Support Board
Date/Time: Tue, 06 May 2025 20:07:16 +0000
SC Version 2716: (GOOD) Best backtest times ever(BAD) No T&S, TAL Window Updates
View Count: 769
| [2024-12-04 06:55:21] |
| User569373 - Posts: 27 |
|
This study takes 3 minutes 20 seconds in SC 2700 / 2712. The fastest ever seen duration was 3 minutes 9 seconds. The duration fluctuated between 3 minutes 9 seconds and 3 minutes 40 seconds. Most often near 3 minutes 20 seconds. Til SC 2700 I used a Chart Update Interval of 250 ms. In 2712/2714 I played around with different CUI. In 2714 I got best experience with CUI 550ms but was far away from 3 minutes 20 seconds. In 2712 I use a CUI of 100 ms and get a duration of 3 minutes 20 seconds. Backtesting speed is: 100000 The GOOD ======================================================================== Since SC 2716: Backtesting times: Start End Duration Chart Update Interval 06:19:23 06:23:02 3:39 250 ms 06:25:50 06:30:04 4:14 100 ms 06:31:15 06:34:56 3:41 500 ms 06:52:37 06:55:33 2:56 550 ms 06:57:04 06:59:59 2:55 550 ms 07:04:28 07:07:21 2:53 550 ms 07:08:56 07:11:50 2:54 550 ms 07:13:25 07:16:17 2:52 550 ms 07:17:16 07:20:10 2:54 550 ms This is the first version giving me such fast backtesting times. Around 2 minutes 53 seconds. This is the first version showing constantly almost the same backtesting times. 4 seconds difference from fastest to slowest backtesting over multiple runs. The BAD ======= Like in version 2711 there is no output in the T&S, TAL and Trade Orders Window while backtesting. SC also doesn’t accept any input till the end of the backtest. At the and of the backtest the TAL Statistics and TAL Trades get populated. Its just the same like in 2711 which got solved really fast in 2712. Due to the missing interaction capabilities in 2716 I will stay with 2712 although the backtesting speed and more stable duration times in 2716 is GREAT. Kind Regards Date Time Of Last Edit: 2024-12-04 12:23:42
|
| [2024-12-04 07:12:45] |
| Sierra_Chart Engineering - Posts: 19460 |
|
There are no changes, for a year or more or years regarding updating of Time and Sales or the Trade Activity Log or the Trade Orders Window during back testing. None at all. Nothing has changed with this. There are no differences between versions 2700 and earlier or 2716. They all function identically in this regard. Absolutely no changes at all. 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, use the Teton service: Sierra Chart Teton Futures Order Routing Date Time Of Last Edit: 2024-12-04 07:13:34
|
| [2024-12-04 07:23:32] |
| User569373 - Posts: 27 |
|
I don’t argue with you. READ my thread “SC Version 2711: Unresponsive SC on ongoing Backtest“ Even “Sierra_Chart Engineering [Direct Messages] - Posts: 17351” confirmed it. “Anyway regarding the issue you have, we understand why it is happening and that will be resolved in 2712.” And it got resolved in 2712. So stop wasting the customers time and ask SC Engineers who know the software !!!! Date Time Of Last Edit: 2024-12-04 07:46:41
|
| [2024-12-04 10:50:18] |
| User569373 - Posts: 27 |
|
After I restored version 2700 from a backup and updated to 2716 I can no longer get the former good duration for backtesting. What remains is just the not accessible SC while backtesting and the TAL Statistics, TAL Trades and T&S Window not getting updated. Kind Regards Date Time Of Last Edit: 2024-12-04 12:24:12
|
| [2024-12-04 12:20:35] |
| User569373 - Posts: 27 |
|
What is going on with version 2716? Does it have some kind of AI in it? After my post this morning I was totally disappointed I couldn’t reproduce the wonderful backtesting duration I got this morning. So I increased / decreased the Chart Update Interval. Although I went from CUI 550 ms to 700 ms back to 550 ms. It doesn’t matter. At some point I even stopped altering the CUI and the backtest duration kept getting better. After 5 to 7 backtests I’m back at the wonderful duration I had this morning. Actually 2 minutes and 42 seconds instead of 3 minutes 20 seconds in 2700. It’s the same behavior like this morning. The duration kept coming down with each backtest. I do just start the backtest, take note of the start time, wait till the backtest is done, take note of the end time. Repeat. I’m absolutely doing nothing to increase the backtest performance. So, what is going on? Kind Regards |
| [2024-12-04 16:56:38] |
| User61168 - Posts: 442 |
|
Backtesting speed is: 100000 Just curious which option are you using in Market Replay? In my experience from past several years, 100K speed is just astronomically high for any of the available options for results to be accurate. Could you or someone please validate my understanding that any changes to CUI ms setting should have absolutely no impact on backtesting results when using "Calculate at every tick" ? My CUI is set to 50ms for last 2-3 years and I do not use advanced studies like footprint etc and no chart overlays and have only one chart in one chart book open at all times during market replays. Also, "Same as realtime" option has been just hopelessly useless for me as there is a vast difference in results 1) between "Same as realtime" and "Calculate at every tick" 2) change in speed results in different backtesting results. Completely useless and frustrating. Since last 12+ months, I am using only "Calculate at every tick" option at no more than 240 speed to maintain consistency in my results. All the other 3 options in market replay are completely useless (to me and my algos). p.s. I am still on version 2691. The last thing I want is to upgrade to the latest version and invalidate all my past basktesting results. One year of simple single-chart backtesting with zero indicators takes me about 50+ hours of running market replay in "calculate at every tick" mode ! Date Time Of Last Edit: 2024-12-04 17:04:39
|
| [2024-12-05 08:49:34] |
| User569373 - Posts: 27 |
|
About speed of 100000: SC is suggesting that on https://www.sierrachart.com/index.php?page=doc/Backtesting.phphttps://www.sierrachart.com/index.php?page=doc/Backtesting.php paragraph “Replay Back Testing - Automatic” for replaying a single chart. It’s just my way of telling the study to run as fast as the CPU core is capable to do. Running the study with a SC process CPU core utilization of, as an example 4%, when it can run at 16% is a waste of time. Chart Update Interval: The results in the Trade Activity Log (Statistics, Trades) need to be the same, but the Replay takes more time. Running my study takes often 3 minutes 20 seconds with CUI 200 ms with SC 2712. Increasing CUI to 1500 ms brings the duration to 4 minutes 29 second. I understand different Replay Modes giving different results. I do not understand using the same Replay Mode but getting different results at different speeds. That feels odd. A study gets called at each tick. My study is tick based. It’s different to a study which calculates just when the bar got closed. So, I do run the study with Replay Mode: Calculate At Every Tick/Trade). It’s just natural to me to use this mode. I did several tests. The tests did run at speed 100000 and speed 240: Standard Replay - Different results at each run at same speed - The number of trades placed is totally unrealistic with my study Calculate Same as Real Time: - Different results at each run at same speed - The number of trades placed is not realistic with my study Accurate Trading System Backtest Mode - Different results at each run at speed 240 - Same results at each run at speed 10000 (over 5 runs) - The number of trades placed is way more realistic with my study Calculate At Every Tick/Trade (High CPU Usage) - Same results at speed 240 (over 5 runs) - Same results at speed 100000 over more than 1000 runs on three different computers of different ages - Funny: I figured out the duration of the replay is almost equal at speed 240 compared to 100000. Since the SC process CPU core is running at his limit (as with speed 100000) they take the same amount of time. My default replay settings: Charts to Replay: Single Chart Replay Mode: Calculate At Every Tick/Trade (High CPU Usage) Speed: Whatever speed gives the fastest result. 100000/240 is the same for this specific study Skip Empty Periods: checked Start Paused: I prefer checked I have one chart in one chartbook. Time & Sales window is open. Trade Activity Log is open. Trade Orders window is open. The T&S, TAL and TO window do not make much sense while replaying at such high speed because you can’t digest the information as fast, they cum in. But that’s my window setup while backtesting slowly. Since I’m currently just interested into the Profit/Loss and the number of trades placed, I disable the chart drawing part of my study. I tried one time replaying with active chart drawing(s) of my study. One run takes more than half an hour instead of 3 minutes 20 seconds. I also disable printing to the SC Message Log in my study as long I'm not debugging. In Visual Studio Community Edition 22 (free) you can attach the performance profiler (Debug -> Performance Profiler) to the SC process and let the study replay run. After a certain amount of time (few minutes), you stop the performance profiler, and you get a summary where your study is spending the most time in code. Although the performance profiler is on the Debug menu you can run the study as release. Currently I run the study from 08:30 till the end of the ES Day session. It often takes 3 minutes 20 seconds. Depends on the day I choose. SC updates: It is advisable to backup/copy/create archive (zip) of the C:\Sierrachart folder BEFORE updating it. If you can restore the Sierrachart folder, you can play with whatever you like to do in Sierrachart. In case you use a backup software you need to perform a restore in another location. If the restore is successful then and just then you know the backup is useful. You don’t’ need to do this with each backup. The moment you know the backup software is doing its job correctly you just need to do a test restore from time to time. Date Time Of Last Edit: 2024-12-05 09:43:56
|
| [2024-12-05 17:04:24] |
| User569373 - Posts: 27 |
|
@User61168 About 2) change in speed results in different backtesting results
This should never happen with Replay Mode: Calculate At Every Tick/Trade (High CPU Usage) replaying one single chart. Although it can happen replaying multiple charts. Since you are using “Calculate At Every Tick/Trade (High CPU Usage)” I guess you should read: “Consistency Between Back Tests” on Auto Trade System Back Testing Date Time Of Last Edit: 2024-12-05 17:24:10
|
| [2024-12-06 00:52:42] |
| User61168 - Posts: 442 |
|
Funny: I figured out the duration of the replay is almost equal at speed 240 compared to 100000. Right. Higher speed does not add value in reducing the overall replay times to complete the backtest. That is why I have intentionally made the decision to stick to 240 speed and with just single charts. Calculate at every tick is the only option that has been "consistently consistent" across back test results.
|
| [2024-12-07 05:32:57] |
| User61168 - Posts: 442 |
|
In Visual Studio Community Edition 22 (free) you can attach the performance profiler (Debug -> Performance Profiler) to the SC process and let the study replay run. After a certain amount of time (few minutes), you stop the performance profiler, and you get a summary where your study is spending the most time in code. Although the performance profiler is on the Debug menu you can run the study as release.
Does this technique also apply to simple alert trading system studies? I am a 100% Simple Alert automated trader with a portfolio of 12-20+ entry and exit strategies coded in a single alert study. I have intentionally not engaged in spreadsheet or ASCIL trading systems nor do I use any of the 100s of available charting tools/indicators or any of the advanced footprint/dom/number bar/market profile features of SC etc. Date Time Of Last Edit: 2024-12-07 05:33:30
|
| [2024-12-07 09:41:48] |
| User569373 - Posts: 27 |
|
All I wrote is about ACSIL. I never used alerts or spreadsheet systems. Without ACSIL the Visual Studio part isn’t helping you. Date Time Of Last Edit: 2024-12-07 09:43:09
|
| [2024-12-09 19:39:55] |
| Sierra_Chart Engineering - Posts: 19460 |
|
If you are having any issues related to new Sierra Chart timer functionality, then update to the latest version. Instructions: Software Download: Fast Update By default the Windows-based timers will be used instead. If Sierra Chart was under a high load maybe this explains why these windows were not visibly updating: Time and Sales, Trade Activity Log, Trade Orders Window 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, use the Teton service: Sierra Chart Teton Futures Order Routing |
To post a message in this thread, you need to log in with your Sierra Chart account: