Outside of formal math, all we have are models, so there will often (always?) be an ordering of "less wrong" on model-based knowledge.
So "wronger than wrong" is putting a heavy spin on the truth. More wrong models might still be useful.
Why do you think you are arguing against the concept? "Wronger than wrong" argues against believing everything is "equally wrong", which is precisely what your comment is also arguing against.
In other words, you agree with Asimov that equating all thing as equally wrong is "wronger than wrong", since there will always (or often) be an ordering of "less wrong".
I'm of the _opinion_ that this is the only way of knowledge itself: less wrong as time spent on understanding approaches infinity. Math and models are between language and tools, the fundamental essence is the same though.
Also, any predictive ability should ever be a side-effect, the core of inquiry is explanation. This is why a good-enough explanation of seasons on Earth is perfectly applicable to any other planet (in any solar system). And this is why ML models like AlphaGo are not helpful (you can't learn anything from them, including "how to be a better Go player").
In other words I stand with Deutsch, Popper and Everett on the interpretation of reality.