UA Roll Trigger Bug

Use this forum to discuss data providers like CSI, charting, or other non testing software.
Post Reply
leslie
Roundtable Knight
Roundtable Knight
Posts: 135
Joined: Tue Apr 12, 2011 11:46 am

UA Roll Trigger Bug

Post by leslie »

This is a general roll trigger bug.
BL2 Milling Wheat is just an easy to follow example, but you can look for your own.

Nature of bug:
Contract Expiry override does not work consistently.

Wheat settings:
Roll Tigger=Open Int
Confirmation=Roll on 1st Trigger
Roll by Days Before Expiry=10
Contract Expiry=15th day of Month One Month Prior

This would mean that a 2004-MAY contract would have to roll on or before 2004-APRIL-5.

Contract failed to roll April 5th, actually rolled on April 19th. This is a bit of a hick up since the Exchange rule states that trading be halted on afternoon of the 15th.
Had you a lot of contracts, could cost a pretty penny.

See data...

20040415 57.75000000 57.75000000 56.25000000 56.50000000 543 5781 200405
20040416 55.75000000 55.75000000 55.50000000 55.50000000 596 3707 200405
20040419 55.75000000 55.75000000 55.25000000 55.25000000 95 3739 200411
20040420 55.25000000 56.25000000 55.25000000 56.25000000 28 3286 200411

The bug is quite general.
Lot of contracts show same bug, that is roll occurs AFTER contract goes off the board even when "Roll by Days before Expiry" is set correctly.


Bogus data = bogus back test result...
Demon

Post by Demon »

I think this is occuring where CSI static data doesn't have an expiration / first notice date for a given market in its database. The issue needs addressing.
leslie
Roundtable Knight
Roundtable Knight
Posts: 135
Joined: Tue Apr 12, 2011 11:46 am

Post by leslie »

Hi Demon,

Yes, good point, this is why I set
Advanced >> Contract Expiry >> "Contact Expiration Rules" manually and set
Roll by days before expiry=10

1 Month Prior to 15th day of month MINUS the 10 days should have triggered roll on 1 Month Prior to on the 5th day of the Month. It did not.

At this time, UA Version 2.7.10.119 is buggy.
See screen clip.

LM
Attachments
milling_wheat_contract_experation_rules_advanced_overrides.jpg
milling_wheat_contract_experation_rules_advanced_overrides.jpg (523.68 KiB) Viewed 5425 times
Josh Reed @ CSI
Full Member
Full Member
Posts: 13
Joined: Mon Mar 07, 2011 1:03 pm

Expiration Editor / Expiration Dates in UA

Post by Josh Reed @ CSI »

LM,
The Expiration Rules are used to know ahead of time when a contract will be expiring, so that we know days before this date when to move. It is not referred to for past historical rolls, the data itself is used, as the last date of the contract is known. This is not the nature of the bug, as you're looking at a roll back in April of 2004. The May 2004 contract ended on May 7, 2004.
This roll was triggered by Open Interest.

With Best Regards,

Josh
Josh Reed @ CSI
Full Member
Full Member
Posts: 13
Joined: Mon Mar 07, 2011 1:03 pm

Contract Expiration Rules assume correct expiration rules

Post by Josh Reed @ CSI »

LM,
When you enter a rule that does not follow the rules that actually occur, you can expects results like this. BL2 does not expire 1 month prior to it's delivery month.

With Best Regards,

Josh
leslie
Roundtable Knight
Roundtable Knight
Posts: 135
Joined: Tue Apr 12, 2011 11:46 am

Post by leslie »

Hi Josh,

Wow, thank you for the answer.
Alas, you are describing a very dangerous piece of software!
Essentially you are saying there are two separate semantics.

First, historic data IGNORES the "...Rolling Overrides" setting.
Most unwise, but let us continue.

Second, for each day, the "...Rolling Overrides" settings ARE applied and MAY modify the Open Interest roll. Fine. This will keep traders quiet since their positions do not go off the board. Traders are happy and will not notice a thing!

But, here is the rub.
Tomorrow, today will have become yesterday.
Each new day, today will fade further into the past.
What will be the rules than?

Is there a 3rd, hidden semantics?
Will you (a) apply the "...Rolling Overrides" rules in the future starting with the past present; or (b) will the historic data rule set revert to the original model #1 --ignore the past rule?

If (a) the rules have 3 different semantics.
If (b) the rules have 2 different semantics.

In any case, the trader will be
--back testing rule #1
--trading rule #2 or rule#2 + (a)

IMHO, this is more then a bug, the 'feature' is a fatal design flaw....
Please give some thought to redesigning it, making the 'generate forward' work correctly; stabilizing the "Roll by Percent" switch effects.

Thank you.

LM
Josh Reed @ CSI
Full Member
Full Member
Posts: 13
Joined: Mon Mar 07, 2011 1:03 pm

Expiration Dates and Rules

Post by Josh Reed @ CSI »

LM,
Your initial claims were of bogus data and bogus results. Yes, you
entered bogus data in the expiration rule editor, you received bogus results.

When using this properly, please present us with a 'dangerous' scenerio with supporting data, to back up your assertion of a 'fatal design flaw'.

Rules change, static data doesn't....

With Best Regards,

Josh
leslie
Roundtable Knight
Roundtable Knight
Posts: 135
Joined: Tue Apr 12, 2011 11:46 am

Post by leslie »

Josh,

Please read concerns about testing rule #1 and trading rule #2.

I can see Father Sluggo's wisdom-- write a back adjuster from scratch.

LM
Post Reply