Confusing MAPE distribution comparision in Model Studio

[whymath]

I am learning the lesson "Forecasting Using Model Studio in SAS® Viya®" from learn.sas.com and trying to repeat the demo "Pipeline Comparision".

I think the result of comparision is very confusing.

Before comparision, the distribution graph of MAPE of "Pipeline1" and "Honest Assessment Auto" are different:

But after I select both 2 pipelines and click Compare button, the distribution graph of MAPE become different and difficult to explain:

Why will this happen? I think the distribution graph of MAPE should be same before and after click Compare button. 

[sbxkoenk · Accepted solution]

I suspect that — when comparing the pipelines — the horizontal axis is scaled identically for all pipelines (i.e. for all champion models within the pipelines). Check for yourself ... if you only look at pipeline 1, you have a horizontal axis with a much smaller range. The x-axis end point for pipeline 1 — when only looking at pipeline 1 — is below 100 (around 35), while for the comparison it is above 100 (around 125).
Because the number of bars (or the bar width) is calculated based on the range, you fall back from multiple bars to a single bar.
So there is definitely no error. But perhaps it could be improved visually?

Ciao,

Koen