The conclusion should probably be that redis is super slow, not that the python 3 implementation is particularly fast. Also it is unclear how python dictionaries would scale for large number of keys compared to redis.
That sounds like the definition of confirmation bias.
How is that substantiated?