Regarding the community vote, I tend to agree. However, to play devil's advocate, part of me thinks people watching a stream would place too much emphasis on map visuals. I'm also not sure that people watching a caster flicking around the map would be able to form a decent overall evaluation. I'd be interested to get other people's opinions on this point.
I' ll give you my opinion.
When watching a stream, I do tend to look at the visuals of a map, since I am there for entertainment and like to look at something interesting and pretty. Obvious issues with maps do become apparent pretty quickly though, because you are objectively watching the battle play out in front of your eyes without the concern of having to move your units around and all. Issues such as garrison placement, cover, resource point placement, base locations and for example pathing issues can be easily spotted whilst watching a twitch stream. However, there is the bias risk of thinking there is something wrong with the map, when in fact it could just be the player being bad or rng being heavily involved.
More detailed testing is needed to find all the issues with a map, especially little things such as that particular piece of cover that makes one side of the map less favorable or a strategic point more favorable to defend or approach from one side of the map or that annoying little bush that keeps screwing up pathing or one cut off having a more favorable garrison or none at all etc. Here lies the bread and butter of the map, the core if you will. This needs severe testing and can't be justified by watching a few Twitch streams.
Both ways have their benefits, but the latter way is the way to rely on more heavily when testing maps for automatching.