Quel est votre processus de backtesting et d’optimisation des systèmes de trading techniques ?

Back testing and optimization should be done on the same language where the system originally or going to be developed. Back testing as a POC can be done in Python which give flexibility. I would go with C++ or Golang for full development.

Use what you are comfortable with. Python is used to analyse data, backtest, and find an edge. Use C++ or C# for faster execution

Because the most part of the strategy creation will be based on datasets and backtesting python should be the go to language for testing, c# and c++ for execution.

Clean data, programming language preferably python because of available libraries that can be put to good use. Optimisation should be avoided to suit our need to magnify profitability ratios. A good strategy would give good results without optimisation

Data Collection: Gather accurate historical price and volume data. Strategy Development: Define entry/exit rules, risk management & parameters. Backtesting: Apply the strategy to past data to evaluate performance. Performance Metrics: Calculate P&L, Sharpe ratio, and drawdown to gauge effectiveness & risk. Optimization: Adjust parameters to improve strategy outcomes. Out-of-Sample Testing: Ensure strategy works on new data to prevent overfitting. Risk Management: Implement safeguards like stop-loss orders and diversification. Implementation: Begin live trading with a small capital allocation. Continuous Monitoring: Adapt strategy to changing market conditions. Record Keeping: Maintain thorough trade records for ongoing assessment.

There is no better way to backtest than a manual and visual backtest. For example, it is great to see the results from a system saying buy/sell when an EMA is crossed or similar indicator. But a visual analysis provides more advanced studies and helps in avoiding whipsaws. We shouldn't forget system is designed by people, and the system will always be dumber than human intellect. Best traders are always the ones who do things manually and not the ones who rely on some coded trades. Mediocrity yes very much possible with the system backtesting and all the hoax around it.

Metatrader would be the most complex and easy to adjust platform but in reality the best backtests are executed with own systems(python again)

TradeStation again as the charting software is top notch as are the backtesting capabilities. They also have great features such as “paint bar” and “show me” which enable traders to easily visualise their trading set ups.

Ensure you are selecting a backtesting platform that replicates your "real world" trading style. For example, you may use Trailing Stops; select a quality platform that takes your strategies into account. As well, many of the free (or budget) backtesting platforms severely restrict features, historical data, markets or securities.

There are many trading platforms that offer backtesting capabilities. Some of the most popular ones are: AI-powered Trade Ideas TrendSpider for automated candle and indicator testing. TradingView for international traders Trade Ideas is the best AI stock backtesting and integrated auto-trading software. TradingView offers great free backtesting software for stocks and auto-trading. TrendSpider has powerful pattern recognition, backtesting, and auto-trading.

EasyLanguage gives one the ability to do fun things like partial exits, for example placing limit orders at different price levels in fractions of the entry size.

Yes AND not all strategies are directional , some strategies require arbitraging, hedging and non TA related aspects. As a result, some strategies would require inventory size instead of stop loss, aggressiveness, ob tracking and kill switches.

Popular trading platforms for backtesting strategies include MetaTrader, Amibroker, TradingView, NinjaTrader, MultiCharts, and QuantConnect. Consider your requirements and programming skills to choose the platform that best suits your needs. These platforms vary in terms of features, programming languages, and costs. It's recommended to explore their functionalities, consider your programming skills, and choose a platform that aligns with your specific needs and trading style.

You need to consider better data structure to be used across the system. Instead of chosing a single data structure we should use different structure basis the need of the logic. Most of the places it would be linear look up in such case Array would give faster access. In some cases where we need to store data basis Key and Value like Symbol and Ticks, here we should use some collection like hash or dictionary

When analyzing backtest performance, it's important to note that not all performance metrics suit every strategy. For example, the Sharpe ratio might not be ideal for trend-following because of the way it treats volatility. It measures risk using total volatility (standard deviation) of portfolio returns, without distinguishing between downside volatility (negative) and upside volatility (positive, like large upward profit spikes). For strategies like trend-following that have asymmetric returns, where positive volatility is a feature, the Sortino ratio is a better metric as it focuses on downside volatility. That said, for symmetric return profiles like index-tracking or mean-reversion, the Sharpe ratio is a reasonable go-to metric.

The processing of back test day should be in time series to get correct signal. For example while comparing multiple symbol at different time frame is important. So while reading the data should be time wise sequencial.

There are many ways to optimise . Exhaustive or genetic or more . This is also an expanding field with more work on AI in general which essentially is a type of machine learning . You’ll need to select your data inputs and data frequencies too . After coding something you’ll also have to select you data samples and your in and out-sample data sets and compare them . If something seems to work based on your research then perhaps you’ll need to find some logic to construct a portfolio of your “ stuff that works “ .

There are 2 way to optimize the trading system, first is to adjust the parameter so the signal should be close to correct. The internal and external parameter to be added / removed to get this. The second is to adjust system performance ie the application spead etc. This will ensure the same result during production also

Non linear trading strategies, inventory based strategies and strategy market allocation as strategies should be separated by market type.