Statistical Report #3
All Expert Advisors are different for they are based on different trading strategies. Even working under similar strategies, Experts can differ by their parameters. Irrespective of the strategy used and parameters, it is possible to estimate Experts for their risk level and stability. Risk level shows what part of its deposit the Expert is ready to lose if the situation develops unfavorably. The Expert Advisors' risk level will be estimated by two factors: maximum amount of lots in positions opened in the same direction and StopLoss levels. Stability of an Expert Advisor shows how steadily its deposit increases. To estimate this parameter, it will be necessary to evaluate the maximum drawdown tolerated by the Expert. Risk Level
One of the factors that determine the Expert's risk level is the maximum amount of lots ever opened by this Expert. If the Expert opens several positions in the same direction, the maximum amount of lots will be equal to the total volume of trades. If the positions are opened in different directions, the difference between their volumes will be calculated. ![]() Fig.1. Maximum amount of lots in positions opened in the same direction. The riskiest Expert, in terms of open positions volume, is ldamiani. It has had up to 10 lots in open positions in the same direction. It is followed by Hendrick, Aver, MAEstro and others. The maximum volume of these Participants' positions is limited with 5 lots. Experts alexgomel and payday have the smallest values, they have never opened more than 1 lot simultaneously. The lot price is not considered when calculating the risk level. It varies depending on the currency pair chosen for trading. Thus, having the same volume, Sashken, that trades on GBPUSD ventures nearly 1.5 times more money than Aver on EURUSD. One more factor that influences the Expert's risk level is the level of its StopLoss orders. This factor defines how much the Expert Advisor can lose if the market goes unfavorable. In the diagram below, the burgundy line represents StopLoss. For descriptive reasons, the level of TakeProfit is shown, as well (blue line). ![]() Fig.2. Averaged StopLoss and TakeProfit, in pips, for all positions of the Participants. Some Expert Advisors do not place StopLoss, but use TakeProfit (for example, sashken). Others, on the contrary, do not place TakeProfit, but use StopLoss (for example,
alexgomel, MAEstro). Experts vgc and paydaycan also be subsumed to this latter group since their TakeProfit is placed at a distance of 4087 and 999 points, respectively. The triggering of such TakeProfit is highly improbable. We can consider that TakeProfit is not placed, in this case. Most likely, the developers place such stop orders as some signals. It must be noted that some Experts place neither StopLoss nor TakeProfit. These are bvpbvp and BMG, for example. It is possible that they do not trust in stop orders and use their own mechanisms of positions closing. To evaluate the total risk level, additional calculations were made: an index was obtained that showed what part of its deposit each Expert could have lost at triggering of its StopLoss. ![]() Fig.3. Expert Advisers' risk levels. The risk levels of ldamiani and MAEstro exceed 100 percent. This means they can be closed by StopOut at an unlucky train of events. Experts that do not place StopLoss can be subsumed under the same category. These are bvpbvp, BMG, sashken. The most Experts's risk levels vary from 25 to 50 percent. This is rather much, too. The least risky Experts in the given list are Fonz.de.Cool7, fizzleboink and via. Their risk levels do not exceed 15-20%. Most of such Experts are not superprofitable, but they are able to increase profits without significant losses. When developing risk management systems, it is necessary to consider the current balance and the currency pair. Without this, Experts' stop orders will only be able to calm the author, but not really control losses. For example, Expert GUMASA by Luis Damiani (ldamiani) places StopLoss for all its positions, but if they all trigger, the Expert Advisor can lose its deposit. It is rather difficult to relate an Expert's profitability to its risk level. There are both "super-risky" and very "prudent" ones among the profitable Expert Advisors. Taking into consideration that all the listed Experts are profitable, the most preferable risk level would be 20-30%. However, this parameter can vary, in each specific case. StabilityExchange rates change all the time. This results in changes in profitability of open positions and equities of the Expert Advisors. If the equity falls below the preceding maximum, this phenomenon is called drawdown. The smaller it is the more stable the Expert is considered to be.
Relative drawdown is a ratio between the maximal drawdown and the value of the corresponding local maximum of the equity. This coefficient shows what losses, in percents of equity, are acceptable for the Expert Advisor. More details about how to calculate the above coefficients can be found in the article named "What the Numbers in the Expert Testing Report Mean". Unlike the Strategy Tester, we use in our report the equity as of the end of day to calculate drawdowns (in Tester, the balance after each trade is used for this purpose). ![]() Fig.4. Relative drawdown. It can be seen in Fig. 4 that different Experts have different drawdowns that vary from 2 to 57%. Experts plj, richplank, RobinHood and alexgomel have the smallest relative drawdowns that do not exceed 5%. Experts sashken and GODZILLAhave the largest drawdowns, over 50%. Drawdowns of the most other Participants vary in the range of 15 to 35%. Of course, trading is not possible without losses or, at least, drawdowns. However, one has to define clearly what drawdown level is acceptable for the Expert Advisor. Thus, for instance, the relative drawdown of the Championship leaders, ldamiani and Hendrick'а, makes just 15-20%. One more important index is the maximal drawdown. It shows the highest non-fixed loss ever happened to the Expert.
Fig.5. Maximal drawdown. It can be seen in Fig. 5 that there is a small group of Experts, the drawdowns of which has never exceeded 1000 dollars. These are plj, richplank, RobinHood and alexgomel. Drawdowns of the most Participants made 2000-6000 dollars. In spite of such data, they managed to gain profits and get into the top page of the list. Moreover, the Championship leaders are in this group, too. The Experts with large drawdowns are seen in the diagram very clearly, too. These are: sashken with the drawdown of 16285, vdiddy38 with the drawdown of 11508, and maje with the drawdown of 7923. If these drawdowns were happened on the first trades of these Experts, they would unambiguously be closed by stopout. However, these Experts have first made some profitable trades, so their equities do not fall below the threshold of concern. Due to this, they were able to continue the competition. The next index is the absolute drawdown. It shows by how many dollars the equity has decreased compared to the initial deposit. Absolute drawdown of some Experts is 0 though they have rather large maximal drawdowns. This means that the Experts have large profits for all the drawdowns. Their equities did not fall below 10 000 dollars.
Fig. 6. Absolute drawdown. Using the data obtained, we can calculate the "stability coefficient" of Expert Advisors. For this, it is necessary to find the ratio between the Expert's profit and its maximal drawdown. The higher the coefficient is the more stable the Expert's work is. For example, if an Expert has stability coefficient of 10, it means that this Expert can have profits 10 times exceeding its maximal drawdown.
Fig.7. Experts' stability: ratio between the profit and the maximal drawdown. It can be seen in Fig. 7 that the diagrams of stability and relative drawdown are in an inverse dependence. This means the smaller the relative drawdown is the more the Expert is. Vice versa, the higher the relative drawdown is, the less stable the Expert is considered to be. Experts ldamiani and Hendrickare leaders of the Championship now. They have the largest equities. However, they cannot be called leaders in stability. They run some risks to gain profits, which allow them to take first places. In the other hand, the most stable Experts (richplank and plj) are still far from the prize places. These Experts are rather cautious and close their positions after they have gained some profit. Both approaches are surely worth to be used. There is no unambiguous opinion about which one is better. They both have their pros and contras. Summary:
Created: 2006.11.29 Author: MetaQuotes
P.S. Oh, about the depreciating, biased comments, I think each one can take his
own conclusions, I just wanted to make some remarks about the whole thing.
2
ldamiani
2006.12.01 15:32
“When developing risk management systems, it is necessary to consider the current balance and the currency pair.
Without this, Experts' stop orders will only be able to calm the author, but not really control losses. For example,
Expert GUMASA by Luis Damiani (ldamiani) places StopLoss for all its positions, but if they all trigger, the Expert Advisor can lose its deposit.”
“Ldamiani has made something, but his Expert Advisor cannot be really used in the market.” I would usually not be bothered with such comments, the problem here is that this is a public and very popular site, so I must say a couple of things so that the ones that are still not used to magic tricks can look to the hand where the real thing is happening. A competition like ATC is more difficult than the real market. That’s because besides all the uncertainty involved in real trading, a contest like this adds extra challenges: payoff is limited to only 3 participants, rigidly defined time period, trading size, etc. Given the rules, participants must do their best and bring in all the knowledge and experience they have gathered in order to WIN (and not promote products, the only ones that have this right here are the sponsors). And that’s the biggest worth of such public competition, cause people can find out what’s possible to be achieved in such challenging environment. With that spirit that I tried to reengineer GUMASA, obviously extra risk was assumed, but always playing by the rules (and never complaining about them). Every competitor adopted a different money management strategy (usually much riskier than usual), but the good news is: trading strategies (=trade choice + entry point choice + exit choice) can still be compared! That’s to say, its possible to dissociate money management strategy (bet sizing) from trading strategy. For that purpose we can use an indicator called Total Normalized Profit (TNP). Here is how to calculate it: 1) Choose a set of EAs that you want to compare and copy there Account History to a spreadsheet. Delete the “buy stops” and the “sell stops” orders. 2) For each trade calculate its Net Profit = swap + profit 3) Calculate the trade’s Normalized Profit = Net Profit / Lots 4) Total Normalized Profit = Sum of all Normalized Profits What is the meaning of TNP? TNP iquals to the sum of the profits of all trades as if they were made with 1 lot, so we have just get rid off the money management strategy. Comparing the TNP of different Experts Advisors is a fair way of comparing edge each one had during that period of time. I calculated it below for some EAs from the ATC for the period of 10/1/2006 to 11/30/2006:
That’s it, no magic, you know how to calculate it, so you can judge for yourself. May the odds be with you !! investor_me wrote: We will do our best to include the stats (amount of trades made by EAs) in our next
statistical report.I wish there was a statistic about the number of trades for each EA and whether there were scalpers. We can say that there are no scalpers among 25 EAs analyzed in this report since the largest amount of trades as of the 20 Nov. was at free-day (46). Approximately, 1 trade a day. 2006.11.30 09:27
I wish there was a statistic about the number of trades for each EA and whether
there were scalpers.
2006.11.29 17:35
Really enjoyable article. Hopefully there will be a next championship with a category
for the most stable expert. Thank you.
|
|








