Volume Bars appear to be built incorrectly

[pbastowski]

Hello,

I use Volume bar charts in my analysis. I have noticed that NT6 is not producing the volume bars correctly. That is, while building the volume bars it is loosing some volume along the way and producing less candles than it should be.

The problem is most apparent on bar charts with low volume per candle, such as YM #F with 49 contracts (volume) per bar. It appears that when each candle is built, and an excess volume is available in a tick, over and above the 49 contracts required per bar, the excess contracts may be simply thrown out. Now, I don't know that this is the exact case, but it appears to be something like this.

I have compared the NT6 volume bar charts to volume bar charts created with MultiCharts, which I know handles the situation correctly. The result are that there is less volume bars on NT6 charts for the same time period than there is in MultiCharts.

This throws my indicator out producing unreliable results.

Is this a known problem? I am happy to provide more info if needed.

Thanks,

Paul

[NinjaTrader_Ray]

Hi Paul,

I will have my developer check on this but I am 99.99% confident that volume is not discarded. Comparing MC to NT is apples to oranges.

You can learn more about why bars can be built different from vendor to vendor here.

http://www.ninjatrader-support.com/HelpGuideV6/HowDoesNinjaTraderBuildChartBars.html

[pbastowski]

Hi Ray,

Thanks for the quick reply. I have read the explanation of how NT aggergates contracts to create volume bars.

Different would be OK, such as slightly different looking bars. Because bars are missing I am lead to believe that volume is being lost.

I noticed a small difference when I was using charts with 2401 contracts per bar, but just recently I looked at faster charts with only 49 contracts per bar and there the difference was much more visible. If we loose a bit on each bar then the more bars we have the more visible the difference becomes.

For example, for a 49 volume bar chart, if we get three ticks:

tick 1 = 20 contracts
tick 2 = 15 contracts
tick 3 = 30 contracts

total of 65 contracts

The first two ticks add up to 35 contracts and we need another 14 to complete the 49 contract bar, leaving 16 contracts for the next bar. The next bar should be started immediately with the 16 remaining ticks and then agregating further incoming ticks to complete it and start the next bar, and so on...

If the 16 contracts from Tick 3 are lost rather than used to start the next bar, then we will see a loss of volume as bars are created resulting in a reduction in visible bars. This is going to be less visible on high contract numbers per bar and more visible on charts with small number of contracts per bar.

I will try to prepare a screen show with some analysis for you.

Paul

[NinjaTrader_Ray]

I understand what you are saying and appreciate your willingness to provide further evidence. The only course of action of an value is for my development team to run tests. I have already asked them to run this test and they will later this week at which time I will report back.

[NinjaTrader_Ray]

We confirmed that there are no lost ticks. If you have 1 volume bars and a trade for 10 comes in, 10 bars are built.

You can have a seriously different chart when dealing with tick based intervals comparing one feed to another. IB for example, sends out snapshot data, maybe once every 7/10 of a second where Zen-Fire and Trading Technologies sends out all ticks, could be 40+ per second. Patsystems throttles data where you may lose 20% of the ticks.