Beyond that, ++ votes seem to depend heavily on what it is that voters see and understand. Buried conversations don't get voted on because nobody sees them. Answers that people can see the value of get voted for. If you talked to the 42 (or more) people who voted for the cheap node but not the one you worked on, I'd bet that most of them never saw the one you worked on, and most of the rest didn't read the long discussion.

As for the +38 on the incoherent question, I would be interested in what the division is between different reasons. Many might vote because it is an easy way to use up votes (and hence gain XP). Others want to encourage people to ask questions. (Perhaps the topic or discussion was interesting despite the question?) Perhaps a few simply can't judge the quality of a question.

The 'buried conversation' anomaly is not that bad - if XP shows the utility of a node than a node that was interesting to many people should for sure have more XP than one read by only a couple of disputants.

Worse is the problem with nodes that are correct but too difficult for casual reader to judge their correctness - because it clearly is against the notion of XP as a utility measure.

Such usually doesn't surprise me much. You can only get one vote per node per voter, so +45 vs. +3 means 45 people got at least a little something out of the first node, not that even a single person found the first node better than the second.

The shorter, faster, and simpler a node, the greater the number of potential voters who can appreciate it (slow nodes are seen by fewer, long nodes are often skimmed or skipped, complex nodes are often not fully understood). Perhaps even more significant is that long/complex nodes just take long enough to process that the impulse to vote gets lost (and are more likely to give the reader some minor point they don't like and so give them reservations about upvoting it).

Or perhaps I just see patterns that make sense to me even when there are none. (:

Luckily, intentionally attempting to take advantage of this quirk usually "rings false" and doesn't work very well for very long, while most authors tire of such XP whoring anyway, so the consequences aren't horrid.

Well, I know I've personally been capricious in voting for your nodes, Abigail-II, as well as for merlyn's and some other prolific saints.

It basically comes to that if I've already upvoted your nodes three times in a day, I figure I'll save some votes and spread the wealth elsewhere. And this has resulted in some cases in which I've upvoted "stand on your foot" articles, and failed to upvote articles that clearly took much greater effort just because I saw the former sooner (or ran out of votes for the day.)

If I had unlimited votes or could apply multiple votes to a single article, I'm sure my behavior would be different (and this shouldn't be mistaken as arguing for either of those things.)

Clearly, I could take notes throughout a given day's reading of Perlmonks to figure out what are the 20 nodes whose reputation I feel most strongly about modifying, and then vote accordingly at the end of the day.

But I don't... I vote as I read, which pretty much guarantees there's going to be a capricous element to it.