Merge backadjusted still flawed

[Harry]

Please let me know, if you could reproduce it.... thank you.

[Harry]

Looks similar on my notebook: All historical data for the rollover day session after midnight local time has a false negative offset. Real-time data is correct.

Chart settings: CL MergeBackAdjusted, Nymex ETH Session Template, Rollover Date March 19, Offset 0.30

There was also a false negative offset that I observed prior to midnight. However, once the day break local time had occured, data prior to day break was correct as shown below. False offset occurs only after midnight local time.

[NinjaTrader_JoshP]

Harry,

The "gap" at session break is expected. That is simply what the data shows and that is what is displayed.

The gap down at PC midnight seems to only happen when you are loading historical data on that portion. If you left it running all in real-time it should have all looked correct. We will look into that issue.

[Harry]

Hi Josh,

thanks for answering. But I totally disagree. The chart displays false data for the new contract on rollover day.

On rollover day you want to trade the new contract and this is impossible, because the data is false.

The gap is not expected. If I want to trade CL today, I simply cannot use NT 7.

Harry,

The "gap" at session break is expected. That is simply what the data shows and that is what is displayed.

The gap down at PC midnight seems to only happen when you are loading historical data on that portion. If you left it running all in real-time it should have all looked correct. We will look into that issue.

[NinjaTrader_JoshP]

Harry,

As demonstrated in my prior post, the gap is simply part of the data at session break. When viewing a "vanilla" contract unmerged you already have the gap. When you try to do something like merging it to a prior contract, that will not remove gap downs that are part of the data and it rightfully should not be expected to do so. Gap downs from the data will be preserved. The only gaps that should be removed are the contract gap differences between different front months and that is done correctly as seen by the different prices at session break.

The issue with what happens going forward from PC midnight with historically loaded data is a completely different issue and we are looking into that.

[Harry]

Please watch the chart #13 carefully

Hi Josh,

thanks for answering. Please have a look again at the charts that I posted. If you closely look at the chart as per post #13, you will see that before the day break occured, there was false historical data displayed. Not only after midnight.

The chart shows CL 05-10 in mode MergeBackAdjusted. Old contract prior to session break is correctly backadjusted (up). The new contract after session break shows a negative offset before midnight. Realtime data is ok.

When connecting again after midnight, the false negative offset for the historical data of the new contract between session break and day break disappeared. However offset was false for the new contract after daybreak.

Harry,

As demonstrated in my prior post, the gap is simply part of the data at session break. When viewing a "vanilla" contract unmerged you already have the gap. When you try to do something like merging it to a prior contract, that will not remove gap downs that are part of the data and it rightfully should not be expected to do so. Gap downs from the data will be preserved. The only gaps that should be removed are the contract gap differences between different front months and that is done correctly as seen by the different prices at session break.

The issue with what happens going forward from PC midnight with historically loaded data is a completely different issue and we are looking into that.

[NinjaTrader_JoshP]

Unfortunately I was unable to see your behavior even when I was connected before midnight as in my post #14. Were you rolling your charts completely in real-time? At what time was the chart opened? At what time was initial data loaded?

[Harry]

This is what I did: First changed rollover date and offset via instrument manager, then added new instrument CL 05-10. This was after session break for CL, but before midnight local time, because I calculated the offset from the difference of the close for CL 04-10 and CL 05-10 on March 18.

Then I connected and opened a new chart. Now this was instable. When connected ro IB, I always got the false chart, when not connected I sometimes got a correct chart and somtimes a false chart. At one stage I had both correct and a false chart on my PC. But could not correct this through a second time loading as for data gaps and spikes.

After midnight local time, the bug appeared to be stable = always false.

Unfortunately I was unable to see your behavior even when I was connected before midnight as in my post #14. Were you rolling your charts completely in real-time? At what time was the chart opened? At what time was initial data loaded?

[NinjaTrader_JoshP]

Harry,

Please delete all of your cache and historical data to start fresh. When you get an improper chart when connected to IB, please do not do anything to try and fix it. Zip up your DB folder and please send that in before trying anything else.

[Harry]

Ok, for the next time. Today loaded CL-05-10 new to my second PC, also shows the negative offset for today's data, but that is confirmed anyhow.
So I will try already to replicate problems on 2 PCs, just to be sure.

[NinjaTrader_JoshP]

Thank you. Please be sure to note at what point in time the chart goes from historical to real-time as this transition may be of importance.

[Harry]

Hi Josh,

I just want to point out that Mike 12345 has reported the identical problem for CL 05-10 on rollover day and shown it on a screencast. So you can actually see how the price jumps up when real-time data follows historical data. It is #1 of the following thread:

http://www.ninjatrader-support2.com/vb/showthread.php?t=27140

[NinjaTrader_JoshP]

Harry,

That is the same issue we already looked into on post #17 and #20 and *should* be working better in next version.

[Harry]

Correct symbol for BC (with IB)

Will test MergeBackAdjusted tomorrow for BC, as tomorrow is rollover day. Just want to check the symbol. I use COIL||||1. Is this correct?

[NinjaTrader_JoshP]

Harry,

I had one of our techs check for you. It would be accurate.