AMQParticipant
Senior
Hi everyone,
I’ve seen the tremendous work @Nicolas has put into Prorealquant and the update released today.
I already use Prorealquant, and I’m planning to try out ProrealAI soon.
I’ve been wondering lately if it might be possible to run backtests using AI. In Spain, I don’t have the option to access 1-minute (1M) candle data through the PRT/IG integration. After discussing this with @Turame, I learned that IG only allows downloading 5-minute (M5) data going back a maximum of one year.
The goal is to speed up strategy development and ultimately validate robustness within PRT.
Given the data update in Prorealquant, I was wondering if it would be feasible to download eight years of M5 data for the Nasdaq 1€ instrument. This would allow for the reconstruction of H1 and D1 candles within PRT, enabling AI-based backtesting of early-stage strategies.
Many thanks.
Alfonso
Through the IG API, it’s indeed impossible to retrieve more than one year of data. The data series available in ProRealQuant don’t come from IG, the main reason being this limitation 🙁
I’ve noticed a few minor issues with the application; I’ll fix them tomorrow. I’ll also add the M5 for the Nasdaq as you requested; the 8 years of backtesting might be too intensive, we’ll have to check!
For your information, I am working on this subject.
The biggest problem has always been the available memory size in a web browser to allow simulations with very large data.
Indeed, for your request concerning the NQ100 in 5 minutes with 8 years of history (minimum), we are approaching a million candlesticks.
I hope to be able to complete an optimization of the backtesting engine that will enable this in the next few days; I’ll keep you posted! 😉
AMQParticipant
Senior
Great, Nicolas! Thanks a lot.
The thing is, right now in Spain, 1 million candles aren’t available for backtesting on PRT with IG. So, I test on a higher timeframe and look at the last 200k candles on the M5 chart as an approximation, but I’m not really convinced that this method ensures system robustness.
If there’s anything I can help with, let me know.
Thanks again.
ProRealQuant is updated with a new version of the engine that allows more datas to be backtested at once than ever before.
It means that 1M candlesticks data is held “easily” on any computer. Try it, I added the data you requested for (NQ100 5-minutes timeframe, back to 2012).