Emil O W Kirkegaard

In many twin studies, broad heritability of IQ is .7-.8. If polygenic scores can capture almost all of that heritability, their prediction of IQ won't be perfect, but will very good indeed. Obfuscation abounds, but if those predictions work, I don't see how one gets around them

The polygenic scores don't even come close to explaining as much variance. They still only predict less than 10% of education. All the current hype is a marketing campaign by genetic testing companies.

Of course they don't come anywhere near capturing that much variance -- today. We have no compelling reason to believe, today, that they won't in the future do so -- though perhaps rare variants may pose a real problem.

Rare variants can certainly be captured in principle by whole genomes. But can they be identified as affecting a trait if they are exceedingly rare and the effect is relatively small? That sounds like a hard statistical problem, even with huge datasets.