The webcomics blog about webcomics

So Simple That Only A Child Can Do It

It seems that a number of different websites have recently taken to using new math to figure out something that might seem at first, important – comic strip “ratings”.

No, I guess I don’t mean ratings – I mean “rankings”. That is, some numerical value arrived at through a supposedly secure mechanism, and supposedly calculated properly, that will indicate in some manner the relative quality of a particular webcomic.

Example 1 of this phenomenon is Top WebComics.

The entire purpose of this website is provide rankings. But it seems that they overlooked a small little security flaw in their methodology. And it strikes me as unlikely that SketchBattle is going to be the only one poisoning the pigeons in this park.

Example 2 is Smack Jeeves.

In addition to “proudly” hosting 13,016 comics (as of current count), they provide a number of different views into their ranking system – Top comics of the month, Top Strips of the Week, Top All-Time Strips, and Top All-Time Comics.

The first time I looked at SmackJeeves was about two weeks ago. The Number #1 Comic of All Time was a really really crappy sprite comic that had a total of 4 Ratings. Yep. 4. The comic “artist” had either signed up 4 phony accounts, or gotten his one friend to sign up a couple of phony accounts as well, and given himself the highest ratings he could.

A reasonable method of calculating rankings of comics on a site – based on user feedback – would be to INCLUDE the total number of rankings in your calculation. Otherwise, why bother letting more than one person rate a comic anyway? And then something sensible would occur – a comic that had 300 people decide that it was a good comic would always be rated better than a comic that had 2 people think it was a good comic.

The huge drama factor that is associated with the SmackJeeves ratings system, in their community, is thus at least an order of magnitude (that’s 10 times bigger for the calculationally challenged among you) more humorous given the major flaws in it. Last I looked there were at least four separate threads devoted to “Someone ranked me badly! WAAAAAAAA”.

Math – it’s not just for getting your G.E.D anymore.

Of course the problem with what you (rather sensibly) suggest is that, Oh GOSH mister! Math is just TOO HARD!!!

While I appreciate and understand completely the neccesity of what you present, the large majority of people slept through math class in high school, let alone college and where to find the required mathmatical formulas, let alone authoring them themsleves is beyond the scope of the common man. It doesn’t take much to outsmart the “no ranking till you have 50+ votes” type of system and asking most people to construct a mathmatical formula to calculate weighted values is simply too much, even something as simple as: number of raters/1000 * average rating (which would give greater weight to rankings the more rankings there were, meaning 500 of the highest votes would still give a crappy ranking while 2000 of the worst votes would still give a great ranking). God forbid we asked most people to come up with a system that assigns a bell curved weight to ratings.

Given the lack of enthusiam or ability for the average human for math and the lack of the average human for wanting to do any kind of research whatsoever, this is just an evil you’ll have to put up with. One can only hope that as people click on a link to read the “best webcomic of all time” that they return and give it the lowest score they can in retaliation for it’s over-hyped crappyness.

That’s all fine and dandy, Rick, but most people who are going to be coding sites that do rankings ought to have some decent programming experience. With that programming experience should come enough mathematical ability with which to make coming up with a suitable formula a reasonably easy task.

My main point is not that coders should be mathematicians – because I’m the former and not the latter.

My main point is – if you are making a website and you are putting in a rankings system… Then you should spend more than TEN SECONDS thinking about how to calculate a ranking properly.


You’re right, you are a coder.


You can hide your semantic ineptitude from the world, but I will always know the truth.

While I agree with the fact that coders SHOULD be able to do at least simple mathmatics (and most of the ones I’ve known don’t know as well as they should) the fact is that (in my experience) they don’t. The statement of that fact was proffered as a reason WHY bad ranking systems exist, not neccesarily as an excuse for it.

In fact, I WILL heartily agree with both you, Mr. Buchwald in that coders SHOULD be able to do a reasonable amount of math and YOU Mr. Lowrey, in that if you’re going to code a ranking system and you CAN’T do the math the least you could do is e-mail a coder who has sucessfully done it and beg some code.

I’ve also been thinking about the best formula for a ranking system and I’m convinced it has to be based on some kind of oddball curve like a hyperbole or a parabola or a statistical Z graph because otherwise you get too much weight adjustment at low or high quantities of votes. Alternately, it could just use multiple formulas based on specific input sets (from 1-x votes use THIS formula, for X+1 to X+Y use THIS one, etc….) but that seems more unweildy and prone to abuse than a single formula with a weighed value based on a curve.

Rick – maybe you need to use a Riemann zeta function.

The first problem with these is actually assuming the ratings have anything to do with quality. They’re a measure of popularity. The second problem is assuming the ratings have anything to do with reality. No matter how perfect the coding there’s still the human factor clicking on things, human beings that can and do lie.

So crank up those Heisenberg Compensators.

If these ratings were merely supposed to measure popularity, then they would merely count page views.

Also, I think if you follow my link on “quality”, you might find that I was making a joke about the assumption that these measure quality.

As for Heiseberng Compensators… I have mine hooked up to a cup of tea. Well, Advanced Tea Substitute.

I have officially decided that the rankings mean very little of anything. In general, they seem to indicate roughly which comics are best at persuading their fans to click a vote button as many times as possible.

Incidentally, as I get more readers to my comic (I gauge this in terms of direct request and bookmark visits per day), less people vote for me.

What does that tell you? I have no clue.

Sorry Jeff. You completely lost me. For “not being a mathmatician” you sure pulled a complex refference out. I looked up the Riemann Zeta function, but I admit I’m at a loss to see how it would help with ranking codings. Luckily for me, I suppose, I’m not planning on ranking anything anytime soon though if you have a link showing how to use the one to do the other I’d gladly take a look.

Anybody looking to implement a ranking system, keep in mind the long-cherished mathematics lesson: Plagiarize. Rotten Tomatoes has a nice weighting formula they use to determine best-reviewed movies in a given year.

Isn’t the Reimann Zeta function one of those million dollar math problems?

Anyway, you make a very valid point, but I ask the question, what is the reason for a rating system to begin with? I believe the answer is simply so that good comics get the exposure they deserve. In that case, what’s wrong with letting a comic with only a few ratings get some good exposure? If the rating isn’t true to that comic, people will see, and rate accordingly.

My comment about the Riemann Zeta function was another reference to Cryptonomicon, intended as a kind of joke.

The problem, Dan, is that – except for professional critics like Gary and I – nobody else is qualified to decide if a comic is ‘good’ or not. (this is a *joke*). An individual opinion that a comic is good doesn’t mean a lot – especially because it could clearly be the author themselves who is saying it’s good.

So the notion of including the total number of votes into the ranking is part of the whole notion of utilizing the collective stupidity of the internet.

Yes, new comics need exposure. And good new comics need exposure more than bad old comics need exposure. But listing a comic as one of the top ten of *ALL TIME*, because the chowderhead who made it got his friends to say it was good, is clearly wrong.

I was actually recently at one of the collectives you mentioned and saw this:
listed as the best webcomic of all time.

The description was intigueing so I checked it out. It truely IS awesome, but I doubt we can truely call a webcomic with less than a dozen pages “The best of all time” In addition to number of votes, a robust voting system might attempt to take into account the age of the comic and the number of issues, etc… While comics like the one above may indeed be masterful there’s simply not enough data for it to eliminate the possibility of it’s quality being an aberation.

I wonder if there’s some sort of market for this sort of thing. Given the propensity for the human condition to avoid math like the plague, it seems like there should be organizations willing to pay people to construct complex mathmatical ranking formulas that take into account multiple specified variables.

As to the joke: Sorry. Rest assured that my failure to get it lies solely in my n00bishness rather than your humor.

So, it’s a bunch of aspiring amatuer kids putting up their webcomics and cheating to make themselves ‘THE BEST!” — No one really gives a shit. I mean, anyone with half a mind knows something is up on those list sites when you look at the top ranked strips.

It will be HARD or take a real mathematician AND a coder to make something legitimate that you can TRUST. But hey, if it makes them feel AWESOMELY GREAT to be #1, they’re not fooling anyone else but themselves.

RSS feed for comments on this post.