Quote:
Originally Posted by Trumgottist
No, I haven't given this much thought yet, but the AB, CD thing shouldn't be a problem. A ranks above B and C above D. There should be other pairs to sort out the rest.
|
Yes and no. I'm specifically thinking of a situation where A>B or C>D are mutually exclusive because of the already "locked" pairs. Let's say that (in a totally made up example),
Myst beats
Zork by two votes, and
Zork beats
Monkey Island also by two votes. However, it's already been determined at an earlier stage that
Monkey Island must rank above
Myst. Thus, we can only work either Myst>Zork or Zork>MI, but not both, into the existing list.
Of course, the algorithm could just pick the tied pairs in random order, but it's not very elegant, since it would mean different lists could be compiled from the same data. I'd prefer an objective criterion, even if a silly one. For example, the age difference. (The tied pairs where the difference between the release dates is the smallest have priority. I'd rationalize it this way: the comparisons between games more distant in time are less reliable, as nostalgia may unfairly sway the balance in favour of the older game, and technological advancements in favour of the newer one.)
Quote:
|
My initial reaction to the XY thing is to let those rankings cancel each other out and disregard the XY pair. If doing that leaves too little data to compile a list than so be it. Then we get no list.
|
Good point, probably. But now that I think of it, the same criterion (not necessarily the same which I proposed, but the same that would be used to fix the first problem) could be used here - we would only have to accept that the final list could also end up having some ties.
Sorry for overanalyzing things.
