Backtest Depth - Long-term Robustness (10 years) vs. Recent Adaptation (5 years)? - page 2
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
the current market doesn't behave the same way it did a decade ago
How is that the current market doesn't behave the same way it did a decade ago ???
Price will either go UP DOWN or SIDEWAYS. This is the SAME behavior from begining of time
Halo komunitas. Saya sedang mengevaluasi keandalan EA saya dan muncul pertanyaan mengenai kedalaman historis optimal untuk backtesting .
Dilema saya adalah ini : pengujian historis 10 tahun secara teoritis membuktikan bahwa EA (Expert Advisor) dapat bertahan dalam berbagai kondisi pasar. Namun, pasar saat ini tidak berperilaku sama seperti satu dekade lalu (volatilitas berbeda, algoritma institusional, dll.). Apakah Anda lebih memilih untuk mengoptimalkan kinerja selama 3-5 tahun terakhir untuk beradaptasi dengan kondisi pasar saat ini, atau Anda membutuhkan sistem Anda untuk menghasilkan keuntungan selama periode 10 tahun atau lebih?
Saya akan sangat menghargai jika Anda berbagi pengalaman Anda.
How many RAM needed to load 10y of tick data in the tester ?
How is that the current market doesn't behave the same way it did a decade ago ???
Price will either go UP DOWN or SIDEWAYS. This is the SAME behavior from begining of time
for 1 min OHLC data basically nothing, for tick data depending on the symbol it can be a lot probably like 3 gigs of ram per agent for 1 symbol over 10 years for me
How long does it take your computer to complete a full test with real tick data? On mine, it takes forever, despite an excellent setup.
In monothread used for a single pass test, here's the actual CPU available on the market making better than mine :
- Top 1 Intel Core Ultra 7 270K Plus / Ultra 9 285K ~25% à 30%
- Top 2 Intel Core i9-14900K / KS ~22% à 25%
- Top 3 AMD Ryzen 9 9950X / Ryzen 7 9700X ~20%
- Top 4 AMD Ryzen 7 9850X3D / 9800X3D ~15% à 20%
- Top 5 Intel Core i7-14700K ~15%
- Reference 0% (Base)
As you can see I couldn't expect more than 25% gain. As a reference, 25% of 10mn, being 2min30 is negligible.So if you have a secret, I'd be please to hear about.
How long does it take your computer to complete a full test with real tick data? On mine, it takes forever, despite an excellent setup.
In monothread used for a single pass test, here's the actual CPU available on the market making better than mine :
- Top 1 Intel Core Ultra 7 270K Plus / Ultra 9 285K ~25% à 30%
- Top 2 Intel Core i9-14900K / KS ~22% à 25%
- Top 3 AMD Ryzen 9 9950X / Ryzen 7 9700X ~20%
- Top 4 AMD Ryzen 7 9850X3D / 9800X3D ~15% à 20%
- Top 5 Intel Core i7-14700K ~15%
- Reference 0% (Base)
As you can see I couldn't expect more than 25% gain. As a reference, 25% of 10mn, being 2min30 is negligible.So if you have a secret, I'd be please to hear about.
Optimize your code. My tick based algo, depending on symbol takes on average 20 seconds to complete a 5 year single pass on a old and slow i5-7400. And it is still slow compared to if i create a math logged custom symbol instead of math logging each and every tick twice (bid/ask) at run time.
Use the virtual library or even better use Math mode. You can achive gains of an order of magnitude on the same hardware.
Optimize your code. My tick based algo, depending on symbol takes on average 20 seconds to complete a 5 year single pass on a old and slow i5-7400. And it is still slow compared to if i create a math logged custom symbol instead of math logging each and every tick twice (bid/ask) at run time.
Use the virtual library or even better use Math mode. You can achive gains of an order of magnitude on the same hardware.