Shipped Develop COPPER Multiple Prices at odds with shipped roll calendar

[littlebigfis]

I was attempting to update the roll calendar and multiple prices for COPPER and noticed the oddity that throughout the shipped roll calendar the current contract is one roll behind shipped multiple prices.

Ending Shipped Roll Calendar:

DATE_TIME current_c next_c carry_contract
2021-06-24 20:00:01 20210700 20210900 20210900
2021-08-24 17:00:01 20210900 20211200 20211000
2021-11-19 17:00:01 20211200 20220300 20220100

Shipped Multiple Prices at time of rolls

DATETIME CARRY CARRY_C PRICE PRICE_C FORWARD FORWARD_CONTRACT

2021-06-24 21:00:00 4.3055 20210900 4.3145 20210700 4.3055 20210900
2021-06-24 22:00:00 4.305 20211000 4.3005 20210900 4.3035 20211200
.......

2021-08-24 18:00:00 4.26 20211000 4.261 20210900 4.2615 20211200
2021-08-24 19:00:00 4.26 20220100 4.263 20211200 4.2615 20220300
.......

2021-11-19 18:00:00 4.392 20220100 4.399 20211200 4.3875 20220300
2021-11-19 19:00:00 4.392 20220400 4.385 20220300 4.3875 20220700

Similarly when, I run: sysinit/futures/rollcalendars_from_providedcsv_prices.py to build roll calendars from multiple prices, the problem of mismatch persists. Here is a just one example where multiple prices has rolled on 2023-03-07 at 20:00:02 from the 20230300 to the 20230700 contract yet the roll calendar derived from this data shows we continue to hold 20230300 on 2023-03-07 at 20:00:02. How can this be? My understanding is that multiple prices reflects the rolls Rob's PST executed. How does the script that works successfully for all the other roll calendar builds from multiple prices that i have worked on so far, fail to produce the same results? Rob's roll calendar seems to confirm the same problem.

What am I missing and how do I fix this?

[tgibson11]

I'm not seeing any inconsistencies between what you show for the provided roll calendar and the provided multiple prices. Can you highlight what you see as inconsistent?

Here is a just one example where multiple prices has rolled on 2023-03-07 at 20:00:02 from the 20230300 to the 20230700 contract yet the roll calendar derived from this data shows we continue to hold 20230300 on 2023-03-07 at 20:00:02

Other than this sentence, you didn't actually provide an example. However, since Rob's provided data for COPPER stops in 2022, this is entirely your own data. If you think the script is doing something wrong, you're going to have to explain what you're doing.

Furthermore, I'm not sure why you would be generating roll calendars from your own generated multiple prices. Once you have the multiple prices, you don't need roll calendars any more.

My understanding is that multiple prices reflects the rolls Rob's PST executed.

This is presumably true of the provided multiple prices. It obviously will not be true of the multiple prices you generate yourself.

How does the script that works successfully for all the other roll calendar builds from multiple prices that i have worked on so far, fail to produce the same results?

Different input, different output.

Rob's roll calendar seems to confirm the same problem.

I don't see how it could, since you're talking about data that is over a year past what Rob provided.

I suggest that you first explain where you are in the process, what you have done, and what you are trying to do. Then focus on one specific example, and point out specifically what you see as a problem.

[littlebigfis]

Here's an example of the inconsistency I am seeing in COPPER:

Rob's roll calendar shows a roll on 2021-06-24 to the 20210700 contract
Rob's multiple prices shows a roll on 2021-06-24 to the 20210900 contract
Rob's roll calendar shows roll to 20210900 occurring on 2021-08-24 not 2021-06-24

This inconsistency persists throughout the roll calendar.

I am updating multiple prices and roll calendars from stale data. First, I am updating roll calendars to match the end date of multiple prices. Next I build a roll calendar from current prices. Lastly, I splice multiple prices using updated roll calendar and current prices. My update roll calendar update from multiple prices appears to have the same anomaly as Rob's.

[tgibson11]

Rob's roll calendar shows a roll on 2021-06-24 to the 20210700 contract

No. It shows a roll from 202107 to 202109. That's what I mean about roll calendars being confusing to interpret. Current and next refer to the situation before the roll.

Rob's multiple prices shows a roll on 2021-06-24 to the 20210900 contract

Yep.

Rob's roll calendar shows roll to 20210900 occurring on 2021-08-24 not 2021-06-24

Also no. See above.

[littlebigfis]

Ah of course, next contract! I'm an idiot

[tgibson11]

I don't know if this will help, but before Rob had the production system all nicely automated, I used to manually update the roll calendars whenever I did a roll. I can't remember the exact reason, but it's not important here.

Anyway, the point is I always had to add a dummy row for the day after the roll that reflected the new, post-roll current, next, and carry contracts. I imagine, that dummy row as being carried down until the roll that it describes actually happens, then it is frozen in place on the roll date.

[littlebigfis]

I like that idea! It definitely will help keep the roll calendar flow straight in my mind. Thank you for your help and patience with a rookie error...