In the book How to Lie with Statistics (pp. 56-57), Huff (1954) gave an example of comparing IQ scores to illustrate the significance of probable errors.
[...] So Peter's indicated IQ might be more fully expressed as $98 \pm 3$ and Linda's as $101 \pm 3$.
This says that there is no more than an even chance that Peter's IQ falls anywhere between $95$ and $101$; it is just as likely that it is above or below that figure. Similarly Linda's has no better than a fifty-fifty probability of being within the range of $98$ to $104$. From this you can quickly see that there is one chance in four that Peter's IQ is really above $101$ and a similar chance that Linda's is below $98$. Then Peter is not inferior but superior, and by a margin of anywhere from three points up.
I do not see how "Peter is not inferior but superior, and by a margin of anywhere from three points up." Can someone explain?