Inconsistent Tick Data - Any recommendations on how to identify algorithmically

[parkinsonbr]

Context:

• Platform: NT8
• Charts: Two (One on a VPS, and the second local machine) all with the same settings
• Bar Type: Tick Bars - 250 ticks

Issue:

I see this happening 3-4 times for month. What I am observing is that the tick bars don't appear to be form identically causing indicators to provide different values. There is regularly a touch of variation, then then periodically there is a larger different. This has caused me to lose money a few times this year when I wasn't able to identify the issues quickly enough.

Attempts to correct and results:

• I refreshed charts by reloading all historical data on both VPS and my local machine. Of note the indicators and their settings are identical Result - Worked initially, but within 30-60 minutes decent size differences had crept back in. So didn't fix the issue
• I then rebooted NT and reloaded all historical data Result - Appears to have fixed it on the second attempt.

Monitoring:

Each morning I check NT's log to verify no disconnects have taken place since most recent reboot. If a disconnect has taken place then i reboot NT. For the situation above there hadn't been a data disconnect other than that tied to rebooting NT and differences still showed up.

Questions:

1. Is there a recommended way to algorithmically identify when data isn't building tick bars or bars in general correctly? Any direction would be great.
2. Why would machines that share the same data provider differ in the creation of Bars? Only thinking about this when there are large differences.
3. Would Kinetick's data offer a more reliable solution to this problem?
4. Is there anything being worked on by Ninjatrader to identify when NT8 has bar creation issues that might relieve users from identifying this issue visually/manually?

Thanks in advance!

[NinjaTrader_Jesse]

Hello parkinsonbr,

In your comparison between two machines there is certainly the possibility that Tick based charts will build differently. If one PC has even 1 tick different that may cause the bars to start building differently from that point on. Generally to do any type of tick based comparisons between two PC's you would need both to start before the session begins and reload all data on both machines. If they both have identical data to start with and then enter realtime they could start building identically so long as all the identical data is received.

If you are starting while data is coming in you will likely see the charts start to build differently after a period of time. It would be nearly impossible to have two PC's start a tick chart at the exact same millisecond in live data.

Regarding your questions, there is not a way I could suggest to detect this situation because the chart is going to be correct based on the data it has to build it. You would effectively need to compare data between machines at all times in order to know if something was off between them. Tick based bars are built from the ticks rather than times so if you have a different starting tick or miss a tick that will cause a discrepancy between the bars based on their starting points.

There could be many reasons why data can differ between machines, some example could include connection problems or timeouts, computer issues, connecting to different servers where one server has different historical data. As noted if you started with any differences or as you mentioned there is regularly a touch of variation then you would see that reflected in how the bars are being built.

I couldn't really say if using a different provider would offer a different result, that would really require knowing the specific problem to know if that would address it. If you are seeing that one machine is missing data sometimes that would be a situation where the logs can be used to look for disconnects or other issues. You can also use the historical data manager to check if there is specific data missing/different. Using that process to find what is different would help to highlight what the solution would be.

I look forward to being of further assistance.

[parkinsonbr]

Hey Jesse! Thanks for the quick response. The answer makes sense and is in line with what i have experienced. I will check into the historical data manager suggestion as that isn't something I have done yet. Thanks again!