Peter's quite right, of course, but I think there's a bit more we can
say.  What makes a good pattern of reasoning good (logically good, that
is) is whether it preserves truth, that is, whether it only leads from
true premises to true conclusions and never from true premises to false
conclusions.  (If it starts with false premises, that's another matter
altogether.)  And the best way to tell whether an argument pattern will
be truth-preserving is to do a truth-table analysis of it:  Assume (that
is, make believe) that the premises are true, then figure out what the
truth values of the atomic propositions are, and, finally, figure out
what the truth value of the conclusion is.  If, whenever you assume that
the premises are true, it turns out that the conclusion has to be true,
then you know the argument is a logically good one; otherwise, it is a
"fallacy", i.e., a logically bad argument.  See any introductory logic
text for the details (I understand that Peter has a very nice one :-)

There are a couple of wrinkles with this:  First, some logically good
arguments are pretty bad because their premises are, in fact, false.
Those arguments are factually bad.  The best kind of argument is one
that is both logically good ("valid") and also factually good; if it's
both, then it's said to be "sound".

Second, some well-known fallacies are logically good!  The best example
of this is "begging the question" or "circular arguments".  They have
the form "p; therefore, p".  They may be logically good, but they are of
no practical use in convincing anyone of the truth of p.