Solar Thermal Gains in HB / EnergyPlus - discrepancy between polygonclipping and pixelcounting methods

I am concerned about the differences that one obtains while looking at Transmitted Solar Radiation when changing parameters the Shadow Calculation module, particularly between PolygonClipping and PixelCounting methods, with very small impact of solar distribution methods (being FullExteriorWithReflections or FullInteriorAndExteriorWithReflections).
While it is useful to know how one method may be more suitable - from a computational perspective - to tackle complex shading, which one leads to more accurate results? Such trade-off appears much clearer when it comes to Radiance parameters for instance, but not on this matter.

I am pasting results from a shoebox analysis (shading elements in transparent purple).

Results with simple horizontal overhang:


Results with simple horizontal overhang + external louvers:


@chris making reference to our past conversation on radiation analysis, many of the LBT users probably realise how solar radiation analyses - per se - are very useful when relative figures are considered but risk becoming pointless when the linkage to solar thermal gains is not clear.
So even the availability of multiple ways to calculate solar radiation externally risks become a vanity project unless this is linked clearly to the calculation methods and results that can be obtained by means of thermal / energy modelling. I may be wrong in this, and I immensely appreciate the work you guys have been doing, so please let me know if there is something I am missing here.

Hey @andrea.botti ,

This type of question is better asked on the UnmetHours forum where the EnergyPlus developers can respond:

Given that the PixelCounting method is relatively new to EnergyPlus, I’m sure there are people there with fresh knowledge about it who could answer the question.

I imagine that the accuracy of the PixelCounting method is limited by the number of cast rays. So there might be a means of dialing this up to get better accuracy.

It would also be helpful to see this compared to what the Annual Irradiance recipe reports when high-accuracy parameters are used since that’s the closest thing we’ll probably get to a “ground truth” version of this case.