Come for the quick hacks, stay for the epiphanies. PerlMonks

### Signal-to-Noise Ratio

by spudzeppelin (Pilgrim)
 on Oct 03, 2000 at 19:32 UTC Need Help??

And I'm going to disclaim immediately any connection between the term "noise" and XP gained from logins and voting. At any rate, the discussion occurred in the chatterbox this morning (princepawn had a large hand in it) regarding the "signal" XP (gained from writeups) vs. "noise" XP (everything else) -- so I suppose this could be taken in a context where decibel levels are ordered similarly to /. rep scores.

The basic SNR formula is R = 20 * log (signal/(total - signal)). So, for example, if your rep was 94, and 54 of those were from writeups, your ratio would be:
20 * log (54/(94-54)) = 20 * log (1.35) = 20 * .1303 = 2.606 dB.

Note that the above assumes common logs; to use natural logs (a la perl), you need to change the 20 to an 8.68589. Which leads to our trivial code sample:

```#!/path/to/perl
use strict;
if (\$#ARGV != 2) { die "Usage: progname <total XP> <writeup XP>\n"; }
my \$txp = \$ARGV[0];
my \$wxp = \$ARGV[1];
\$txp =~ tr/0-9//cd;
\$wxp =~ tr/0-9//cd;
if ((length(\$txp) < 1) || (length(\$wxp) <1) || (\$wxp == 0) || (\$wxp ==
+ \$txp)) { die "Unable to compute value!\n"; }
my \$snr = 8.68589 * (log(\$wxp/(\$txp-\$wxp)));
print "Signal-to-Noise ratio is \$snr dB\n";
exit 0;

Spud Zeppelin * spud@spudzeppelin.com

Replies are listed 'Best First'.
RE: Signal-to-Noise Ratio
by BlaisePascal (Monk) on Oct 03, 2000 at 21:44 UTC
S/N ratio might be an interesting statistic to monitor... but to be honest, I don't know where my XP comes from. All I know is that occasionally the XP nodelet tells me I've gained some XP, it doesn't tell me where it came from. Is that info even available?

I assume that you mean XP when you say "rep"... how do "I" get rep, as opposed from my nodes getting rep, and how do I get rep from things other than writeups? XP I can see, but rep?

But it occurs to me... that S/N formula looks like it's based on the intensity of the signal, not the power (which would be 10*log(...)). Which would make more sense to use?

XP is what users get, reputation is what articles get.

XP can be gained (or lost) from having a change in the reputation of an article. There's a complex set of rules that define the odds of whether you gain or lose XP based on how the article was voted. See Voting/Experience System.

Reputation is what articles get. Based on how people feel about articles (and in theory NOT the person writing it, but being people, there are some with political motivation. The practice is supposed to be 'vote the articles, not the person'), they may vote them up, vote them down, or not vote them at all.

By clicking the writeups number on your homenode, you will be presented with a list of articles you have written, and the reputation of each article. Ideally, you'd like the reputation of any of your articles to be above the \$NORM value, described at Voting/Experience System. An article that is lower is ultimately responsible for dragging down the \$NORM value, affecting everyone. It sorta indicates that the article is substandard, ASSUMING that everyone has seen it. In reality, if an article is nested too far down, it may not show up on the page, and will be overlooked/ignored.

A good place to look is site how to for information on the site. (At the moment, it seems someone has let the font tags get carried away, and it's a little harder than it should be to read.)

There have been several theories as to what represents a good indicator of a person's standing in the monastery. Overall XP is a little more meaningful than it used to be, but still can misrepresent a person. In the olden days (what, 3 months ago?), people were given XP for 'voting out', or using all your votes in a day. This (IIRC) was 25% of the number of votes you had. So, if you had 20 votes, you were guaranteed 5 XP just for using all your votes, plus any random XP gained for voting (as according to the rules at Voting/Experience System.

This meant that there were (and still are) some people with very high XP, but virtually no articles. Thus, dividing thier XP by the number of articles written is not representative of how much that user contributed (or, more aptly, didn't contribute) to the site. A more accurate number is the total sum of reputation of your articles, divided by the number of articles.

Obviously, if you have 100 articles, and each article has a reputation of 1, you're not doing too well. On the other hand, someone like Erudil, who has 5 articles, but has a combined article reputation of 262 (as of this node being posted), has an extremely high average (52.40). His articles are well thought of, which of course, is always desirable. If you have a number of articles, you can use luke_repwalker.pl to calculate your totals, rather than adding and dividing by hand (and no, this isn't a shameless self-promotion for something I wrote. Well, OK, maybe it is...) There's also a cool XML version here

Now, does this mean you shouldn't post if you article won't be thought highly of? I'm of the opinion that 'thank you and me too' articles, unless solicited, are not of much value, and tend to pollute the article-space (also known as 'bozon articles'. heh). I don't think they add much to the overall information base. And if you really want to thank someone, send them a /msg. It's more personal, anyway. But this doesn't mean you shouldn't. Maybe it's just a followup note to clarify something. It may not gain any reputation, but it does enhance the informational content of the site, for someone browsing the topic. So you SHOULD post something like that.

There's also the 'XP Whores'. These are people who post articles or vote as much as possible for the sole purpose of gaining XP, but without making any real contributions to the site. It's a bad thing to be called an 'XP Whore'.

These, of course, are my opinions, and don't represent the opinion of the site as a whole, vroom, or the The Everything Development Company

--Chris

e-mail jcwren
I am but a lowly monk, with much more to learn here than to teach. But your post makes me wonder.

Is it really desireable to reduce the site's content to an information base? After all, if it's to be a community, a certain amount of non-informational content would seem not only to be tolerable, but necessary.

I think "thank you" posts are a good example. Having a real community requires a degree of civility, and on the Internet that's a scarce commodity. So why not /msg your thanks? Because then everyone doesn't see it. Gratuitous posts of gratitude establish a tone, and establish for everyone an obvious standard of courtesy. Saying thank you, from a practical standpoint, is a waste of time in real life, just as it's a waste of space and bandwidth here. But it makes a difference; I'm guessing that difference could be increasingly important as many new monks arrive.

jcwren: "Ideally, you'd like the reputation of any of your articles to be above the \$NORM value"

Now there's something I don't think is of much value... an impossible ideal. I realize you surely aren't suggesting that everyone should be above average, but is dragging down \$NORM even something with which posters should be concerned? It seems to me that the voting/experience system is supposed to be modeled after the nature of conversation on such a forum, helping to improve the quality. Modeling our conversation after the voting/experience system does not seem like a good idea to me.

I would suggest a different ideal. If your post has a positive reputation, that's good. It means that at least someone thought well of it, and that more people thought well of it than poorly. If it's higher, that's better. You're making people happy.

Because of the technical nature of the site, you're not likely to feel compelled to say something unpopular because you feel it's true. Nevertheless, this is a good example of why an ideal which, at its root, is strictly majoritarian may not be a good idea. The voting system can help improve content, but can't replace conscience and good intent.

jcwren:"This meant that there were (and still are) some people with very high XP, but virtually no articles. Thus, dividing thier XP by the number of articles written is not representative of how much that user contributed (or, more aptly, didn't contribute) to the site."

If by voting you make no contribution to the site, then it seems clear that you should get no XP for voting. Perhaps this is what you mean. But if you assume that it is useful and good that posts should be popularly rated, then people who simply vote are contributing to the site (assuming they do so responsibly). Perhaps the relative weights of voting vs. posting don't reflect the real contributions made, but that's a different story.

Further, people who achieve higher levels are invited to participate in the maintenance of the site, editing faqlets, moderating content, and perhaps performing other chores I don't yet know of. In this way also people can contribute without posting... the average reputation of your posts just seems like too narrow a metric to me.

Improving the signal-to-noise ratio for those browsing a topic is very important; this is a technical site, after all. But the voting system does not seem designed to do this, but rather to act as a more general index of a post's contribution to the community.

I'm afraid I'm becoming long-winded, so I'll stop. I should say that I don't mean to imply that your views are opposite to everything I've said here, only to offer a different perspective. Having a single standard of virtue (epitomized by the very impressive Erudil) seems somehow un-Perlish.

That isn't what I meant... I'm familiar with the XP system.

According to the Voting/Experience System node, I get a 25% chance of getting 2xp for being logged each day. I get a 25% chance of getting 1xp for every vote I cast. These methods of getting XP are not related to my writeups or my rep.

When I see that I have more XP in the XP nodelet, it doesn't tell me this is because someone voted up a writeup I've made, or if it's because I was logged on 24 hours ago. True, I can guess that if I get a +1 after voting, it was because of the vote, but if I leave a good post at 10PM, and log back on at 8AM, and see that I have 5 more XP, was that 5 from the rep alone, or 3 from the rep and 2 from the "logged on within 24 hours" scheme?

I can't tell from looking at the XP nodelet, or from my homepage, which XP is from signal, and which XP is from noise, so I can't calculate my S/N ratio.

RE: Signal-to-Noise Ratio
by spudzeppelin (Pilgrim) on Oct 03, 2000 at 20:35 UTC

Oops, zejames pointed out that there is a typo in the "trivial code sample". \$#ARGV should be compared to 1, not 2. *g*.

Spud Zeppelin * spud@spudzeppelin.com

Create A New User
Domain Nodelet?
Node Status?
node history
Node Type: monkdiscuss [id://35123]
Approved by root
help
Chatterbox?
and the web crawler heard nothing...

How do I use this? | Other CB clients
Other Users?
Others meditating upon the Monastery: (2)
As of 2023-09-30 12:36 GMT
Sections?
Information?
Find Nodes?
Leftovers?
Voting Booth?

No recent polls found

Notices?