Aries Results

[markCBE]

<style> </style>

Case	Oil Volume
Exponential	1,512,131.75
Flat	365,000.04
Harmonic	1,985,771.54
Hyperbolic	1,950,108.18
Linear	547,500.02

TestCasesummary.xlsx

[grovesNL]

Thanks @markCBE! This is great. I started adding these results in #8.

I was able to match most of the cases by hand, but the harmonic results seem a bit different (still close, but different enough that it seems unexpected). The explanation I have so far is here https://github.com/pod2co/forecast-comparison/blob/f8da805394ddc8ddb512c5b3a2783d4c8ffb10fd/test-cases/04-harmonic-simple/README.md#aries Do you get the same result using HAR Desi instead of B/1 Desi?

Which ARIES version (5000 vs. 6) and engine (classic vs. new) were used? I don't expect it to matter much for this but I know the new engine uses 64-bit precision in some places while the classic engine always used 32-bit.

[grovesNL]

Just for fun I wanted to check the 32-bit calculation in Fortran in case there's anything interesting going on:

program main
    implicit none
    real(4) :: d, q
    d = (1.0 / (1.0 - 0.7) - 1.0) / 365.0
    q = (5000.0 / d) * alog(1.0 + d * 1825.0)
    write(*,*) q
end program main

This gives 1,985,840.25 though, which is closer to the hand-calculated value than ARIES.

[grovesNL]

Interestingly if you hand-calculate with an exponent of 0.9998 then you get ~1,985,771.376. I wonder if B/1 jumps down to 0.9998 to avoid an exponent of 1. This might just be coincidence but I wonder if HAR would give different results.

[grovesNL]

Confirmed that HAR gives the same result as B/1 (same as B/0.9998)