Statistical Analysis

Analysis goal

This is the statistical analysis of RobinVOL 2.0 based on the official 13 year backtest. I will try to be very thorough on how to do the analysis so that you can recreate it by yourself. Our goal with this analysis is:

To understand how RobinVOL 2.0 works
To understand the losing periods and the winning periods, and know how to deal with both
To understand how to decide if a losing period is just a temporary drawdown or if the EA has lost its edge and must be stopped
To understand how to configure the risk
To know what to expect in terms of return and risk

Basic statistics extracted from the backtest

This is the basic statistical information extracted from the RobinVOL 2.0 backtest:

Explanation of the main statistics:

Average annual return: Represents the return rate that RobinVOL 2.0 averaged in all 13 years and a half. It averaged a return rate of 72.37% every year.
Maximum drawdown %: Represents the deepest losing period that RobinVOL 2.0 had during all the test. To achieve the Average Annual Return of 72.37%, it had to keep trading during losing periods of -12.58%.
Maximum drawdown period in days: Represent the longest losing period of RobinVOL 2.0. You can see that it had a 10 months long losing period until it reached a new equity high. We will deeply analyze the drawdown behavior of FOREX RobinVOL 2.0 later.
AAR / Maximum drawdown ratio: Represents the overall quality of an Expert Advisor. Shows how much risk the Expert Advisor took to achieve the Average Annual Return. This value can be used to compare the quality between different Expert Advisors.

Exposure to high efficient markets

Regardless of general statistics, there are periods in the market where efficiency is high and any trading strategy reduces its edge and increases its losses. That can be caused by events that are impossible to predict.

As the trade frequency increases, any strategy is more susceptible to this efficiency periods of the market (such as August/2011). On this analysis, we calculated how much would be the drawdown if we trade one of such periods of 4 weeks, where we lose 80% of the trades:

So, during a four week long high efficient market where we have simulated losing 80% of the trades taken, we see that RobinVOL 2.0 is able to preserve the capital and lose only 10,74% of the total balance. Compared to the Average Annual Return, we see that it can recover from this loses very quickly.

More basic statistics

First we have the main ratios that are used by the industry to compare the quality of trading systems. I am putting them here to be able to compare with other systems:

Below you can see some more interesting statistical information about RobinVOL 2.0:

First, you can see that the Overall reward to risk ratio is very good. As we saw in the backtest, the average size of winning trades almost doubles the size of losing trades. This is very good as it help to recover from drawdown periods very quickly.

The win to loss ratio is 48%. That means that a bit more than half of the trades taken end as a lose. This is due to cutting losses very quickly if market goes against us.

The Profit factor is the profit generated by profitable trades divided by the losses generated by losing trades. In essence it represent the money made on winning trades compared to the money lost in losing trades. The higher, the better. This ratio is not as good as the AAR/Maximum Drawdown we saw previously to measure the quality of an Expert Advisor, but can be used as a reference as it is a value that appears on any Metatrader 4 backtest. You cannot use the Profit factor ratio to compare between different Expert Advisors.

On average, RobinVOL 2.0 took 7.8 trades per week. Please, note that there will be weeks without trades and weeks with a lot of trades. This is just an average.

RobinVOL 2.0 shows a very stable behavior in terms of monthly profit. It has faced winning strikes of up to 12 consecutive months. But it is much more important to focus on losses:

RobinVOL 2.0 has faced in the past two consecutive losing months. Statistically this will happen again, even worse periods with three or more losing months (we will focus on this later on the Montecarlo Analysis). One must be prepared to trade through this drawdown periods to be able to achieve the potential earnings.
RobinVOL 2.0 has faced in the past up to 19 losing trades and in the future it will go through this kind of losing strikes or even worse.

It is very important to set the risk level so that this periods do not affect us psychologically thinking too early that the EA does not work and stopping it or lowering the risk just before the recovery period. This is the main reason why most people with good strategies lose money on FOREX.

You will see many times how RobinVOL 2.0 has a profitable basket of trades opened and suddenly it ends closing them at a loss. You have to be confident in that this is the statistically proven optimum behavior. Sometimes this happens, but in the long term, the benefits of letting profits run are much bigger than cutting profits short.

Making money should be boring. If you feel excited or worried while trading, you are risking too much.

Balance curve

Here you can see the logarithmic balance curve from 2000/01/01 to 2012/11 of RobinVOL 2.0 (the linear balance curve is the one shown in the backtest). A logarithmic equity curve is not constant in the Y axis, but logarithmic. It has the ability to see the drawdown periods better hiding the effect of money compounding.

Looking at the equity curve we can see that there was periods with good slope and periods where the equity increase is more flat. But in general, RobinVOL 2.0 was able to make money on all market conditions during 2000/01/01 to 2012/11. Remember that this does not mean that it always wins. It means that in the past we had constant positive years.

Looking at the equity curve you see that there are dips and losing periods. In the curve they might be small, but remember that there can be more two consecutive losing months, probably more in the future.

No one can predict what will be the market conditions in the future. The only thing we can do is to test our strategy on as much different market conditions as possible. The fact that RobinVOL 2.0 is able to make money even in the first years of the backtest where the volatility was low and the EURUSD was so different than it is now shows the high robustness level of the Expert Advisor.

Drawdown analysis

This is the most important section of the analysis in my opinion. It is essential to understand the drawdown periods of FOREX RobinVOL 2.0 if you wish to make money trading it.

Here is the drawdown periods sorted by length of RobinVOL 2.0 from 2000/01/01 to 2012/11:

The most apparent thing in the graph is that there is a single period of 300 days long that faced a 12.3% drawdown. If we go deep and analyze the causes, it was a single losing month followed by a slow recovery in sucesive months.

The rest of drawdown periods are less than 5 months, so I would say that it is not common in RobinVOL 2.0 to have long drawdown periods.

The conclusions I get about this graph are:

To trade RobinVOL 2.0, one must be prepared to trade through similar (or worse, we will see later) drawdown periods without reducing the risk and with confidence that it will eventually recover and make money as it did so many times in the past.
Drawdown periods are very usual. The majority of the days trading RobinVOL 2.0 are days inside a drawdown period. This is very important to understand. Losses are cutted short, so when market is adverse, RobinVOL 2.0 will have losing days and winning days, while our balance don’t grow or we even lose some money. But when good conditions come, it makes enough money to cover any small losses and reach a new equity high.

There is no way to avoid this. This is how RobinVOL 2.0 works. If you try to avoid the drawdown periods, you will not be there for the recovery period and you will end up losing money.

This is very important. If you are not prepared to trade through the drawdown periods described in this document, please, don’t buy this Expert Advisor.

All trading strategies in the world are doomed to fail in the future. The most robust strategies might last for many years or even decades. But even strategies with very serious effort in robustness such as RobinVOL 2.0 will eventually degrade it’s performance. We will see later in the Montecarlo Analysis section how to identify this and consequently identify when to stop trading it.

For most people, 16% to 20% drawdown is the maximum they can psychologically accept. Please, be sure to set the risk of RobinVOL 2.0 so that you will not be worried while trading through a long and deep drawdown period.

It is important to dedicate some care to the risk settings in the beginning and don’t let fear and greed influence you.

Profit analysis

Here is the chart of annual returns (in red) compared with a strategy of buy-and-hold in the S&P500 (in orange) of RobinVOL 2.0 during the period of 2000/01/01 to 2012/11.

We can see in this graph that the distribution of earnings is stable, specially on the recent years. We can see very good years with returns of 117% and 173%, and discrete years with just a 25%. One thing to note is that obviously, the increase in volatility of the markets in recent years is very good for this strategy.

On this chart we can immediately see that the majority of months were winning months, and that winning months were bigger than losing months.

We can see that the average monthly return is near 5%. You can see that sometimes there are two consecutive losing months and that easily could be three or even four and that losing months are not rare at all. Every 4 or 5 months there was a losing month on average.

On the worst months FOREX RobinVOL 2.0 lost around 10% of the trading capital. Contrasting with the losing months, there was incredibly good months too with earnings of almost 25%.

You have to realize that losing months can happen at any time. Even the first month you trade this strategy. This is the reason why you must be sure you know how FOREX RobinVOL 2.0 operates and it’s statistical characteristics described in this document.

Before trading it live, be sure you understand this section and this whole document about how FOREX RobinVOL 2.0 works. Do not hesitate to ask us whatever you don’t understand.

Montecarlo Analysis

Montecarlo Analysis is extremely important, as it focus not on how FOREX RobinVOL 2.0 traded in the past, but on how it will probably trade in the future. And we make money in the future not in the past.

If we classify the trades of the EA in terms of outcome, we can calculate the probability of getting each kind of outcome. With this data, we can run a simulation called Montecarlo analysis that generate the same number of trades as the original backtest of the Expert Advisor but with random outcomes complying the trade classes described above.

In essence, what we get is other possible outcomes of the EA within similar kind of trades.

We run 100,000 simulations. I give much more importance to the information obtained from Montecarlo simulations than the information obtained in backtests. The reason is that a backtest is just one possible outcome obtained from all the universe of outcomes that the statistical characteristics of the EA can generate.

The data we obtain from the Montecarlo simulation is the following:

The first important thing to note is that of all 100,000 iterations, the average drawdown of all but the 5% worst is 17,09%, and the Average Maximum Drawdown is 11.25%. So the interesting conclusion is that the backtest is not a statistical oddity. There are very good probabilities that FOREX RobinVOL 2.0 will perform according to the backtest in the future.

We will talk about the Worst Case values later, as they are extremely important to identify when the strategy has lost it’s edge in the market.

We see that on average in all simulations, the Average Compounded Yearly Profit is 58.85%, which is close to what showed us the backtest (67%). The last two values compare the Expected Annual Return with the Worst case drawdown and the drawdown of the first 95% simulations. In the case of FOREX RobinVOL 2.0, those values are very good (2.08 and 3.44), which gives us a lot of confidence on the robustness of the strategy.

In other words, even trading near the worst case performance, we should keep making money.

Here you can see three example simulation results (a good one, an average one and a bad one) from all the 100,000 we ran in the Montecarlo Simulator.

On all pictures, the red line represents the result, the purple line is the average of all iterations and the green line is the worst case values (both purple and green lines are the same in all three simulations, only the red line changes):

The green line is extremely important for us. It is the line that separates the worst 5% simulations. Above the green line happened 95% of the 100,000 simulations launched.

As I said many times, all strategies in the world are doomed to fail in the future. It can take decades for the most robust ones to start degrading or just some weeks for the weakest strategies, but all, with no exception, will end performing bad.

We will use the green line to identify if the strategy has lost it’s edge and decide what to do then (stop trading the strategy, trade other configuration, reoptimization, etc.).

The fact that RobinVOL 2.0 makes money even if it were trading on the worst case is something I like a lot. This means that the strategy has a solid edge, is able to adapt to many market conditions and it will be very difficult for the market to break this strategy.

Looking at many different iterations, I can see that the possible scenarios are not too wide. This is good, as the possible results you will get live will probably not be very different than the one from the analysis.

So, how would I know if the strategy is in drawdown or it just stopped working? One good way to answer is: stop trading when our EA is trading near the green line (worst iteration in 100,000).

On the table where we had the Montecarlo results, the green line is the Worst Case Drawdown Scenario. For the risk specified in the backtest, the Montecarlo Analysis would tell us that if we get a drawdown of 28.2% we should stop trading the Expert Advisor and rethink what to do, as we would be trading outside the statistical boundaries considered normal (95% of the iterations).

As we are trading concurrent trades, I think it is wise to give a little more room to the Worst Case Scenario. Instead the 28.2% I would personally prefer to set it at around 35% for this risk settings. But this level is very personal. It is your business decision.

This worst drawdown level might be too far for some traders, so as an early warning we can use the “Average DD% values higher than 95% of cases”, which in our case is 17.09%. At this point of drawdown we should start to watch carefully how RobinVOL 2.0 is performing. Until that point, we are in the normal zone. From 17.09% to the Worst Case Scenario we are in the warning zone.

Once we reach the Worst Case Scenario, it is your responsibility as a trader and business manager to take a business decision with all the information available. We may reach the Worst Case Scenario due to a bad weekend gap against us caused by a news event or maybe just the inefficiencies this EA exploits with this settings have been erased from the market.

So even if you are facing a 25% drawdown of your account (if you set risk=2) you should keep trading it without modifying the risk settings, as that is statistically normal and it should recover from the drawdown and reach a new high.

One interesting way I like to see this is: “Once I reach 28.2% profit, I will be in a free ride. Until then, the money on risk belongs to the market.”

Conclusion

This EA has all the characteristics I need to classify an Expert Advisor as long term profitable: trading at the close of the bar, robust default settings, good risk to reward ratio, is not a scalper and it has a very profitable parameter space that makes FOREX RobinVOL 2.0 very robust against market changes (you could put almost any random parameter settings and it will probably manage to make money in the end).

I like that in the end it will probably be able to end generating money even in the worst case, but one has to be prepared psychologically to keep trading it keeping the same risk during the drawdown periods.

Final words

It is a bit strange to analyze my own Expert Advisor. I tried to be as unbiased as possible, following the same analysis procedure as I did on many other Expert Advisors as you can see on DonnaForex site. I tried to be as explicit as possible so that you can generate your own analysis.

With FOREX RobinVOL 2.0, I tried to create an EA based on a solid edge so that I would be confident to trade my own money.

I put a lot of effort on robustness, implementing all my knowledge to make FOREX RobinVOL 2.0 trade in the future in a similar way as it traded in the past.

There are very few Expert Advisors on which I trust to manage my personal money. I invested a lot of effort to make FOREX RobinVOL 2.0 one of them.

Fernando Monera