Anomalies-in-Fidelity-Select-Sector-funds-DataSet

If you’re backtesting over a period of more than three years why would you NOT want to have that present??

Unlike all the other WorldIndices DataSets which get dynamically updated every week, that one and "Major U.S. and world indices" are hardcoded. So it's out of date. If you know an up to date source of the select funds why not share it so we could try to include it in the weekly run?

QUOTE:
why would you NOT want to have that present?

Because FSCGX is a discontinued fund that will never trade again. Useless.

Fidelity changes the Select Sector funds once in maybe 10 years, so there's no need to update it weekly. You just need to update it manually every 10 years.

QUOTE:
Fidelity changes the Select Sector funds once in maybe 10 years, so there's no need to update it weekly. You just need to update it manually every 10 years.

OK I'm all ears on how it should be done.

Fidelity typically makes an announcement when they are going to discontinue a fund. They rarely change their lineup, but it does occur on rare occasion. And there's no reason to try to trade a discontinued fund.

By chance you know where we can find these announcements and the current composition of the portfolio?

It’s hardly “useless” to include discontinued funds. It’s a valuable way to prevent survivorship bias.

QUOTE:
... where we can find these announcements and the current composition of the portfolio?

This is a good question. Fidelity once listed the Select Sector funds on a webpage with daily NAV prices, but I don't see that anymore. I think that may be replaced with their mutual fund screener.

I would visit https://www.fidelity.com/mutual-funds/fidelity-funds/overview then select the "Sector" tab. That should give you all their sector mutual funds without using the screener. Now if you have your browser highlight the word "Select", that should reveal which of those funds are "Select Sector" funds (which have more opportunity to profit if you can handle the greater volatility). All Select Sector funds have a 30-day minimum retention period.

You'll also see a Real Estate category in there, but no "Select" label in that category. Most of those Real Estate funds have a 90-day minimum retention period, but the FRESX fund has a 30-day period and can be traded like any Select Sector fund. And that's why I'm recommending it be included in the Fidelity Selection Sector funds World Indices grouping. Check it out.

There's also a Fidelity 800 phone number (at the top of that above link) you can call for questions since my information isn't that scrapable for automated updating. But Fidelity changes this information so rarely that automated updating would be an overkill.

QUOTE:
hardly “useless” to include discontinued funds ... to prevent survivorship bias.

I wouldn't simulate four years back in time for mutual fund models, so a fund discontinued four years ago is useless. And including discontinued funds may slant your simulation in more harmful ways by ignoring current sentiment. For example, the Biden presidency creates different market pressures than the Trump presidency.

You wouldn’t, but I (and perhaps others would. I would want to see the effects of survivorship bias eliminated in as many cases as possible. But nothing is preventing you from just creating your own DataSet that does not include the symbol.

@superticker

Thanks for the walkthrough. 8 constituents are gone compared to our DataSet and 4 new funds are in.
Not sure if converting this into a dynamic DataSet with arbitrary rebalancing date would be worth the effort. So we can simply add the new ones to the DataSet but retain the removed symbols.

Sounds good, we wouldn't want to introduce any unexpected bias by removing the defunct funds, it's good to keep them for a true historical perspective.

QUOTE:
8 constituents are gone compared to our DataSet and 4 new funds are in.

Interesting. I didn't know that. You might want to call that 800 Fidelity number (on the website link) to confirm. And the consultant might provide you with a better (scrapable) web link. Post it if they do.

I can't imagine 8 funds are gone because when I run my Select Sector funds, I get 40 "active" ones (counting FRESX). If I lost 8 funds, I don't think that would happen.

QUOTE:
Why Fidelity sector funds [?]
* Over 40 actively managed sector and industry funds

That agrees with what I expect. There's about 40 active Fidelity Select Sector funds. They all employ the "efficient frontier" Modern Portfolio Theory (MPT) method of active management to minimize turnover.

According to the Fidelity web site there are 48 (active) Sector Funds:

CODE:

Fidelity Select Semiconductors Portfolio (FSELX)
Fidelity Select Technology Portfolio (FSPTX)
Fidelity Select Automotive Portfolio (FSAVX)
Fidelity Select Communication Services Portfolio (FBMPX)
Fidelity Select Software and IT Services Portfolio (FSCSX)
Fidelity Select Construction and Housing Portfolio (FSHOX)
Fidelity Select Medical Technology and Devices Portfolio (FSMEX)
Fidelity Select Computers Portfolio (FDCPX)
Fidelity Select Brokerage and Investment Management Portfolio (FSLBX)
Fidelity Select Wireless Portfolio (FWRLX)
Fidelity Select Retailing Portfolio (FSRPX)
Fidelity Select Gold Portfolio (FSAGX)
Fidelity Select IT Services Portfolio (FBSOX)
Fidelity Select Consumer Discretionary Portfolio (FSCPX)
Fidelity Environment and Alternative Energy Fund (FSLEX)
Fidelity Select Leisure Portfolio (FDLSX)
Fidelity Select Health Care Portfolio (FSPHX)
Fidelity Select Consumer Finance Portfolio (FSVLX)
Fidelity Select Pharmaceuticals Portfolio (FPHAX)
Fidelity Select Financial Services Portfolio (FIDSX)
Fidelity Select Insurance Portfolio (FSPCX)
Fidelity Select Biotechnology Portfolio (FBIOX)
Fidelity International Real Estate Fund (FIREX)
Fidelity Select Consumer Staples Portfolio (FDFAX)
Fidelity Select Health Care Services Portfolio (FSHCX)
Fidelity Select Telecommunications Portfolio (FSTCX)
Fidelity Real Estate Investment Portfolio (FRESX)
Fidelity Select Industrials Portfolio (FCYIX)
Fidelity Select Utilities Portfolio (FSUTX)
Fidelity Select Materials Portfolio (FSDPX)
Fidelity Real Estate Income Fund (FRIFX)
Fidelity Select Transportation Portfolio (FSRFX)
Fidelity Real Estate Index Fund (FSRNX)
Fidelity Global Commodity Stock Fund (FFGCX)
Fidelity Telecom and Utilities Fund (FIUIX)
Fidelity Select Banking Portfolio (FSRBX)
Fidelity Select Chemicals Portfolio (FSCHX)
Fidelity Select Defense and Aerospace Portfolio (FSDAX)
Fidelity Natural Resources Fund (FNARX)
Fidelity Select Energy Portfolio (FSENX)
Fidelity Agricultural Productivity Fund (FARMX)
Fidelity Disruptive Automation Fund (FBOTX)
Fidelity Disruptive Communications Fund (FNETX)
Fidelity Disruptive Finance Fund (FNTEX)
Fidelity Disruptive Medicine Fund (FMEDX)
Fidelity Disruptive Technology Fund (FTEKX)
Fidelity Infrastructure Fund (FNSTX)
Fidelity Water Sustainability Fund (FLOWX)

QUOTE:
there are 48 (active) Sector Funds

Thanks for posting that list! But ...

1) Are all part of the "Select" Sector Fidelity offering?
2) Do all have a 30-day minimum retention time? Some may be 90-days, instead.
3) Are all managed using the "efficient frontier" MPT method to reduce position turnover (which would have tax consequences)?

If they are really part of the Select Sector offering, then the answer should be "yes" to all three questions.

---
It might be nice to have a public ETF sector dataset too (with no minimum retention time), but that's another topic.

QUOTE:
8 constituents are gone compared to our DataSet and 4 new funds are in.

I stand corrected: the 4 aren't Select funds.

Stay tuned for updates.

Got my hands on some web magic technology which lets me restore the Select Sector portfolio's historical composition over time. Mark is right, Fidelity makes changes to it at a minimal pace: for example, only twice in the last five years.

Build 24 will convert this Fidelity Select Sector funds DataSet into a dynamic one, with additions and deletions of constituents taking place roughly on an annual basis. We'll get the estimated dates hardcoded for now, and any future changes will not be tracked by our index composition bot.

Please do not include nonexistent funds like FSCGX in the Fidelity Select Sector funds dataset.

In addition, PreExecute fails when it hits a nonexistent fund. The statement

CODE:
public override void PreExecute(DateTime dt, List<BarHistory> participants)

Returns a "1" for participants.Count when I'm expecting "39" when it tries to process FSCGX. To some degree that's expected because the Data Range is set to the 400 most recent Daily bars which starts at 5/7/2020 for "existing" funds. But for FSCGX, this starts at 2/3/2016. Since it's the only fund at 2/3/2016, participants.Count=1 when PreExcute encounters FSCGX (which crashes my strategy).

Maybe the 400 most recent Daily bars Data Range should be translated to a specific date from the present so participants.Count doesn't get messed up (in PreExecute) for stocks with no recent data like FSCGX.

I'm wondering if other users are having problems with their PreExecute implementation when dealing with stocks with no recent data?

Commenting on just this one:

QUOTE:
Please do not include nonexistent funds like FSCGX in the Fidelity Select Sector funds dataset.

We decided to keep FSCGX in Fidelity Select Sector Funds, turning it into a dynamic DataSet. Consequently, said fund does not appear in backtests after 2018/1/26.

Use most recent 80 weeks instead of 400 bars.

QUOTE:
Use most recent 80 weeks instead of 400 bars.

What you're saying is that using an "N bar" Data Range is no longer possible. If that's true, then why is it being offered as a Data Range choice in the first place?

The right solution is convert Data Range "N bars" to a "real date", and then set the startbar to that real-date bar internally. Then all the stocks--valid and dead--will synchronize correctly. As it is now, dead stocks don't synchronize correctly, which is a problem for PreExecute/PostExecute procedures.

As more users create PreExecute procedures, they are going to run into this dead-stock problem with N-bar Data Ranges. WL needs to be designed robust enough to avoid this dead-stock problem automatically.

--
Obsoleting "N bars" as a Data Range is not a real solution.

There is a complication because certain DataSets like ASCII or Metastock have files that contain only old data. We've run into other problems with trying to tie Most Recent N Bars to the current date. In these cases, these ASCII files then fail to load data and it leads to confusion for another reason.

Most Recent N Bars will continue to do just what it says on the tin, and load the most recent B Bars of the symbol, even if it's outdated.

You can use Most Recent N Weeks to compensate, and we can also introduce a new Recent N Days that would work on calendar days which could help also, if you don't want to conver to weeks.

QUOTE:
We've run into other problems with trying to tie Most Recent N Bars to the current date.

Thanks for your explanation. I didn't realize there were complications. And I don't understand why switching to weeks fixes the problem.

QUOTE:
we can also introduce a new Recent N Days that would work on calendar days which could help

Well, I like uniformity to reduce confusion. I'm now helpless to suggest a solution because I don't understand all the complexities involved. But whatever you do, pick a solution that is as uniform as possible.

I'm going to create a new dataset without dead stocks in it. The dead stocks create synchronizing problems with the Pre/PostExecute routines in my strategies. I would be interested in knowing how others using Pre/PostExecute deal with this problem, but that's another topic. Maybe there could be a strategy option to disregard dead stocks (at least in the Pre/PostExecute routines).

The reason it works for weeks (and the suggested new "days" option) is that these are tied to a calendar, and it calculates start and end dates. Most Recent N Bars goes purely by whatever data is in the file.

I've rebalanced the list of select funds for B44, deactivating 6 funds in backtests spanning 2022 and up because they're no longer found on Fidelity's page.

Just so you know, Fidelity has recently change the description of FDCPX from Fidelity Select Computers Portfolio to Fidelity Select Tech Hardware Portfolio. However, all the third-party data providers (such as Yahoo) are still using the old name at this time, 1/1/2022. The fund itself hasn't changed really.

Just an update:

No changes to the Fidelity Select DataSet in 2022 according to my tracking.

Heads-up:

Symbol FIUIX will be deleted from our DataSet in B67 for it's never been a Select Sector ETF. My bad.

Other than that, no changes made (by Fidelity).

QUOTE:
Symbol FIUIX will be deleted from our DataSet in B67 for it's never been a Select Sector ETF.

These are managed mutual funds, not ETFs. Technically, FIUIX is not a Fidelity Sector fund, but it trades like one with a 30-day minimum ownership (without penalty), so I include it in my Select Sector fund dataset anyway. The real Fidelity Select Sector utility fund is FSUTX. Both of these utility funds perform similarly.

The only other change is the FBMPX fund, which was called the Fidelity Select Sector multimedia fund, is now called Fidelity Select communication services fund. And Yahoo knows about this name change, but unless you delete and recreate it in the WL Data Manager, it will continue to appear with the old name.