How parameters of EDGE trade expert can be optimized

When working with EDGE it is important to keep in mind that using as many data for parameter optimization as possible is recommended. This is substantiated by the fact that for consistent results one should undergo a test with as many examples of price recoilless movement as possible. Why more of those examples are required? Because the structure of unidirectional movements is very varied and the only way to select consistent parameters is to go through as many of such examples as possible. EDGE trade expert contains all the necessary settings to select such combinations and to consider all the unpredictable outcomes. Signals are calculated on the current chart timeframe, on which the expert is downloaded. You can use it on any timeframe, but before that, one must find optimal parameters for that timeframe.

Absence of required resources for finding such combinations may turn out to be one of essential hindrances on your way. For doing this, you need a computer with a big volume of RAM as well as a many-core processor which cores will be engaged maximally during parameter optimization in strategy tester. Those, who can afford assembling a network of computers which will simultaneously participate in optimization, will have even a greater advantage. Also, using MQL5 Cloud Network service from MetaQuotes would be a great opportunity. For a reasonable payment one could perform optimization, using resources of thousands of computers globally.

It would be a mistake to think that for five-minute (M5) or one-minute (M1) timeframes, one or two years of data would be sufficient for searching optimal parameters. A “the more - the better” rule works here. There is never too much of data. The algorithm must be tested on all the data available and it is about hundreds of thousands of transactions. If the trader has serious intentions to earn, instead of losing his deposit, his job is to get prepared for it maximally.

Why do we need more data? This is caused by the fact that on the time range selected for optimization, the required examples on which problems begin, simply might be not available. As we have stated already, there are recoilless unidirectional movements of various types. You can make sure in it by yourselves, simply analyzing test results in visualization mode, looking in detail through the process of deal making on a multitude of recoilless movements. Suppose, you have chosen a time range for optimization for the last year. However, this period could be calm enough, with prevalence of lateral price movements. Even if there were trends, they were with quite big rollbacks. There is no difficulty to set the expert on such price behavior. But this is a mistake to think that it will contiue this way. Immediately after the optimization period, prolonged recoilless unidirectional movement can start and your “optimal” settings will be not relevant at all. Due to that very reason, forward-testing is not the proper method to use for checking and searching optimal parameters. This method is always worse than that used when you conduct optimization on all the available data. Many cannot pass this test because they fail to get interesting results on large series of trades.

As well, it is important to realize that historical data with many brokers might have different properties. Even the same broker and on the same type of account may have different properties of data. For instance, on several years in the very beginning of historical data there may be spreads of less size. Whereas, the last year of data and on the same symbols may contain spreads of bigger size. We observe a slant towards the period when spread was less. And the earlier optimized parameters do not already fit to new data. Therefore, parameters must be optimized more often. Even the moments when manipulations over historical data are carried out must be taken into account. For this purpose, you must study all the available data provided by the broker with whom you are going to open an account for trading and further to make double checks for changes.

Unfortunately, historical data do not contain such changes as stop / limit levels. This can also have substantial effect. To reduce such effect parameters may be optimized on larger timeframes. In such a case, the best option is to choose the broker having more minute data for building the timeframe we need, as well as real tick data for final tests. Besides, the best broker for you will be the one having no stop / limit levels, no commissions and spread is minimal on the plenty of various symbols. The perfect variant for parameter optimization would be a possibility to use the same symbols but from different brokers with different conditions. Probably, this is even the only way to obtain really consistent combinations of parameters for various types of trade algorithms. But, unfortunately, for now it is difficult-to-implement. It is quite possible that such feature will be available in future versions of MetaTrader 5.

Parameters of EDGE expert allow to set both conservative and aggressive trade types. If you want to trade with minimal risks and plan to attract investors using Signals service or PAMM-account, then your choice is conservative trade type. If you prefer quick deposit growth and know when currencies are in a long flat (lateral price movements), your choice is aggressive trade type. You can find such combination of parameters where trading with strong risk you will double the deposit for several times, each time withdrawing the profit received. And, whereas, only once having lost the initial deposit already having several earned deposits of such amount in stock.

Further, we will consider in detail certain important moments for working with EDGE trading algorithm.

Currently, the best option is to optimize parameters in Open price only mode, so far as you considerably cut optimization time. After conducting optimization, having chosen the parameter combinations you are interested in and having saved them in set-files, you should make tests  in mode Real ticks (Every tick based on real ticks). As a rule, test outcomes in both modes are practically identical. If no, you should search reasons for serious incompliance in test visualization mode.

EDGE expert provides a feature to optimize parameters by user’s criteria. This is an improved version for selecting optimal parameters, which allows to find the most effective parameter combinations and to display them in upper lines of optimisation results. In order to use it, select Custom max in the drop-down list of optimization modes (see screenshot below).

The below screenshot shows an example of one of the possible variants for expert parameter optimization.

You can find a detailed description of parameters on EDGE application page in the market. Here, let’s consider which parameters should be chosen for optimization in strategy tester. 

The parameters not intended for optimization are designated by inactive check-boxes. Also, these parameters have no feature of specifying ranges: start, step, stop. Therefore, they must be set manually. The below screenshot shows an example of recommended parameter ranges for optimization. Once the parameter combination is selected, these ranges may be narrowed a little and we can look which results are received in the vicinity of those values. Pay attention that there are parameters which are used only for accounts of Hedge (only hedge) type. If you have an account of Netting type, these parameters should not be used during optimization and tests, because they exert no influence over the result.

Let us consider more closely some parameters which require explanation.

Using Symbols spread application you can easily and quickly look through key parameters of a variety of forex-symbols on your forex-broker’s server and to choose between the ones most fitting for your trade. This is specifically actual if a decision was made to apply one or another trading algorithm on short timeframes, where a big deal frequency is implied and trading levels should be set close to the current price. The fact is different brokers set their own values for such indices as spread and stop / limit levels. The shorter the timeframe is, the more these indices will influence the final result of trading. Symbols spread program might be useful for each trader who considers intraday trading for the purpose of gaining maximal profit.

The program will select symbols by user set criteria. Besides the time range filter, you can use filters for specifying the limit by spread amount (Spread filter) and by symbol name (Name filter). As well, the data type may specified: Bars (bars) or Real ticks (real ticks). If Bars data type is selected, then you might specify which timeframe should be used for viewing the spread in historical data. 

Let’s see which conditions are provided by broker RoboForex. Let’s request historical data for the last year, data type - Bars, timeframe M5 (five minutes). Once the criteria are specified, press Request button. After a while, the program will download from the broker server the data missing on your hard drive. Then it will generate series by the specified timeframe and conduct synchronization.

In the end, you see a result similar to the below shown on application chart (see screenshot below). The program has downloaded data by 38 symbols available for trading. It is obvious that many of them do not fit at all for trading due to huge spread.

As an example, here are test outcomes in Open price only mode. Timeframe M5 (five-minute data). Type of account Hedge. Time range is one year (2017.01.01 - 2018.01.01).

Symbol: EURUSD

EURUSD,M5: 281877 ticks, 74300 bars generated. Test passed in 0:00:01.453.
282883 total ticks for all symbols
EURUSD: passed to tester 282883 ticks
466 Mb memory used including 8 Mb of history data, 64 Mb of tick data

The outcomes of the test above show that for a one-year period the test on one symbol lasts for 1-1.5 seconds.

Now, try to make a test of the currency pair which has no account currency. For example, if you have an account in US Dollars (USD), for making the test take a symbol without USD. For example, EURCHF. The fact is, that for correct calculations of margin requirements and profit, in such case symbols EURUSD and USDCHF will be used in the test. In its turn, this will increase the test time.

Symbol: EURCHF

EURCHF,M5: 281063 ticks, 74273 bars generated. Test passed in 0:00:01.860.
846826 total ticks for all symbols
EURCHF: passed to tester 282468 ticks
EURUSD: passed to tester 282883 ticks
USDCHF: passed to tester 281475 ticks
467 Mb memory used including 8 Mb of history data, 64 Mb of tick data

As we can see, a test for cross-rates will be approximately twice longer. Here, the test took 1.5-2 seconds. Now, let’s try conducting a test on several symbols.

Symbols: EURUSD,GBPUSD,USDJPY

EURUSD,M5: 282881 ticks, 74300 bars generated. Environment synchronized in 0:00:00.031. Test passed in 0:00:07.172.
EURUSD,M5: total time from login to stop testing 0:00:07.203 (including 0:00:00.031 for history data synchronization)
853054 total ticks for all symbols
EURUSD: passed to tester 282883 ticks
GBPUSD: passed to tester 285067 ticks
USDJPY: passed to tester 285104 ticks
628 Mb memory used including 26 Mb of history data, 64 Mb of tick data

Symbols: EURCHF,AUDCAD,AUDNZD

EURCHF,M5: 282465 ticks, 74273 bars generated. Environment synchronized in 0:00:00.031. Test passed in 0:00:12.109.
EURCHF,M5: total time from login to stop testing 0:00:12.140 (including 0:00:00.031 for history data synchronization)
2264405 total ticks for all symbols
AUDCAD: passed to tester 284995 ticks
AUDNZD: passed to tester 285398 ticks
AUDUSD: passed to tester 282069 ticks
EURCHF: passed to tester 282468 ticks
EURUSD: passed to tester 282883 ticks
NZDUSD: passed to tester 282153 ticks
USDCAD: passed to tester 282964 ticks
USDCHF: passed to tester 281475 ticks
854 Mb memory used including 26 Mb of history data, 192 Mb of tick data

When testing several symbols the speed test falls. Unfortunately, now it is impossible to perform without losing in test accuracy. But, as we mentioned before, terminal developers in the nearest updates will extend features of MQL5, having added a feature of conducting multi-symbol tests several-fold faster.

As an example, try optimizing the parameters on data of Alpari broker on different symbols in Open price only mode. Timeframe M5 (five-minute data). Type of account Hedge. Time range is one year (2017.01.01 - 2018.01.01). Expert settings were demonstrated in the section above: Expert settings for parameter optimization

Symbol: EURUSD

result cache used 7953 times
genetic optimization finished on pass 15616 (of 504330836375520000)
optimization done in 28 minutes 56 seconds
local 7663 tasks (100%), remote 0 tasks (0%), cloud 0 tasks (0%)

Symbol: EURCHF

Symbols: EURUSD,GBPUSD,USDJPY

Symbols: EURCHF,AUDCAD,AUDNZD

Very soon terminal MetaTrader 5 will be updated and testing & optimization speed will be several-fold faster. Probably, then we can conduct optimization even in All ticks mode. Besides, using MQL5 Cloud Network service will become more beneficial because optimization speed will increase.

To compare, try conducting optimization with the same parameters as in MQL5 Cloud Network. The use of the service is paid. After each optimization, record the amount of funds frozen to pay for the service.

In this test, in order to speed up optimization, use simultaneously both local processor cores and cloud agents.

Symbols: EURCHF,AUDCAD,AUDNZD