Support Board
Date/Time: Mon, 07 Jul 2025 02:12:17 +0000
[Locked] - does sierra chart plan to improve bectests
View Count: 1692
| [2019-04-08 15:01:21] |
| User30743 - Posts: 365 |
|
when [if ever] will sierra chart engeneering start working on improving the becktester for acsil studies. because so far only ohlc barbased bt can be used, when using replay in higher speed [not much higher, lets say 5x times which is nothing..] , the results are completely wrong Date Time Of Last Edit: 2019-04-08 15:02:26
|
| [2019-04-08 19:58:11] |
| |
|
There is nothing to improve here. The solution is with your own trading system. More information about this is here: Auto Trade System Back Testing: Consistency Between Back Tests However, make sure you are also using Accurate Back Testing Mode: https://www.sierrachart.com/index.php?page=doc/ReplayChart.html#ReplayMode And if you use the option to Calculate at Every Tick when using tick by tick data, this is as accurate as it can be. So any further issue is going to be for you to solve on your side. Sierra Chart Support - Engineering Level Your definitive source for support. Other responses are from users. Try to keep your questions brief and to the point. Be aware of support policy: https://www.sierrachart.com/index.php?l=PostingInformation.php#GeneralInformation For the most reliable, advanced, and zero cost futures order routing, *change* to the Teton service: Sierra Chart Teton Futures Order Routing Date Time Of Last Edit: 2019-04-08 19:59:09
|
| [2019-04-12 12:43:43] |
| User30743 - Posts: 365 |
|
i mean it is really slow with Accurate Back Testing Mode. lets say i want to test a year or two, 4 markets simultanously because the system is based on intermarket situations. with 1x speed i get the same results as in realtime live mode - thats ok - but if i would like to test a year back, i would have to wait half a year for the results!!!! and if i make it faster (lets say 10x or even 5x) , it returns completely wrong entries on places where they should not be and places that should be taken it doesnt make an order. so it is incorrect. and this is not by far only my problem. as far as i talked with others who develop under siera, they have the very same problem. you should really consider what i am saying. if it was only me, then ok, it may be on my side, but since it has pretty much everyone, it is not on my side. what works fine is OHLC bar based beckest - yeah, there the results are correct and consistent. but when u need more precise test, on replay or tick by tick data, is really absolutelly incorrect. u should try it - you will see even with your own code Date Time Of Last Edit: 2019-04-12 12:46:41
|
| [2019-04-12 17:10:45] |
| |
|
Accurate back testing can be slow but refer to: Auto Trade System Back Testing: Improving Back Test Performance if i make it faster (lets say 10x or even 5x) , it returns completely wrong entries on places where they should not be and places that should be taken it doesnt make an order. so it is incorrect. What you are describing is a problem with your system. When you say it returns completely wrong entries this is because that's what your trading system is doing! It's your system! You are only talking about what your trading system is doing! All of this is for you to resolve. We have no ability to resolve what your system is doing. This is point-blank obvious by what you are saying.One thing we discovered with one user was they were relying on the current date-time variable which can be inaccurate and that was affecting how their trading system was working, and this is explained here: ACSIL Interface Members - Functions: sc.GetCurrentDateTime() This is the only Date-Time which could vary in the ACSIL interface during a back test based on the replace speed. Otherwise, everything is always 100% consistent. Maybe you are doing the same thing. and this is not by far only my problem. as far as i talked with others who develop under siera, they have the very same problem. you should really consider what i am saying. if it was only me, then ok, it may be on my side, but since it has pretty much everyone, it is not on my side. There simply is no problem for us to resolve. There could not be. This is a complete impossibility. Sierra Chart Support - Engineering Level Your definitive source for support. Other responses are from users. Try to keep your questions brief and to the point. Be aware of support policy: https://www.sierrachart.com/index.php?l=PostingInformation.php#GeneralInformation For the most reliable, advanced, and zero cost futures order routing, *change* to the Teton service: Sierra Chart Teton Futures Order Routing Date Time Of Last Edit: 2019-04-12 19:18:48
|
| [2019-04-12 17:57:17] |
| User30743 - Posts: 365 |
|
this is really sad that you believe that the problem is always only on our side. simple logic: the systems behave properly when running realtime or with 1x replay speed on beckests. when running faster backtest, sierra is simply unable to calculate and puts wrong entries. what other technical explanation do u need? i will provide you whatever necessary Date Time Of Last Edit: 2019-04-12 17:57:55
|
| [2019-04-12 18:35:39] |
| |
|
when running faster backtest, sierra is simply unable to calculate and puts wrong entries. You need to find out why your system is doing that. We gave you one possibility and that is that you are relying on a Current Date-Time variable which will be inaccurate during an accelerated replay and we gave you the relevant documentation.It is your system doing the calculations and is your system doing the wrong entries. Refer to: Automated Trading Management: Troubleshooting Automated Trading System Behavior You need to take responsibility for what you are doing and solve your own problem. Sierra Chart Support - Engineering Level Your definitive source for support. Other responses are from users. Try to keep your questions brief and to the point. Be aware of support policy: https://www.sierrachart.com/index.php?l=PostingInformation.php#GeneralInformation For the most reliable, advanced, and zero cost futures order routing, *change* to the Teton service: Sierra Chart Teton Futures Order Routing Date Time Of Last Edit: 2019-04-12 18:36:12
|
| [2019-04-12 19:21:21] |
| |
|
Are you running a multiple chart back test? In older versions, when using the Accurate Trading System Back Test Mode, the behavior is different in that case and is fully documented here: https://www.sierrachart.com/index.php?page=doc/ReplayChart.html#ReplayMode This is something to be aware of. That can cause a variance in the behavior of a trading system. This has been changed in recent versions where the behavior of Accurate Trading System Back Test Mode for a single chart replay also applies to multiple chart replays. Sierra Chart Support - Engineering Level Your definitive source for support. Other responses are from users. Try to keep your questions brief and to the point. Be aware of support policy: https://www.sierrachart.com/index.php?l=PostingInformation.php#GeneralInformation For the most reliable, advanced, and zero cost futures order routing, *change* to the Teton service: Sierra Chart Teton Futures Order Routing Date Time Of Last Edit: 2019-04-12 19:23:56
|
| [2019-04-13 08:03:08] |
| User30743 - Posts: 365 |
|
yes i am running multiple chart backtests on version 1800 or 1830
|
| [2019-04-13 08:49:08] |
| Ackin - Posts: 1865 |
|
User30743) I stand on Sierrachart's side that it works fine (BT / replay does exactly what's written in the documentation), if you want other else: https://www.sierrachart.com/index.php?page=doc/SierraChartStudyAndSystemProgrammers.php You write: i mean it is really slow with Accurate Back Testing Mode. lets say i want to test a year or two, 4 markets simultanously because the system is based on intermarket situations. You want suggest Sierra to improve their program because you think it's slow to calculate your ATS. I understand that?Programming does not mean knowing commands, but thinking as programmer. Understand that the back test is at the end of YOUR system. You can program in c++ which follows from your posts. It is really such a problem to create this as a two phase program? which saves the situation to an external file from each market in one step, and then loads all of it in the next step into one chart and tests it without the load-calculation of each study. You have one chart, one study. Is it really so hard ? You have to think how to help yourself if the program doesn´t offer you what you want. Sierra Chart Engineering) I'm (and others) waiting for a deeper inner measure that can do only Sierra: to support OpenGL, support multi-core or other system acceleration, develop such great things as 200 levels of depth or improve tradeslog (thanks for this) .... I think that's really important for all ...and not to solve unfinished work of individual devolepers. Date Time Of Last Edit: 2019-04-13 11:37:07
|
| [2019-04-13 10:25:03] |
| |
|
yes i am running multiple chart backtests on version 1800 or 1830 Therefore, it does relate to this which is in the documentation:When you have enabled Charts to Replay >> All Charts in Chartbook | Charts with Same Link Number and you are using Accurate Trading System Back Test Mode, then in this case there is a multiple chart synchronized replay being performed.
Therefore, in the case of a multiple chart synchronized replay, prior to version 1863, the studies are not called/calculated whenever the High, Low, or Last prices of the current bar changes or there is a new bar added to the chart during the processing of the data records in the Intraday chart data file. Instead the studies are called/calculated after a specific amount of data based upon time, the default is 60 seconds, has been read from the Intraday data file into the chart. At that time the study functions are calculated on the chart. For more information, refer to the Auto Trade System Back Testing page. This amount of time is prompted for when the replay is started during a multiple chart synchronized replay with the Enter Processing Step in Seconds prompt. So update to the current version and the issue will be resolved for you. Or you should change the Enter Processing Step in Seconds to a small value like 1 second. Sierra Chart Support - Engineering Level Your definitive source for support. Other responses are from users. Try to keep your questions brief and to the point. Be aware of support policy: https://www.sierrachart.com/index.php?l=PostingInformation.php#GeneralInformation For the most reliable, advanced, and zero cost futures order routing, *change* to the Teton service: Sierra Chart Teton Futures Order Routing Date Time Of Last Edit: 2019-04-13 10:26:10
|
| [2019-04-13 12:53:02] |
| User30743 - Posts: 365 |
|
Akcin, there are many other people who encounter the same problem with SC backtest, one of them is a pro developer who was troubled so much with that, that at the of the day was forced to built his own becktester that improved the speed more than 100x times with absolutelly correct results even for fastest timeframes and tickbytick data. so yes, there is a space for improvemet. this is just a fact. dont get me wrong - i like sc a lot. it is the best platform i have ever used. but the backtests are weak. Date Time Of Last Edit: 2019-04-13 12:55:31
|
| [2019-04-13 14:45:56] |
| Ackin - Posts: 1865 |
|
Akcin, there are many other people who encounter the same problem with SC backtest
I know ... but the debate is academic when they do not use Sierra studies but their own (often users studies = high memory and processor load due to bad code). It isn't an objective assessment.so yes, there is a space for improvemet. this is just a fact.
no.. what you want is a completely different tool, you want a tickbytick tester for long-term testsone of them is a pro developer who was troubled so much with that, that at the of the day was forced to built his own becktester that improved the speed more than 100x times with absolutelly correct results even for fastest timeframes and tickbytick data.
I also know one - me :) I have own tester which test data (one year) on a 1trade chart in 15-25 minutes (YM,NQ,CL) verified by a group of 20 traders - also in live, my ATS contains 30 IDs studies, but it's a very specific tool and doesn't work on big timeframes (I didn't need it for big TF, it could be reprogrammed). And I'm not alone, I know about more developers, what they were crying for orderflow tests with Sierra backtester, but we made ours and we are satisfied. It doesn't change on this that I wrote "(BT / replay does exactly what's written in the documentation)" for what you want it was not created dont get me wrong - i like sc a lot. it is the best platform i have ever used.
ok, we agree on this ;)
Date Time Of Last Edit: 2019-04-13 21:26:06
|
| [2019-04-13 15:07:02] |
| User30743 - Posts: 365 |
|
I dont need orderflow tests. I need test on multiple charts. Properly sychronised OHLC would be enough but OHLC tester doesnt work for multiple charts. So i need to use replay - but Accurate Replay doesnt skip empty periods. HA. What an insanity. I want to test only two hours every day - the rest 22 hours i want to skip within a second. But there is NO WAY to skip. - Instead, SC tells me I should use "higher speed" replay. So I use higher speed and even with that "higher speed" I need to wait couple of weeks for a result. So it is completelly useless - Dont tell me it is not! Who can wait weeks for just one simple backtest???!! What is the advantage that it is all written in C++ when u need to wait weeks for one simple test?? (which will have wrong results at the end anyway because it was "too fast" replay)?? Explain me this and I will be satisfied! ------------- and they say: "There is nothing to improve here in backtest. The solution is with your own trading system." --->> This is absolutely not. Making SC to skip empty periods or multiple-chart synchronised OHLC bar based tests is NOT SOMETHING I CAN CONTROL FROM WITHIN MY STUDY. Date Time Of Last Edit: 2019-04-13 15:23:27
|
| [2019-04-13 15:33:48] |
| Ackin - Posts: 1865 |
|
OHLC tester doesnt work for multiple charts.
that would be useful, I agreeExplain me this and I will be satisfied!
this is a question on Sierra not for memy advice in your situation (I'm not solving the tester now): try to think if you can't recode your program into one chart, you don't need it online in the tests so you can do in more steps as I wrote above. This is meant in good. Even if Sierra processes your request, it's not a matter of the week. You can still work without delay, if you want to advise something, write me a direct message. |
| [2019-04-16 07:56:53] |
| User553714 - Posts: 184 |
|
I was following this post to see the outcome; Wasn't there a post to this thread with text file results, where did this post go?
|
| [2019-04-16 08:28:11] |
| |
|
It was hidden because the post has incorrect information. There is a 100% consistent outcome but the problem is they are not using the correct back test settings. They must be running a multi-chart back test, for no reason for that trading system, on an older version and probably using a large processing step in seconds. In that case, yes there will be inconsistent results but that is a complete disregard to the information we have posted already on this. We are not sure what they are trying to prove here. This thread is locked and we have no tolerance, for the disregard for what we are saying. At this point we are going to start to delete posts if needed. And we still want to verify what the user is saying and we are going to hide it until then. We cannot have misleading information here. There is also mention of skipping periods during a multi-chart back test. This inherently is very complicated while also still maintaining synchronization among charts. Sierra Chart Support - Engineering Level Your definitive source for support. Other responses are from users. Try to keep your questions brief and to the point. Be aware of support policy: https://www.sierrachart.com/index.php?l=PostingInformation.php#GeneralInformation For the most reliable, advanced, and zero cost futures order routing, *change* to the Teton service: Sierra Chart Teton Futures Order Routing Date Time Of Last Edit: 2019-04-16 08:33:25
|
To post a message in this thread, you need to log in with your Sierra Chart account: