers are pretty unfavorable by today's standards.

- The data was purchased at forextester.com and it is 1-minute data. No need to purchase tick by tick history when we are dealing with long term trend following strategies that use distances in the hundreds of pips for both Stop Losses and Target Profits, unlike other strategies such as intra-day, short-term scalpers, where tick data becomes more necessary.

- Back-tests performed on the Metatrader 4 platform, build 950, powered by GoMarkets, an Australian Forex broker regulated by the Australian Securities and Investment Commission (ASIC)

- Portfolio simulations performed through the Asirikuy Draw-Down Analysis Tool, version 1.41a.

- No affiliation whatsoever with any of the previously mentioned sources. They are just enumerated for clarity so that the reader knows exactly where it all came from. No Recommendation.

Becoming a systems designer.

In order to become a systems designer, you need to understand the following concepts:

 

1. What a historical statistical edge means and how to quantify it.
2. What curve fitting is and how to avoid it.
3. How to implement rock solid risk managent in Forex by using dynamic stop loss distances according to the current volatility of a curency pair and consequently reduce position size or increase it depending on the stop loss distance.
4. Understand and perform advanced studies such as Walk Forward Analysis and Rank Analysis for parameters selection with greater likelihood of positive performance in the future.
5. Understand worse possible draw downs of your system in terms of percentage as well as days by using advance tools such as Monte Carlo simulations. (Ultimately, this ability is the one that will help you stick to the system during the inevitable bad times, because you'll know what to expect and you'll know when there is enough evidence to assume that the edge exploited by the system has dissapeared).

Almost nobody talks about any of these things when discussing specific strategies. And that's sad.

Cheers,
LT

the-lazy-trader.com and all its products are Copyright by The Lazy Trader and property of Lazy Trading LLC. All Rights Reserved.

The Lazy Trader posts are great content Bass. There is enough there to spend a lifetime on.

Something i posted on another thread.
http://www.forexfactory.com/showthre...64#post9152564