But how can advice for winning a game be either bad or good if you will win whatever you do? That's the part I don't understand.
The advice was "good advice" in the sense of "if you had followed the advice, you would have found the item before the taking-pity-on-you time limit." But it's not "good advice" in the sense of it having any relation to where the item was ultimately found—and therefore it can't be inferred to have been good advice just from the evidence the child has at that point.
On the other hand, if the game is played iteratively (like an iterated Prisoner's Dilemma), it would eventually become clear (presuming the child managed to find the item by themselves some of the time) that there is this "taking-pity-on-you time limit", and that some of the advice has a correlation to where the item is before that time limit. The child would likely intuit the "taking-pity-on-you" moment as the true loss condition of the game, and become upset when it happened, just as if they had been told they lost.
But yes, it's hard to determine whether this is how the experiment worked from the article. Unfortunately the actual paper at http://www.sciencedirect.com/science/article/pii/S0022096516... is paywalled in the usual way.
Searching for
"title of article" PDF
There was actually 4 similar studies described in the article. But the first one is the most important.
The setup was that the kid would watch a video where a toy was hidden in one of two cups. The helpful person would point out to the kid in which cup the toy was hidden. The unhelpful person would not give any hints at all but rather just explain which toy was hidden. The kid would of course follow the helpful persons advice, but would have to guess in void of any advice.
The kid would then get to guess in which cup the toy was hidden, whereafter the answer was revealed in the video. However, in the successful condition, the answer was modified so the kid would guess right no matter what.
This was repeated eight times, after which the kid got to choose which person would give them advice for four more trials.
It was in this situation it showed that the kids who had always been successful did not seem to value the better advice they would expect to get from the helpful person while the kids who had failed due to lack of advice chose the helpful person.
In the article they discuss this being due to the kid trusting their own ability to guess, not needing any help.
I'd say that the results are academically interesting as they increase our understanding of children's decision making. But I don't think they should be taken out of context and be applied as a some kind of answer on how to do parenting and teaching.