Lifecycle of algorithmic trading portfolio
Running trading strategies on a live Forex account is a very important part of real algorithmic trading, that benefits not only those who use platforms for their trading like StrategyQuant X, but also every algorithmic trader. This article will include techniques that on applying will give a wider view on how trading robots are behaving in a real environment and at least but not last, improve trading results.
Algorithmic trading is characterized over other trading approaches by giving the possibility of having a detailed view of what is happening with the running trading robots. very trade and order can be managed on an unlimited number of trading instruments simultaneously. The key now falls in the way of getting the most out of this advantage. in other words, the cornerstone is not just collecting data, but how to evaluate collected data and obtain real outputs for real trading. So, let’s start;
On coding a Forex trading robot either using a strategy builder for that or not, its algotrading lifecycle passes through five milestones likin the graph above. All start with getting a portfolio of trading robots. It's not recommended to use just one robot for that as diversification is so important for the success of algorithmic trading and it can b achieved by using multiple robots. Portfolio composition can b discussed in a separate article.
Next, you will need to prove the strategies in the real market even if it was created in the best possible way, some issues might still exist and won't be discovered till the robot starts to deal in the real money world. The testing period will let you discover these bugs and give you a better idea of what is expected from the engaged trading strategy. By gathering enough real data, you can then decide what should be changed in the portfolio, in a process that will be discussed later in this article.
Now the valuation has been completed and the strategy can b applied live. Live trading valuation should be continued, and we will discuss it in detail. Basically, the two moments you should compare live trading results with the backtest are at the end of the testing period and three months after deploying the strategies live. A periodical check that can b regularly once a year is very important too.
The first step in this process that follows starting live trading on your trading platform is registering your trading account in one of the online monitoring platforms like FXBlue or MyFxBook, the first offers easy access to all necessary statistics, so it will b used in this presentation. It's shown in the graph below disclosing how to export your live account trading history into a CSV file, this saved report will b used in the further steps of this tutorial.
The FxBlue.com free online monitoring platform
Export of live trading results into a CSV text file
With Quant Analyzer software “Compare results” module, a bulk comparison of live strategy results and backtest obtained from multiple sources can b performed to introduce several scenarios.
It allows comparing multiple results between trading platform / live results and backtests in StrategyQuant X / AlgoWizard.
This can be done manually by loading each report into Quant Analyzer and then combining it in the portfolio, but this new module can do it automatically for all files at once.
There is a new icon Compare results on the top left
“Compare results” Quant Analyzer module
With this option, it will recognize all Magic numbers from the reports loaded from Folder 1, and it will try to find strategies in Folder 2 that have this Magic Number in their name.
So for example, if Magic Numbers are 12345, 76543 then it will look for files with strings “12345” and “76543” in their names in Folder 2. If found, it will combine these found results into one portfolio.
When you click on the "Compare" button it will perform the comparison and writes the log. All the comparison results are stored in the Portfolios databank.
Performing results comparison
Result comparison graph
In this example, the backtest of strategy in TradeStation (first blue line) was compared with the backtest in StrategyQuant X (second, red line). Some differences in trading can be seen but the equities are quite close to each other.
Many brokers use their own symbol syntax that might not be defined in the Quant Analyzer setting file. In such a case, Quant Analyzer want be able to recognize them when you load a report of your account into it. Here is how to define all symbols at once in the Quant Analyzer setting file:
Matching symbol’s definition in Quant Analyzer
Results presentation should b switched from money to pips for correct comparison of results.
Switch presentation of results from "money" to "pips"
This is the most common scenario for the traders who use StrategyQuant X for creating strategies:
Compare fxblue live data with StrategyQuant X
Comparison graph: Greenline is the backtest and the red line is the result from FxBlue. You can see that backtest is quite short compared to the live results. Regarding Live/BT accuracy results matches well
If you have a trading robot in MT4/MT5 code and you can perform backtest in MetaTrader, this scenario can be employed. Noting that most brokers do not provide long enough historical data for the trading symbols, but the free QuantDataManager can give you access to a longer history. Continuing with the comparison, steps are similar to the previous scenario:
Comparing MT4/5 and FxBlue results
On the image, you can see that results fit almost perfectly
TradeStation live report can also be loaded and compared with strategies generated in StrategyQuant X. MT4 backtest results can also b compared with StrategyQuant X results. Numerous ways how to use this module exist.
Real results will be usually slightly worse than the backtest as real trading always involves slippages, spikes, etc. So the percentage of similarity between the backtest trades and the real ones greatly matters! Significant differences between them indicate that we should stop trading this strategy, as there is some discrepancy in the execution of the logic of the strategy, while if the matching for example reaches 80%, it is then confirming the quality of the strategy.
Live results deviation from the backtest can be seen in the graph below. A red rectangle mark on a situation during the testing period suggests stopping to trade the strategy.
Real-backtest results mismatch
Besides matching the results, one more important aspect is needed to be evaluated during live trading especially while evaluating the testing period. It is comparing a live performance to backtest where the overall profit of a trading symbol e.g. EURUS, can be taken and compared with the backtest to find out the percentage of real performance to the backtest that could be for example 70%. Generally, 70% up to 100% is fine as live conditions are always tougher than the ideal backtest. But if that percentage falls to 30%, then further research is needed to find a reason for this inconsistency, that could be for example lack of liquidity on the symbol which leads to big slippages.
Several years ago, there was a lack of methodology and tools to evaluate live results. Many custom solutions exist now for that purpose for everyone with a wide range of supported platforms like Quant Analyzer. This article focused on practical examples as well as on rules on how to evaluate results during live trading. Having good rules will help you to make clear decisions during your lifetime of portfolio and will allow you to be more consistent with your results.