This means you would like to predict success with or without intervention, and choose kids with the largest difference.

That's it. I think it's reasonable to expect that poor children who have had little resources spent on them would see a larger increase in performance that wealthy students who have already received a lot of resource on them.

However, you're assuming that education is viewed as an investment. This isn't necessarily the case.

True. The purpose may be to indoctrinate their minds, to crush souls, or to destroy futures. More likely the purpose of the educational bureaucracy is to enrich itself with an indifference to rival Cthulu and Clippy toward the harm they cause beyond that goal.

But you asked:

Should we try to make (other people's) children more successful?

Not being a moral objectivist, I didn't have a true morality to refer to, but I do have my own, so I provided an answer according to my own.

I haven't got the article here, but IIRC McClelland said something to the effect that it would be more fair to single out high-IQ students as kids not to spend money on.

Even more "fair" to single out smart kids to surgically regress to the mean, if the metric of fairness is equality. That's the problem with equality merchants - destroying value is probably the easiest way to achieve equality.