Gendalph

Done

Have you read this one?

Ore to Kawasu-san no Isekai Houroki

I'd highly reccomend toaru ossan no vrmmo katsudouki

these ones aren't quite as amazing to me, but still very good

https://vatoto.com/comic/_/comics/how-i-stalked-some-dude-with-an-exposed-nipple-and-stumbled-upon-the-zenithian-sword-r1319

http://vatoto.com/comic/_/comics/yureka-r101

-edit-

I take it back. The latest chapters of 'How i stalked a dude...etc.' are awesome.

http://vatoto.com/comic/_/comics/onepunch-man-r7139#commentsStart

A whole bunch of people like Meow Zorg are posting open spoilers for the next few chapters.

To make things easier, these comments are the current comments with spoilers:

Today, 03:14 AM

Done.

My main concern with Batoto's current rating is the lack of dynamic range. Almost everything I look at has a rating of 4.5 +/- 0.25. Effectively the raters are using ~ 1/10 of the dynamic range of the rating scale. (Or I only look at ratings for good-to-very good comics.)

Wilson's lower confidence interval penalizes ratings with sparse info.--making it partially redundant with a display of the number of ratings given. It does expand the useful dynamic range, but only in the downward direction. I can see where a Bernoulli distribution might help derive an approximate null hypothesis for intervals around up-down rating averages. (See Evan Miller page.) Apart from which approximate intervals are most "correct" for binary ratings (e.g., up or down), extending Wilson's original logic for binary choices to multi-level ratings is never explicitly justified. An appropriate null hypothesis for arbitrary ratings would be a categorical distribution--using math derived from a multinomial distribution. This is a hoary old problem in psychometrics and I'm a little appalled to see web statistics just adapting Wilson's (1927) approach to averages of binary responses.

I think multilevel ratings by humans gain reliability faster than just scaling the Wilson minimum curve (red line above) to non-binary ratings implies. Also randomly sampled multinomial averages converge toward an asymptotic value more quickly than the formula implies. For 5 stars by the time there are 15 to 30 ratings, additional doublings of the sample size don't continue to narrow the sampling interval that much. For very "rating rich" environments like Reddit or Digg, it doesn't matter that much whether the Wilson's approximation to a binomial distribution is conservative. For Batoto, it might discourage the crucial steps of going from ratings based on 3 votes (unreliable) to 10 votes (starting to get reliable) to 30 votes (pretty stable).

As usual, the Bayesian approach is appealing in theory and with a somewhat problematic implementation. A straight Bayesian rating would give the site average rating to hentai/ero guro titles locked for unloading with no user ratings. That seems a bit . . . generous. The threshold value (min. # of ratings) appears to be chosen arbitrarily. Also the rate at which the ratings transition from the prior to the empirical average doesn't seem to founded in multinomial sampling statistics. The changeover should be a function of the number of integer steps in the rating system, when the number of steps is small.

-- Argh. The first couple of pages of Google results are answers for the wrong problem. Now I want to go do a proper literature search on this.

While visiting batoto.com today, I kept that developer console open that Firefox has. I think I ran into that evil ad that redirects the browser. While the developer console is open, it seems it won't redirect to that fake flash update page, but it still cleared the page and started doing something.

Sadly, the moment the browser gets redirected, the console gets wiped. I can't show exactly what happened last, what thing exactly did the redirection, but this is the new stuff that shows up in the console for the new page that starts to get loaded:

Empty string passed to getElementById(). ............. check.php:40
TypeError: document.getElementById(...) is null ............... check.php:40
GET http://data06.adlooxtracking.com/ads/image_visibilite2.php ................ [0ms]

-------

EDIT:

I found out how to keep the developer console from getting cleared when visiting a new page. I made a screenshot of how what happens for me looks like:

http://i.imgur.com/EcJKFTx.png

The old follows is still there too. http://vatoto.com/follows_comics

For those that liked the old follows much better, it's still there. Link on the right. But, I do plan on eventually removing the old interface with a merged interface that'll have features of both. Because... why have two?

SEARCH! Some people might have had the same problem you're having. It doesn't take a while to search at least two pages. You can also use [ctrl + f] to help.