X^2 for Aggregate Freq Data

[Bühlmann]

Loss Models discusses a modification of the chi-square for aggregate frequency data (pp. 435 - 436). The key being you divide by the variance (instead of "expected") and your degrees of freedom is the number of intervals minus the number of estimated parameters (note no "minus 1").
My question is how important is this topic and what are the odds of it being asked? It may not be too hard to do, but if it is really not that important, then why cloud my brain space?
There are no released exam questions on it, but the discussion does take up 1.5 pages of LM. I saw it in the ASM manual as well and I don't recall the Mahler notes going over it.

[inexactuary]

I think it is unlikely to be asked, but it is easy enough to remember if you think of it as adding up the squares of "z values" (observed minus mean divided by sd). You don't subtract one from the df because the sample size isn't fixed. That said, your time might be better spent honing your skills on more important topics.

[JohnGalt]

I wouldn't ignore it. The calculation isn't any more difficult than the usual chi-square test, so I think it's a reasonable exam question.

[NewTubaBoy]

I disagree. I think that you can ignore it.