"Unknown Indicator ID"

Some of the Time Series ID’s are returning “Unknown Indicator ID” (such as OIL, INFLEXP). Also, the documentation is incomplete- (some are “obvious to a casual observer from a cursory inspection” as we used to say at LTI, others not so much, such as “UNTEEN,” “UNWANT,” etc. )

Sorry blanking out. Where do you see these problems? (our ids start with # or ##)

##OIL, ##INFLEXP, for example.

ok. in what context?

in the screener these rules seem ok

showvar(@oil_0,close(0,##oil))
showvar(@oil_20,close(20,##oil))

I have a Formula defined:
ema(2,0,PITSeries(##PPI))>ema(20,0,PITSeries(##PPI))
that works fine, but
ema(2,0,PITSeries(##OIL))>ema(20,0,PITSeries(##OIL))
doesn’t work with error:
Unknown indicator ID 51981

I find that ema(2,0,##OIL)>ema(20,0,##OIL) without the PITSeries works though…

(All done in sims, not screens)

OK I see now. This is an embarrassing and confusing solution we offered for series that have future revisions. It will be corrected (see below)

Turns out ##OIL is not a PIT series (Point In Time series), past values are never revised. Therefore you should NOT use PITSeries to specify it. Just use ##OIL.

But other series are PIT series, like ##CPI. The only way to know if a series has revision is in the description . For ##CPI you will see the following “Can be used in conjunction with PITSeries function”. And if you do not use PITseries(##CPI) and just use ##CPI you will get a fully revised value when evaluated in the past (which has no real value and is confusing)

Sorry about this confusing mess. Saying ema(2,0,##CPI) should just do the right thing without you needing to think about whether you should use PITSeries. In other words we’ll be deprecating PITSeries completely. It will become a no-op so it won’t break any code.

But for now please be mindful of which series have revisions by looking for this in the documentation of the series “Can be used in conjunction with PITSeries function”

Uhm, just read the original post about why PITSeries was introduced https://www.portfolio123.com/mvnforum/viewthread_thread,12374

Apparently the PIT data for some series is not clean, so the fully adjusted data is preferable in the past. BTW, ## series come all from FRED. Some get revised, some do not.

But it’s all very confusing with current documentation. We’ll try to figure out a way to have it clear in the documentation

Thanks-

Sorry. scratch what I said. This is NOT correct:

This is also not correct from the original post:

The point in time series does not “jump” in a backtest. It’s always safe, and recommended, to use PITSeries(##cpi) when the series has revisions. The data appears jumpy when you see it in the fundamental chart because the fundamental charts is all point in time. See the attached image

The jumps in the point in time RED line are not really there in a backtest. RGDP is revised every 5 years or so and FRED restates previous data. But when you access it in a backtest you will always see a smooth line (except covid years of course bc the jumps were real). The YELLOW line is the fully revised RGDP as of today to account for inflation

On Aug 27 2011 the point in time RGDP was 13,260 , but today that value is 15,870 to account for inflation.

Hope this clarifies a bit. Long story short for now:

  1. In a backtest always use PITSeries(seriesid) if possible . There are no “jumps” or “pitfalls”.
  2. Refer to the documentation as to which series can be used with PITSeries()
  3. We will try to deprecate PITSeries() , it is unnecessary. The backtest should use the right series by default.
    Sorry. again.