 Posted 2/3/2009 12:46:20
|
|
|
|
Thank you
|
|
|
Posted 2/3/2009 19:36:19
|
|
|
|
| Edition 2 Beta 6 feedback after backtesting an intraday trading system on 1 minute bars of the S&P 500 component stocks: The memory saving works well, which is very welcome. Backtesting is drastically slower than in Edition 1, which I hope will be fixed. Speed is not affected by an assignment to Bars.MaxLookBack.
|
|
|
Posted 2/4/2009 12:17:43
|
|
|
|
| Is the backest slowness a known issue, and is a resolution likely in the final release?
|
|
|
Posted 2/9/2009 12:22:57
|
|
|
|
| We will profile and optimize the code before the final release. It may be that because of the new frequency system there is simply more work to do, and we won't be able to make it significantly faster. If this is the case we may enable a mode to bypass a lot of the frequency processing if the frequency of the bar data you are using matches your system frequency. Thanks,
Daniel
|
|
|
Posted 2/9/2009 20:03:35
|
|
|
|
On the speed question.
I have written a MetaStock data plugin. This removes the need to run SQL server or any other database system that just adds overhead. I found any database always lags a flat file system.
The speed gained was very noticeable when running a back test over 200 stocks for up to 10yrs.
For the RE guys: Have a look at Amibroker for real speed.
|
|
|
Posted 4/15/2009 12:09:40
|
|
|
|
| To the RE developers: I've been waiting for the Edition 2 backtest slowness issue to be fixed before working with RE again. Has this been resolved in the current beta? If not, what is the expected time frame for resolution? Thanks.
|
|
|
Posted 4/15/2009 13:33:58
|
|
|
|
Optimization of code is really the last step, so no, Ed 2 is still slower backtesting than Ed 1. It will likely improve before release.
Steve2008 (4/15/2009)
To the RE developers: I've been waiting forthe Edition 2 backtest slowness issue to be fixed before working with RE again. Has this been resolved in the current beta? If not, what is theexpected time frame for resolution?
Thanks.
|
|
|
Posted 4/15/2009 15:27:11
|
|
|
|
| Is it likely that backtests are faster as in Edition 1 after your optimization or will it still rather be slower?
|
|
|
Posted 4/15/2009 15:30:37
|
|
|
|
My guess is that they will be on par with Ed 1. But we haven't even run through the optimizer yet.
Tradix (4/15/2009)
Is it likely that backtests are faster as in Edition 1 after your optimization or will it still rather be slower?
|
|
|
Posted 4/15/2009 15:46:51
|
|
|
|
| Will the memory leak also be fixed in edition 2? As you probably already know Edition 1 has a leak at optimizations. At least Daniel was able to reproduce it. I discussed it with him 1 month ago and attached a empty project. To reproduce simply do an optimization with > 10 markets and 10 years of EOD data or similar data. The attached simple project is without any code and results in out of memory when doing an optimization. It's the same with every project, so i reduced all my own code parts until i came to a empty project which shows that it must be something directly in RE which doesn't free up all the used memory after each optimization run. It would be great to have this bug fixed, so we could make optimizations. Right now out of memory happens after about 50 iterations which isn't a lot.
|
|
|
|