Composite node ranking direction

Hi

Just noticed a curiosity. Maybe I am being thick.
Anyway, so say you have a ranking node that is ‘value’. Ranked ‘higher is better’. Say it contains some value factors like P/, P/Sales, ranked lower is better.

If I invert the direction of the factor rankings, and also invert the direction of the composite ranking, I would expect to get the same result for a simulation but I do not by a long shot.

Basically I was looking into inverting a long ranking system for a shorting system, but was not sure if to invert the factors (and not the composite), or just invert the ranking direction of the composite (and leave the factors alone). I assumed these would be identical. I also assumed if you inverted both the factors and the composite, that the result would be unchanged. But this is not the case.

i.e. combining of rankings appears not to be commutative.

Maybe there is a simple reason for this but I was just curious.

It’s best to keep the arrows of the composites unchanged and flip the individual nodes. The reason is NA’s.

For example if one of your nodes is PE and for a long system you want low PE stocks and for a short system high PEs. Since a lot of stocks have NA for PE (50% or so) if you flip the composite node arrow you will be getting the higher score for the NA’s which is meaningless.

Short simming is next in our list and one of the tools we’ll add is a button to automatically flip all input nodes arrows.

Thanks Marco, that makes sense.
Looking forward to short sims, keep up the good work.