Consider why the comic cites categorical reasons, not continuous ones.
Specifically, I can put a mean and a standard error down on polling, approval, and using a factor like incumbency calculate a probability of re-election based on a given approval or polling metric.
Polling and approval data, is something at least hypothetically ‘exists’ for all candidates, ever, even if it went unmeasured.
And it does exist for these candidates. Don’s felony would fall within the bounds the premise of this comic, but not polling or approval. The relationship between polling, approval, and incumbancy doesn’t because we do actually have those information on those things. We can look at all presidents prior to now that we have data for, we can divide them into ‘re-elected’ and not ‘re-elected’, calculate a mean and standard error of their polling, and their approval, anything we can measure, and look at the probability of occurrence for the thing given their polling. We couldn’t actually do that with any of the factors in the XKCD because we’d be dividing by zero. We literally couldn’t create the statistic to get a probability distribution from because there are no examples of President running has parameter “thing B”, which is the actual point of the comic. “thing B” gets more and more ridiculous as the comic goes along.
Why the current example isn’t that case is that we do have examples of incumbents with low approval trying to be elected. The “thing B” about the incumbent exists and has been tested, so we can calculate the probability distribution.