Support Board
Date/Time: Sat, 23 Nov 2024 23:24:58 +0000
Post From: Why is the fastest backtesting so relatively slow?
| [2024-07-19 14:31:59] |
| User386194 - Posts: 3 |
|
Sierra Chart is by far the best platform available today, with terrific functionality, engineering quality and value-for-the-money. There is however one area in which I think Sierra could be even better - backtesting speed. The fastest way currently to backtest in Sierra Chart is using AutoTrade System Bar Based BackTest. On my PC, backtesting (with 250 Bar Processing Increment) a strategy on a 1 Minute Chart with the Date Range 2023-01-01 to 2024-07-19 takes 10 seconds. Calculating a study on the same chart takes 102 ms. Shouldn't a bar based backtest execute in a time more similar to a study? A study loops through all the bars and applies some sort of logic, too. A somewhat reasonable scenario would be setting up an optimization with ASCIL for 100 symbols, with 500 iterations per symbol. With the same speed as in the example above - using bar based backtest - that would take: 100 * 500 * 10 = 5.78 days If a backtest instead would execute in 102ms, the same calculation would take: 100 * 500 * 0.102 = 1.42 hours. Is it possible that a backtesting function of this speed could be implemented in Sierra Chart? This would be a significant improvement for Sierra Chart and offer new opportunities for analysis and strategy development. And also be more in line with what one would expect from a C++ based platform with otherwise superb quality that Sierra Chart has. One of the main features of Sierra is: Sierra Chart is extremely fast with a definite focus on high performance in all areas of the program.
I find the backtesting speed to be the one thing that does not live up to that description. Would it be possible for the engineering team to look into this? Best regards, J |