Feedback request for series operations

[marco]

Dear All,

In a previous @geov post a problem was reported due to a recent change which we had to reverse out. The change was meant to catch misuses in the “series” parameter found in many technical functions. Users are doing stuff like this

EMA(2, 0, #SPYieldBlend - #CPI)

They want the EMA of the difference of two series (SP500 yield and CPI) so for the series parameter they used the difference of two series ids. What ends up happening is that the operation results in some other random id

53,666 (id for #SPYieldBlend) - 45,314 (id for #CPI) = 8,352 ( id for ??? )

I’m not sure what 8352 but it doesn’t matter. It’s nonsensical. You can check it yourself with this rule in the screener @s:#SPYieldBlend or @c:#CPI or @diff:(#SPYieldBlend - #CPI)

The original intent is very cool, and we want to support it because it cannot be easily done now.

This series enhancement is a wide ranging project that includes:

1. Making our “language” more robust
2. Adding series transformation functions
3. Adding series compatibility checks (like mixing weekly with daily data).

Do let us know what kind of operations you would like to be able to do with time series, whether benchmark, macro, imported, etc.

Thanks

[WalterW]

I terms of functionality, making a time series when the values are cross-sectional is something I’ve wanted to do often. See;

Calculating Industry Growth with Regression

Calculating an industry’s growth in sales or operating income should not start with the industry’s average or median sales or operating income. Instead, it has to start with the sales growth or operating income growth of each company in that industry, and then take the average or median of that. That’s how all industry growth factors are calculated, and it’s the standard way of looking at industry or sector growth. One can calculate the sales growth or operating income growth of a company using…

GetSeries accesses pre-built series. If it could be extended to building a series dynamically, that would be cool.

Maybe something that piggybacks on the Aggregate Series syntax;
GetUnivMedian(“Sector=Tech”,“OpIncTTM”)
so this would work;

EMA(20,0,GetUnivMedian("Sector=Tech","OpIncTTM"))>EMA(20,10,GetUnivMedian("Sector=Tech","OpIncTTM"))

But even that’s not dynamic enough since I don’t want to specify the sector/industry ahead of time. I want the sector to be implicit so that I can use the rule in the Buy (Sell) sections of a simulation.

I don’t know what that syntax would look like.