Previous Bar Price Changes

[tradertom]

I am seeing this ALL the time. After a bar is completed and the new bar begins, the previous bar's prices change. Open seems to stay the same, but the close, high and low often change. See the images below. The image on the left is a 5-minute chart snip (in Excel) just after the bar closed. The image on the right was snipped a minute or less later. Note that the previous candlestick high and close have changed significantly from their previous values. As a point of comparison, QCharts painted the bars exactly as the first image below and it stayed that way.

To correct this, I have to switch to a different symbol or change the market hours or chart timeframe and then back again for QLink to requery the data, but if I do that too soon after the erroneous bar, the bad bar comes back. Needless to say, changing symbols or market hours over and over is tiring, tedious and should not be necessary. I have also noted that the source data for these charts changes after the bar closes so it is not an Excel chart issue.

Can someone please correct this anomalous behavior in QLink? QLink should produce a chart in Excel that matches QCharts chart (for the same symbol, timeframe, market hours, etc) and not be changing the previous bar's H,L or C after it has closed. In other words, it should mimic the data displayed in QCharts or eSignal.

Thanks,
Tom

[tradertom]

To clarify:

The symbol I was charting in the example is NQ #F. It is not a delayed symbol. This is a problem on all intraday timeframes that I use, not just the 5 minute timeframe. I do not know the impact on Daily or Weekly bars.

Once an erroneous bar is painted wrong, it stays wrong; it never self-corrects unless the data is requeried. This leads to charts and prices that look very different from the same chart in QCharts.

The frequency of this behavior is about every third bar but it is not constant or predictable. Four bars may pass just fine, then start changing on the fifth and sixth bar. It seems to change within the first minute or so after a new bar has painted.

Thanks,
Tom

[tradertom]

So, no response? Here's this morning's QLink flase data. Top image is the chart as it painted in real-time and the bottom image is after refreshing the data (changing timeframe). Note the different shape candlesticks.

I can keep documenting this if you need more data.

Thanks,
Tom

[tradertom]

Still no response... Jay? Anyone?

This morning, I lined up my QLink spreadsheet chart next to QCharts and took screen images. The QLink charts are the ones with the parchment background while the QCharts charts are the ones with the greenish background. The left pair of images shows QLink data (with changing previous bar prices) and the pair on the right shows QLink data after requery of data (by changing the chart timeframe and back again). This is the NQ futures on a 5 minute chart.

On the QLink charts (parchment background), the left one has candles that differ from its paired QCharts chart. Counting from the left, note candles 3, 4, 5, 6, 7 and 8 differ on the QLink chart when compared to the QCharts data. The first two candles appear the same because I requeried their data after the second candle was painted prior to this demo. The right hand pair of charts shows the QLink data after requery. I also requeried the QCharts data by changing its timeframe then changing back (the "control" of this study). Note the candles listed above are correctly displaying on the right-hand parchment chart of QLink. This indicates to me that the previous candle's incorrect data is carried forward after it is no longer the previous candle.

This is most often seen when the current candle is painting higher or lower with some distance from the previous one (e.g. a bigger move, relatively speaking). When the current candle paints, the previous candle's high, low or close seems to shift in the direction that the current candle is painting. It IS the QLink data and not an artifact from Excel. I have watched the QLink quote values change while the new candle is being formed. It also seems to occur most frequently in the early minutes of the current candle's life. After a minute or two, changes in the current candle don't seem to affect the previous one any longer. As I mentioned, this is on the 5 min chart but I have also noted this behavior on other timeframes as well

I am wondering if anyone else sees this and also if eSignal (at least Jay) is aware of this fully reproducible bug in QLink. If so, are there plans to fix this?

Thanks,
Tom

[tradertom]

Buhler? Buhler?

Below is more specific evidence of a bug in QLink changing previous bar data after the bar has completed. The two images below represent the QLink data from this morning on NQ futures, 5 minute interval. The lowest row of each (row 0) represents the currently painting bar so it's price should be changing. However, please look at the gray area which is OHLC data for the first half hour or so this morning.

The upper spreadsheet image is how QLink recorded the data with no intervention. In other words, the top spreadsheet image is simply a screen print of the data as it painted with no requery or refreshing of the data. The lower spreadsheet image is the QLink data AFTER requerying the data (forced by changing the QLink interval and back to 5 minutes).

Please compare the upper and the lower spreadsheet images (specifically the High, Low and Close values). Specifically, compare rows 8, 7, 6, 5, 3 and 2. Only two rows out of eight (not counting row 0) are correct without requerying. Once a row is finished, it should NOT change. I don't understand how a high, low or close on a bar can somehow change after that bar has completed painting... They do change though, almost every time. And I am certain that they change several times before they seemingly settle down but with incorrect data. (The two spreadsheets below indicate that it is the QLink data that is incorrect and not some artifact of Excel.)

Jay, any comment on this, please?

[JayF]

Hey Tom,

Sorry for the extended delay on a reply. We are looking into this issue and will correct it once it is nailed down within the code.

[JayF]

Thus far, we haven't duplicated it yet. Here's a screen capture where the left side was built based on locally captured data (i.e. streamed data over time) and the right side was after a refresh (switched to 4-min and back to a 5-min.) I also tried an Excel and QLink restart to a new spreadsheet and had the same results.

Because this test PC was on our internal network (essentially the same network as our feed), it likely not to show any discrepancy if the result of the differences are caused by packet loss or other such data drop. We will retest this again next week using a wireless network through a local ISP and see if the problem occurs through that method.

[tradertom]

Jay,

Thanks so much for checking into this. I appreciate you getting back to me. If you need more information, I'd be happy to help.

Thanks!
Tom

[JayF]

Ran a spreadsheet for the whole day on NQ #F, 5-min (Time Range of 9:30 - 16:00 ET) on a wireless DSL connection and refreshed it after the close, and the entire dataset matched the refresh from the server. So, I still haven't see the issue yet. Perhaps our settings simply don't match (i.e. I'm not using cell referencing, but perhaps you are?)

Tom, could you attach a zipped example of the spreadsheet you are seeing this on? It doesn't have to have your full workbook, strategies, etc... Just something simplified down that still shows the issue.

Thanks for the help in tracking this down.

[tradertom]

Hi Jay,

Below is a screen shot before and after refreshing on all sessions NQ 5 minute data on a simple spreadsheet (which I will zip and attach in the next posting). Note that even in thin trading, the fourth candle from the right differs between the pre-refresh chart on the left and the post-refresh chart on the right.

I will continue to check this spreadsheet and post more pictures if you need me to. Yes, I am doing cell references, including the interval and the candle number.

Thanks,
Tom

[tradertom]

Jay,

Attached is the spreadsheet containing the embedded chart from the previous posting.

Thanks,
Tom

[tradertom]

Apologies for the large, wide screen shot... Below is a better representation of the differences... top image is before refresh and the bottom one is after refresh. Note that this was taken during evening hours. NQ futures, 5 minute chart from the same spreadsheet as below.

Thanks Jay!
Tom

[JayF]

I was able to duplicate this issue once today, but in digging into my network log, I did have the wireless network drop during the time of the problem. This doesn't look like anything systemic, because when I run this on a hard wire DSL line, the data seems to flow just fine. So far the only correlation I've seen is related to packet loss / network issues.

Anyone else seeing similar issues to Tom's?

[tradertom]

Thanks Jay for testing this. I am not sure how dropped packets or network issues could change the previous bars that have already completed though. I don't understand how the data stream is formatted so perhaps that is why this makes no sense to me at all. Seems if a bar is in the past, the software (QLink) shouldn't allow it to be changed after its closing time has passed. QCharts does not do that at all. I would think that QCharts and QLink should paint the same shape of bar patterns and produce the same indicators and moving averages for a given symbol and timeframe but this is not at all the case. I guess I could move my office to the eSignal server room just to get a solid connection ;-) but that might not be the best solution! (Unless your cafeteria is nicer than mine!)

I would hope that others are seeing this on their computers as well. It is certainly a constant frustration for me because it shows up so very often. It alters indicators and moving averages of course, because the data is simply wrong. I hope there is something that can be done to filter these extraneous data changes within the QLink software.

Thanks,
Tom

[cpatte7372]

Originally posted by tradertom
I would hope that others are seeing this on their computers as well. It is certainly a constant frustration for me because it shows up so very often.
Tom

OMG, I'm sooooo glad its not just me having this problem. If you look at past threads I posted on this issue you'll find I've spent literally days - AND I MEAN DAYS - on the phone with esignal support trying help resolve this issue. In the end, the told me to submit a thread to this forum.

JAYF has been great trying to help resolve this issue but he couldn't help in the end.

TOM, I feel your pain.

I would love this to be resolved, as it sooooooo frustrating...

Cheers

Carlton