I've gone over some older posts from people sharing their reporting stats, and I think a reasonable guess for how the accuracy is being calculated (in PHP), is something like:
$accuracy = ceil($good / ($good + $bad) * 100);
Hazarding a guess on top of a guess, I'd say that theymos probably likes doing it this way (that is, using
ceil rather than
round) because it biases things toward more people feeling better about their stats (probably he's trying to avoid a situation where too many people get into score-maintenance mode and do less reporting than they otherwise might).
If someone wanted to calculate things differently, I suppose they could do it from a userscript or something. For example, here's what my own "Report to moderator" page looks like:
But, if I hit
Ctrl+Shift+K (I'm in Tor Browser on Linux, BTW), and I execute the following JS:
(function() {
const div = document.querySelector('tr.windowbg > td > div');
const info = div && div.innerText.match(/ \d+% accuracy \((\d+) good, (\d+) bad,/);
if(info && info.length == 3) {
const good = parseInt(info[1]);
const bad = parseInt(info[2]);
const accuracy = (good / (good + bad) * 100).toFixed(2);
div.innerHTML = div.innerHTML.replace(/\d+% accuracy/, accuracy.concat('% accuracy'));
}
})();
Then my page looks like:
If anyone wants to embed or adapt the above JS into their project, then, please feel free, but, be aware of two things:
(1) I wrote it very quickly and haven't tested it much (and, even if I had written it carefully and then tested it thoroughly, my JS is often unidiomatic).
(2) Don't trim the leading space from the first regex unless you understand why it's there (that is, if you do remove it, then you should probably find a different way to achieve idempotence).
Hmm... I should have done this beforehand, but, after writing the above, I
searched BitList for things that theymos might have said about reporting, and found this:
Since so many people have so many reports, I changed the percentage on the reports page so that it uses proper rounding instead of ceil() and shows 1 decimal. Actually, I changed my mind. This would probably make people feel subconsciously paranoid about reporting.
I also bumped into OP's reply to that post:
So, I think we can consider this question to have been "asked and answered" (a long time ago).
One final thought: I don't think many people realize the
extent to which theymos believes that users shouldn't lose any sleep over their accuracy. For example, I thought I had internalized the whole "one accurate report is worth many inaccurate reports" thing, but, I obviously hadn't, because bumping into the below quote was still something of a surprise to me:
Any accuracy percentage above ~25% is probably fine.
