[throwaway_4ever]

No because "smart" genes don't sustain. Like 50% is volatile noise. So that would actually line up. I.e. Two 150 IQ don't have a 150 IQ on avg. They have a 125 IQ, and then the two 125 IQ have a 112, and the two 112 IQ have a 106 and voila.

[yunohn]

I was being sarcastic to OP; but your math would imply that our ancestors were 1000 IQ and every generation decreases IQ further to 0.

[xyzzyz]

No, because this regression towards the mean tends to pull the children of below-mean parents up, just as much as it pulls the children of above-the-mean parents down.

That said, the grandparent's figure are rather wrong, for two reasons. First is that the heritability of IQ is typically estimated to be around 0.8 instead of 0.5, which means that the expected IQ of parents with IQ of 150 is 140, instead of 125. Second is that this is only the expected IQ. If they have multiple children, some will typically be above the expectation, and some below. More specifically, given standard deviation of 15, around a quarter of children of parents of 150 IQ will have IQ of 150 or higher.

[astrange]

That post is also wrong because "heritable" doesn't mean "caused by genetic differences".

[xyzzyz]

No, in fact it means exactly that. This word has a technical meaning.

[astrange]

Should get a different word then, since you can't get rid of confounders like prenatal diet and environment that aren't genetic but can cause you to be the same as your parents.

[xyzzyz]

In fact, you can. See, for example ACE model, which explicitly attempts to separate generic causation from shared environment and from non-shared environment. This can be done using twin studies, by comparing correlations between monozygotic vs dizygotic twins on various variables of interest, see Falconer’s formula for example.

This research is not new, it has been done for many decades now. The word “heritability” has a well established technical meaning.