Support Board
Date/Time: Sun, 12 May 2024 18:21:42 +0000
Post From: Backtesting is inaccurate or slow
[2020-07-28 18:46:26] |
Marion2025 - Posts: 13 |
I need some advice improving my backtest compute bottlenbecks. I've read through the documentation but still have a general problem: 1) Bar based backtesting is very inaccurate for my trading strategy. When compared to Replay based backtesting, the results are completely different. 2) For this reason I prefer the more accurate replay based backtesting, but the compute performance is very slow making it frustrating and time consuming I don't understand why the backtest speed is so slow. I have a 48 thread Xeon with 128gb of ram and a RTX 2080ti GPU but SC only uses a fraction of computing resources during a backtest Are there specific causes to why the two backtest methods would cause such different results? Is there a way to modify my strategy to generate consistent results between the two methods? It would be great if I could work with bar based backtesting since it is much faster, but it is useless if inaccurate. Alternatively, can I improve the replay backtesting speed without introducing inaccuracies? Any advice beyond what is in the documentation is much appreciated |