OnBarUpdate misses prints

[egan857]

Again, not my problem if I have a workaround. But missing prints in OnBarUpdate can't possibly be unique to my setup, not if I get those prints when I use OnMarketData.

[sledge]

Originally Posted by egan857

Again, not my problem if I have a workaround. But missing prints in OnBarUpdate can't possibly be unique to my setup, not if I get those prints when I use OnMarketData.

How many trades a day?

I guess you don't have market replay recording enabled?

There have been reports/complaints about market replay becoming slower because of trades "filling up the database".

There was a recent post from someone doing thousands of trades a day running out of memory(this may have been you).

I don't have Kinetick, but can't you download one of these stocks that have an issue that day, and run it through market replay with this OnBarUpdate/OnMarketData checker program? If that proves out a difference, then NT Support might have something more to chase down.

This could be IB related otherwise.

I have TDA... but I'm not sure how I'm going to figure out which out of 200-500 stocks is going haywire easily with missing OnBarUpdate calls.

[sledge]

Originally Posted by egan857

Again, not my problem if I have a workaround. But missing prints in OnBarUpdate can't possibly be unique to my setup, not if I get those prints when I use OnMarketData.

RE: Missing prints

Is it more pronounced when the market it *glitchy* for the day?

Do you have any filtering enabled in NT?

Tools->Options-> Data Tab-> Filter bad ticks 10% off market?


I guess that could explain why OnBarUpdate (filtered) and OnMarketData might not be same (assuming OnMarketData doesn't use this filter option).

[sledge]

Originally Posted by egan857

Again, not my problem if I have a workaround. But missing prints in OnBarUpdate can't possibly be unique to my setup, not if I get those prints when I use OnMarketData.

I played around tonight in 6E and DX ...

OnMarketData - Last was not always equal OnBarUpdate ( Closes[1][0] )...

This doesn't explain "dropped" things..

I assumed OnMarketUpdate was called first. THen OnBarUpdate.

So I set my local varialbe in OnMarketUpdate and then compared to OnBarUpdate Closes[1][0], and *sometimes* it didn't match.

I was careful with BarsInProgress everywhere, as this was an added series in my indicator (BIP=1).

With 6E, I multiplied it by *1000 to make sure it wasn't some dumb double error issue. The print looked fine.

[NinjaTrader_Brett]

Hello,

Thanks for the testing.

With the NT Core being multi threaded you cannot actually count on OnMarketData() getting called before OnBarUpdate() could be the other way around and also gets more complex when you add a secondary series on which gets fired when.

-Brett

[egan857]

Hi Sledge/Radical,

I experienced this problem with equities, typically smallish ones (100-200K shares per day). The problem tends to happen when the stock is indecisive (30 seconds between trades) and tanking (15-cent drops on $40 stocks, not enough to trigger the 1% filter I have activated).

I don't think that market replay is a valid way to test this. That's probably what Ninja's QA team used to "prove" that everything was alright.

When I was experiencing this problem a few months ago, the market was a lot more volatile than it is now, so it might take a little more difficult to reproduce the problem in August - when was the last time you saw a stock be indecisive AND tanking in the last few weeks? When I switched over to OnMarketData, that situation was common, and the problem immediately disappeared. I did not test the two in parallel; instead, I ran a simple program that printed time and price for every print using OnBarUpdate and then quickly modified it to OnMarketData for the same symbol that was experiencing this problem all day long, and then flipped it back and forth a few times. In all instances, I saw dropped prints (compared to my IB screen) on OnBarUpdate and not on OnMarketData. I figured what was the point of testing one NT product vs another? Why not compare each against an independent source? This can only be done with smallish stocks because otherwise there's no way to keep track of them visually.

[Radical]

I don't think you've mentioned explicitly in this thread yet, are you using Kinetick or IB through NT? Which are you having this problem with?

[egan857]

Hi Radical,

I'm using Kinetick as the primary data feed and Interactive Brokers as the broker. The IB screen shows the data from IB, which is usually (except in the situation we're talking about) identical to the data in NinjaTrader. The problem isn't with either feed, but with OnBarUpdate vs. OnMarketData.

[Radical]

If you're comparing Kinetick data with IB data, I think it's pretty possible that the difference you're seeing is between the feeds themselves, and has nothing to do with NT. IB data is aggregated, Kinetick is not, and it would be very possible to see the difference if you're watching tick by tick.

[egan857]

Hi Radical,

I totally agree that that's a possibility (and mentioned it in post #7 of this thread). If I thought it was just a difference between IB and Kinetick, I would have dropped the issue a long time ago and continued using OnBarUpdate. But, the only time I've ever seen a difference between the two feeds is when I was using OnBarUpdate, and never when I've been using OnMarketData, so that implies a difference between OnBarUpdate and OnMarketData, not a difference between IB and Kinetick.

[Radical]

Frankly I think that could just be a coincidence though. Seeing that behavior just a handful of times could have many different causes, especially if you weren't recording the data at the time. I just think it's a big waste of time to focus all of your energy on OnMarketUpdate vs OnBarUpdate being the problem when really there are other things that could just as easily be the source of the problem. And given that the IB vs Kinetick difference is a documented, known difference, I would say that is more likely the problem.

[egan857]

This is probably why the problem has never been addressed. Everybody thinks it works 100% of the time because it works 99% of the time, and there's no incentive to go hunt for problems, and it's comforting for a trader to think that his data is perfect. If I ever run into CPU problems and need to return to OnBarUpdate, I'll run some real tests and present those results. Until then, I'll leave it as a problem for someone else to prove. I just presented the problem and a description of how to test it. It's not my job (or yours, unless you're experiencing bad fills when your stocks are tanking) to prove it. That job belongs to NT's QA department.