Inconsistent RSI calculations - very worrying

Questions about MCFX and MCFX Data Feed.
MKlein
Posts: 147
Joined: 14 Aug 2007

Inconsistent RSI calculations - very worrying

Postby MKlein » 06 Dec 2007

There appear to be serious problems with the calculation of RSI in MCFX. I often notice that the RSI curve doesn't look right - for example the direction of the RSI may reverse for no apparent reason when the price curve gives no justification. Reloading the window often causes the value of RSI to change significantly. For example, I hit "reload" on my 2-day EUR/USD chart, and RSI dropped from 72 to 46 with no change in spot price.

This is a very worrying bug: we need to be able to trust our indicators.

M Klein

User avatar
Marina Pashkova
Posts: 2758
Joined: 27 Jul 2007

Postby Marina Pashkova » 11 Dec 2007

Dear M Klein,

What you have observed is not inconsistent RSI calculations. We have run numerous tests and can guarantee that the calculations are perfectly consistent.

The reason why you see different results after reload, is that reload gets new data into your chart. Whenever you request the data, the reference point is the current moment and the request goes from the current moment and into the past - not the other way round. Thus, once the reference point has shifted the data will be updated and the study will have different input data and hence different calculation results. And this is exactly what happens when the reload is done.

MKlein
Posts: 147
Joined: 14 Aug 2007

Postby MKlein » 11 Dec 2007

What you have observed is not inconsistent RSI calculations. We have run numerous tests and can guarantee that the calculations are perfectly consistent.
I currently have two GBP/USD 30m charts running in two separate workspaces, using exactly the same indicators. I refreshed both charts a couple of hours ago, and both showed the same price and RSI at that time. There's now a discrepancy of nine points between the two. There's clearly a problem.
The reason why you see different results after reload, is that reload gets new data into your chart. Whenever you request the data, the reference point is the current moment and the request goes from the current moment and into the past - not the other way round. Thus, once the reference point has shifted the data will be updated and the study will have different input data and hence different calculation results. And this is exactly what happens when the reload is done.
Of course the RSI will be changed if the data it is based on has changed, and it's true that there are errors in nearly any arbitrary period due to the problems with your servers. However, it's possible to find periods of good data, and there are still inconsistencies.

When you say that refreshing the chart changes the direction in which the data is loaded, surely that shouldn't change the close price of any given bar, assuming the bar close occurred when your server was serving. And presumably the RSI calculation is still performed in the same way once the data is downloaded.

MKlein
Posts: 147
Joined: 14 Aug 2007

Postby MKlein » 11 Dec 2007

I currently have two GBP/USD 30m charts running in two separate workspaces, using exactly the same indicators. I refreshed both charts a couple of hours ago, and both showed the same price and RSI at that time. There's now a discrepancy of nine points between the two. There's clearly a problem.
Here are some example screen grabs. The two charts in different workspaces were refreshed within a few moments of each other, and the two screen captures were made roughly two hours later, again within a few moments of each other. The red line in subchart 4 is RSI, and there's a discrepancy of over seven points ( 26.45 vs 19.08 ) between the two charts.

There is a one pip difference between price in the two charts in the final bar (because the price had moved in the few seconds it took to make the two screen grabs), but that doesn't account for the RSI difference.

I manually checked the closing price of each bar in the relevant range (13 bars) and there was no difference between the two charts.
Attachments
GBPUSD1.png
(37.04 KiB) Downloaded 883 times
GBPUSD2.png
(43.13 KiB) Downloaded 895 times

User avatar
Marina Pashkova
Posts: 2758
Joined: 27 Jul 2007

Postby Marina Pashkova » 13 Dec 2007

Dear MKlein,

The screenshots you attached do not illustrate inconsistency of RSI calculations because they are about 2 hours apart. If you could post screenshots showing the same time but different RSI values that would demonstrate the inconsistency of calculations.

Best regards.

MKlein
Posts: 147
Joined: 14 Aug 2007

Postby MKlein » 13 Dec 2007

The screenshots you attached do not illustrate inconsistency of RSI calculations because they are about 2 hours apart. If you could post screenshots showing the same time but different RSI values that would demonstrate the inconsistency of calculations.
Please re-read what I wrote and compare the two charts. The screenshots were made within seconds of each other.

M Klein

User avatar
Marina Pashkova
Posts: 2758
Joined: 27 Jul 2007

Postby Marina Pashkova » 13 Dec 2007

Dear MKlein,

I re-read your post, but the way the screenshots are scaled it is hard to see if it is indeed the same time or not. Could you please adjust the chart in such a way that would make it clear? Thank you.

MKlein
Posts: 147
Joined: 14 Aug 2007

Postby MKlein » 13 Dec 2007

I re-read your post, but the way the screenshots are scaled it is hard to see if it is indeed the same time or not. Could you please adjust the chart in such a way that would make it clear?
You can observe that these are 30 minute charts and compare the bars in the main (price) window of each. If, as you say (and I've no idea on what grounds you make that claim), the screenshots were taken two hours apart, you would expect there to be four additional bars chronologically in the later chart. There are not. It's quite clear that the active bar in each is the same.

How would you like me to re-scale the screenshots to make them clearer?

User avatar
Marina Pashkova
Posts: 2758
Joined: 27 Jul 2007

Postby Marina Pashkova » 14 Dec 2007

Dear MKlein,

We still haven't been able to reproduce the problem you are describing.

We would like to check the RSI behavior on your computer through HelpDesk, a remote desktop connection utility.

Could you please contact us through LiveChat http://messenger.providesupport.com/mes ... pport.html and let the customer support representative know what the issue is. He/she will then give you instructions on how to download and run HelpDesk.

Please note that the customer support hours are 6:30 am - 4:00 pm EST, Monday through Friday.

Best regards.

MKlein
Posts: 147
Joined: 14 Aug 2007

Postby MKlein » 17 Dec 2007

We would like to check the RSI behavior on your computer through HelpDesk, a remote desktop connection utility.
I will if I'm able to, but I'm extremely busy during your office hours.

Here are two more screen grabs showing an RSI discrepancy. I took care to include the clock in the image - you can see that these images were made eleven seconds apart.
Attachments
usdchf2.png
(52.46 KiB) Downloaded 877 times
usdchf1.png
(53.05 KiB) Downloaded 874 times

User avatar
Marina Pashkova
Posts: 2758
Joined: 27 Jul 2007

Postby Marina Pashkova » 17 Dec 2007

Hello MKlein,

On these screenshots it's really obvious that something weird is going on with the calculation of RSI. However, since we still can't reproduce the problem on our end we would appreciate if you could find time and contact us through LiveChat so that we could use HelpDesk to see what's going on on your computer.

Best regards.


Return to “MCFX”