datatest
-Interested User-
Posts: 23
Joined: Feb 29, 2012
|
Posted: Mar 15, 2012 03:12 AM
Msg. 1 of 6
Hello
I have analyzed historical corn contracts (EOD Data) and have seen that quite often the close is higher than the high, or lower then the low. Why is in these cases the high/low not equal to the close? thanks
|
DTN_Steve_S
-DTN Guru-
Posts: 2093
Joined: Nov 21, 2005
|
Posted: Mar 15, 2012 10:12 AM
Msg. 2 of 6
Can you post some examples?
The close value in EOD data (for futures) is the settlement price. Its possible that they are cases where the contract settled outside the High/Low for the day but we will need to investigate it to be sure.
|
datatest
-Interested User-
Posts: 23
Joined: Feb 29, 2012
|
Posted: Mar 15, 2012 10:34 AM
Msg. 3 of 6
yes, I put them in an XLS File, these are 2665 examples of wrong records...
one question I don't understand: you have all this data in your database. why can you not simply run a check and compare where close/open is lower than low or close/open is higher than high? I am spending hours and hours in data validation, data which should be in principle be checked with some minimal quality checks. And if there is an explanation for this specific "pattern", e.g. due to the settlement price, so it should be in your FAQ. I have the feeling I am the first one looking at these data and asking these questions...
|
datatest
-Interested User-
Posts: 23
Joined: Feb 29, 2012
|
Posted: Mar 15, 2012 10:36 AM
Msg. 4 of 6
seems I cannot upload excel, so I send you enclosed a screenshot
File Attached: Corn High Low Errors.pdf (downloaded 1421 times)
|
DTN_Steve_S
-DTN Guru-
Posts: 2093
Joined: Nov 21, 2005
|
Posted: Mar 15, 2012 01:29 PM
Msg. 5 of 6
Hello, I have confirmed that these are cases where the settlement was outside the High/Low.
IQFeed does not stretch the High/Low to include the settlement in this scenario. We don't stretch the High/Low because that allows the 3rd party or end user to decide if they want to do this on thier end (if we did it automatically, they would have no indication that it is the case).
As far as your more general concerns: I can assure you that you are not the first person to look at this data. Our history servers process several million requests for data per day. You are looking at a very small subset of the data that our servers provide. They contain trade/minute/daily/weekly/monthly history for millions of symbols (both current and expired). I don't mention this to make excuses but rather simply to illustrate that it is not as simple as running an SQL query where Low > High or and spitting out a report. With that said, we do have processes in place that check the data and we have a team of market data specialists to investigate and correct errors. Additionally, as in this case, we also investigate all issues reported by customers as well. What you found today wasn't an error. The high/low is left "unstretched" so that the end user (3rd party developer) can decide how they want the data.
|
datatest
-Interested User-
Posts: 23
Joined: Feb 29, 2012
|
Posted: Mar 16, 2012 01:21 AM
Msg. 6 of 6
Hello
Thanks for the investigation. I was assuming that the reason is the settlement price, this is why I wrote above that I am surprised that this answer is not coming immediately or written somewhere, because obviously it happens not so seldom. I was searching in the internet, but could not find an explanation: Why exactly can the settlement price be outside of low/high boundaries?
|